SaaS has long celebrated a 3x LTV/CAC

This heuristic, first popularized by David Skok, has become the default for measuring the health of a SaaS company’s go-to-market engine. At around 3x, growth is considered efficient and sustainable. As the ratio falls below that level, businesses require progressively more capital to grow, making the economics less attractive.

In reality, consistently achieving a 3x LTV/CAC ratio is difficult, which is why it has become something of a “benchmark” for excellent SaaS businesses.

The calculation itself is straightforward:

As someone obsessed with AI Roll-ups, I was wondering how those businesses fare on the LTV/CAC metric?

To answer this question, we first need to adapt the formula to the reality of AI Roll-ups.

1. Re-defining LTV/CAC for AI Roll-ups

SaaS companies (primarily) scale organically through direct/PLG/partner-led customer acquisition, while AI Roll-ups scale through a combination or organic customer acquisition and M&A + AI transformation. So the LTV/CAC metric does slightly different things for SaaS businesses vs AI Roll-ups.

For SaaS companies, it estimates the lifetime value of a cohort of customers, and compares it to the (Sales and Marketing) cost of acquiring them. Usually, this cohort is defined by a combination of a time-period (e.g. Q1 2025) and a channel (e.g. direct sales). For AI Roll-ups, LTV/CAC should serve the same intent, but the cohort of customers is the entire book of business acquired from one M&A transaction.

1a. Calculating Lifetime Value (LTV)

The core concept of calculating LTV remains the same - estimate the economic value generated by a recurring stream of revenue for a cohort of customers. The ideal way to define this economic value is some measure of profitability - for SaaS, that measure is Gross Margin. It is directly related to a customer cohort and it is what ends up paying for the CAC (Sales & Marketing costs), and also for R&D, and HQ expenses (G&A). For AI Roll-ups, we believe EBITDA is a better measure. The economic stream acquired through an M&A is the [Gross Margin - SG&A] (i.e asset EBITDA1). That’s what ends up justifying the acquisition price, and also HQ-level expenses like R&D, and HQ team.

The fundamental premise of an AI Roll-up is that AI can automate a meaningful portion of COGS (typically 30%+). Since EBITDA = Revenue − COGS - SG&A expenses, successful AI transformations also improve EBITDA Margins.

Therefore, rather than using today’s EBITDA, we calculate LTV using the post-transformation EBITDA2, since that better reflects the economics we expect over the lifetime of the acquired customer relationships3.

Our adapted formula therefore becomes:

LTV = (ARR * Post-transformation EBITDA Margin %) / Annual Churn

Where ARR = Annual Re-occuring Revenue4

1b. Calculating Customer Acquisition Cost (CAC)

The formula sees a meaningful departure from the SaaS version here.

Instead of using S&M expenses, the relevant measure for AI Roll-ups is the cost of acquiring the entire business; because M&A is a key mode through which customer relationships are acquired.

Clearly, an acquisition buys much more than customers - it also includes employees, brand, contracts, licenses and other assets. So this is not the same definition of CAC used in SaaS. I’m intentionally adapting the framework to better reflect the realities of an acquisition-led business model. For simplicity, let’s also assume the acquired business has no excess cash or debt (which is usually adjusted while calculating the Acquisition Price).

That means:

CAC = Acquisition Price = Enterprise Value

EV for services businesses is usually calculated using an EBITDA multiple. So that gives us,

CAC = Enterprise Value = EBITDA × Acquisition Multiple

For smaller service businesses, acquisition multiples typically range from roughly 2–5x EBITDA.

One important nuance is that the AI transformation happens after the acquisition. Therefore, CAC is calculated using pre-transformation EBITDA, while LTV reflects the economics after transformation. This asymmetry is precisely what creates investment returns: you acquire a business based on today’s earnings while creating value through future operational improvements.

Enterprise Value (EV) = Pre-transformation EBITDA × Acquisition Multiple

In reality, you also need to incur some additional acquisition-related costs (e.g. legal, due diligence) and integration costs. Depending on the size of the acquisition, these usually end up being between 5-10% of the EV. To make our estimate robust, let’s add a 10% loading on the EV to account for these expenses.

CAC = EV + Acquisition and Integration Expenses

So our final formula becomes:

CAC = (Pre-transformation EBITDA x Acquisition Multiple) x (1 + Load Factor)

The “Load Factor” here is simply 10%

2. Simulating LTV/CAC for AI Roll-ups

2a. A typical services business - pre-transformation

Let us first apply this formula to an acquisition of a typical white-collar service business.

The image below shows the P&L of a representative example.

Applying our adapted formulae to this business shows something very interesting:

A typical M&A transaction already beats the 3x LTV/CAC threshold even before the AI transformation!

This looks great on paper, but it comes with an important caveat: without AI transformation, these economics are difficult to scale. Every additional dollar of revenue still requires proportionally more employees. If ten tax professionals serve $2 million of revenue, you’ll likely need twenty to serve $4 million, forty to serve $8 million, and so on. Very quickly, organisational complexity (which grows exponentially, not linearly to scale) becomes the limiting factor. Many owner-operators deliberately choose not to keep growing because managing increasingly large workforces becomes disproportionately difficult.

Now let us look at how AI transformation affects this equation.

2b. Impact of AI transformation

For the same business, let’s assume AI transformation can reduce labour-related COGS by 50%. As the below chart shows, that productivity flows directly through the P&L increasing Gross Margin to 1.75x and EBITDA to 3x!

Our LTV now becomes:

LTV = 100 × 45% / 5% = 900

CAC remains unchanged at 82.5, because we purchased the business before the transformation.

Giving:

LTV/CAC = 10.9x. The rarest of the rare in SaaS-land!

The above figures are not a work of fiction: They are based on actual results delivered by AI Roll-up pioneers across 100+ acquisitions! We can call them the “Base case” scenario.

But for the sake of intellectual rigour, let’s tone down these assumptions significantly.

2c. Impact of (conservative) AI transformation

Let’s reduce the COGS automation from 50% to the lower-end (30%), and also let’s increase the annual revenue churn to 7% assuming there is heavy drop-off of customers during the integration and transformation.

Even then, we end up with a >5.5x LTV/CAC. Still comfortably top-quartile in the world of traditional SaaS!

And even these assumptions remain very conservative.

• Many of the verticals targeted by AI Roll-ups exhibit relatively low recurring revenue churn. In the majority of cases, service quality and responsiveness improve after transformation, creating revenue expansion rather than contraction. Increased throughput capacity is often used to drive margin expansion via growth (more customers and products, without adding costs) rather than just via cost (same customers and products at lower cost base). Even if customer churn were to occur immediately following an acquisition, it would likely stabilize after the first one or two years. (For context, a 7% annual churn rate implies 40% of all revenue is churned away within 7 years!)

• So far we’ve only considered improvements to COGS. In practice, AI can also streamline SG&A functions such as billing, scheduling, reporting and internal operations, creating further operating leverage.

• Finally, while LTV/CAC should theoretically be independent of financing structure, AI Roll-ups are somewhat unique in that acquisitions are often funded with only 20-50% upfront equity, with the remainder financed through debt, seller notes and earn-outs. While not part of the LTV/CAC calculation itself, this further enhances the overall economics. Quantifying how leverage and deferred consideration reshape these returns is a meaningful exercise in itself (one we may return to in a future post!)

So we now know AI Roll-ups can deliver exceptional LTV/CAC characteristics, under conservative assumptions, and challenging definitions.

Next, let’s flip the question on its head …

3. Working Backwards - “Goal Seek” for LTV/CAC 3x

As one final exercise, let’s do the reverse calculation: what would lead to the LTV/CAC of AI Roll-up to drop to 3x?

There are infinitely many combinations, but here are some possible scenarios:

• Goal Seek 1: Acquisition discipline disappears and buyers consistently pay 10x EBITDA rather than 2–5x; OR

• Goal Seek 2: Annual churn explodes to 14% (this means total churn is 65% in 7 years); OR

• Goal Seek 3: AI transformation fails to deliver any meaningful productivity uplift, but churn still increases to 7% annually

These are all fairly extreme assumptions - but they drive the point home:

For SaaS, achieving a 3x LTV/CAC often requires being a positive outlier.

For AI Roll-ups, reaching the same ratio may require becoming a negative outlier.

Stepping back

We think LTV/CAC is a very interesting lens to compare business models. When viewed through this familiar framework, the economics of scaling an AI Roll-up appear remarkably compelling compared to a SaaS business. AI Roll-ups almost eliminate GTM and PMF risks, and shift the risk profile more towards careful selection of the terrain, and execution.

We have been SaaS investors for several years, and have had front-row seats to market transformations using technology. Our belief in the power of AI and Software continues with Tenet, but our strong conviction is that AI Roll-ups are a better delivery mechanism for the diffusion of this technology into European small businesses. Attractive LTV/CAC dynamics are just one of many reasons for this optimism!

If you’re curious about our views on the AI Roll-up playbook, do give us a shout!

Over a year ago, I shared a financial model to test out whether AI Roll-ups could be an outlier financial strategy. Today we are sharing one you can actually use to build.

When I published our early model, I was upfront about what it was and wasn’t. It was a simplified scenario-testing tool: could you acquire several fragmented services businesses, transform them using AI, and generate returns that justify the complexity? The answer was “yes” under a range of assumptions. But the model was always meant to be more a proof-of-concept than practitioner-grade.

At Tenet, we recognize that the gap between a model that proves a thesis and a model that helps run an AIPRO business is significant.

From scenario testing to actual planning

My partner at Tenet, Alex, brings a very different vantage point to financial modelling. He spent years at Clearsight Investments, operating deep in the buy-and-build world. He thinks of models the way a CFO does: as a budgeting and business planning tool. A living Instrument that tracks real assumptions against actual outcomes. The model he’s built is constructed from the ground up, based on actual planning models used in the micro-PE space.

Today, we are open sourcing this model for all AI Roll-up founders.

You can download the model here.

About the financial model

We’ve designed this spreadsheet model to enable you to actually run your business planning. It covers the full lifecycle of an AI Roll-up, with all the knobs and dials that a founder will care about when it comes to planning and budgeting.

For example:

• Choose the frequency, size, margin profile, and EBITDA multiples of your acquisitions

• Granularly define your team scaling profile and costs

• Built-in controls for debt capacity levels

• Explore sensitivities across key input assumptions

The overall philosophy has been to make it as granular as needed, without being so complex that it’s not useful.

Why did we create the model?

We built this model and are sharing it publicly because we believe that AI Roll-up founders deserve better than starting with a blank spreadsheet (and Claude for Excel 😉). We saw several founders repurposing generic SaaS financial models, which are not sufficient to capture the full complexity of building an AI Roll-up (Debt Mechanics, Margin Expansion, Organic vs Inorganic Growth, Working Capital etc.) - hopefully this model serves as a good starting point and saves a few weeks of effort. We also hope this model helps expose founders to some additional parameters that they might not have thought of before.

Read Me: some caveats before you use the model

The model ships with illustrated numbers, pre-populated throughout: it’s easier to react to a number than to conjure one up from nothing. But treat every pre-filled assumption as a provocation, not a recommendation. Replace them with your own numbers before you share this further. A model is only as useful as the inputs behind it, and of course the inputs are yours to validate. Garbage in, Garbage out holds here as well 😊

Of course, no model can survive an interaction with the real world, and the same is going to hold true for this as well. What the model can do is give you a rigorous structure to make those calls and track them over time.

What’s next?

We plan to keep this model updated based on feedback and inputs from founders. We will keep releasing versions (with changelogs) every time there are significant updates to ship. We really enjoyed simulating various scenarios on this model internally, and would love to share some observations in subsequent posts.

For now, please kick the tires with this spreadsheet, and if you have any questions or feedback, please reach out directly to Alex (LinkedIn) or me (Twitter / LinkedIn).

You can download the model here.

[Originally posted on AI Roll-up Nexus: https://www.ai-rollup.fyi/essays/definitive-guide-raising-debt]

It’s not a stretch to say that debt will likely account for at least half of the total capital raised by most AI Roll-ups over their lifetime. As a corollary, being thoughtful about how you raise that debt is critical. Yet, it’s also the topic that the majority of founders I’ve spoken to are least comfortable with.

A few months ago, we hosted a credit masterclass with Linus from TechCredit Partners for a small group of 25 AI Roll-up founders, and the response was phenomenal. Still, the questions kept coming. No clear founder-friendly guide to raising debt for AI Roll-ups exists right now, so I asked Linus to co-author this with me. Together, we’ll unpack what you can expect when raising debt for your AI Roll-up — the key concepts, lender psychology, and the trade-offs that actually matter.

We expect this guide to be especially helpful for founders navigating the debt raising journey for the first time.

Before we dive in, one of Linus’s favorite quotes sets the tone perfectly:

“Venture capital is all about making a possible $100 from $1, while debt is about making a certain $1 from $100.”

It’s a simple but powerful framing, and an excellent lens for understanding how lenders think.

1. Debt: Basic Terms and Concepts

The credit world loves complexity. You’ll find structures for every imaginable scenario, but in practice, three concepts matter most when you’re starting out: how interest is repaid, how principal is repaid, and how the debt will help you to generate overall equity value. Mastering these gives you a good foundation for navigating most debt discussions.

1a. Mechanisms for Interest Repayment — Cash Interest or PIK?

Cash interest is the portion of a loan’s interest that is paid in regular cash installments during the life of the loan, typically monthly or quarterly.

PIK (Payment-in-Kind) interest is interest that is not paid in cash but instead added to the loan balance, increasing the principal owed over time.

For the same principal and interest rate, PIK structures allow you to carry more debt since no immediate cash payments are required.

1b. Mechanisms for Principal Repayment — Amortization or Bullet?

Amortization of principal means repaying the loan in regular installments that include both interest and principal, reducing the outstanding balance over time.

Bullet repayment means paying off the entire principal amount in a single lump sum at the end of the loan term.

In amortizing loans, it is common to have an interest-only period, during which the borrower pays only interest, allowing time to deploy capital and generate returns before starting principal repayments.

Typically, when maximizing debt capacity or preserving cash flow is key (as in acquisition-driven or growth-heavy roll-ups), Bullet repayments and PIK interest are preferred structures. Though they often come with higher cost and lender scrutiny, and are not always compatible with all lender types (more on that in the next section).

1c. Equity Value Generation

Debt is often described as non-dilutive capital because, unlike issuing equity, lenders do not typically receive ownership upfront. However, certain debt structures can still affect equity value directly or indirectly depending on their terms.

Some debt agreements include equity-linked features that compensate lenders for taking additional risk, particularly in venture debt, growth debt, or higher-risk structured financings. These commonly include:

• Warrants: options granted to the lender that allow them to purchase a small portion of equity at a pre-agreed price, usually exercisable upon an exit or IPO.

• Conversion Features: provisions allowing part of the debt to convert into equity under specific conditions, such as a future financing round, refinancing, or default event.

• Profit Participation Rights (less common): rights that entitle lenders to share in future profits or liquidity proceeds without direct equity ownership.

Depending on the instrument, dilution may range from zero in pure debt structures to approximately 5% in many venture debt facilities, and in more specialized or higher-risk structures, can reach 15–20%.

At the same time, debt affects equity value beyond explicit dilution. Interest rates, amortization schedules, covenants, collateral requirements, and repayment timing all influence how much capital remains available to reinvest in growth versus servicing obligations. Two debt facilities with similar pricing can therefore produce very different outcomes for equity holders depending on how they interact with the company’s growth profile and operating model.

The most effective debt solution is not simply the cheapest capital, but the structure that extends runway, preserves strategic flexibility, and allows management to deploy capital into initiatives that increase long-term enterprise value.

It is also worth remembering that your first debt structure is rarely your last. As your roll-up scales and your track record matures, refinancing into larger, cheaper, or more flexible facilities becomes not just possible but expected. The goal early on is to get the right structure for where you are today, one that gives you enough runway and flexibility to prove the model. Better terms will follow.

Subscribe now

2. Debt Instruments Commonly Used in AI Roll-ups

As an AI Roll-up progresses through its lifecycle, both its risk profile and financing needs evolve — and different debt instruments become suitable at each stage. The chart below outlines the most common instruments used by AI Roll-ups and their key characteristics.

AI Roll-ups will be able to unlock different types of debt at different lifecycle stages. This is what a typical progression looks like This is what a typical progression looks like (even though, every journey is different. For e.g. AI Roll-ups typically hit positive EBITDA much sooner in the journey!):

But what about your first acquisition?

For the first acquisition, most AI roll-ups should predominantly rely on equity and seller notes or earnouts rather than pull in outside debt. Early on, you should prioritize flexibility to iterate on AI transformation and change management to create a transformation playbook, rather than navigating debt covenants.

This shifts if you already have real conviction on cashflows — for example if you have tested the product and validated margin uplift with design partners. Another exception could be when the minimum viable acquisition size is large, say above $10m in enterprise value, and you want to limit the initial equity hit. But if you do not yet have proof points, expect more rigorous scrutiny on your assumptions, and potentially less favourable terms. Make sure any debt you take is an accelerator, not a drag.

3. Who Are the Players

It’s easy to assume that only a handful of institutions offer debt to AI Roll-ups. In reality, the market is large and complex, with a wide range of lenders and solutions available.

Your chances of securing the right kind of debt improve dramatically when you target the right kind of lender — one that matches your company’s stage, scale, and risk profile.

The table below outlines the key categories of debt providers and the typical structures they offer.

Continue reading the full article on AI Roll-up Nexus: https://www.ai-rollup.fyi/essays/definitive-guide-raising-debt. The full guide covers who the players actually are, what lenders really look for beyond the numbers, how to benchmark your debt capacity, the real price-tag of a raise, and a collection of horror stories worth reading before you sign anything.

About the Authors

Linus Eriksson Noren is the Founder of TechCredit Partners, a London-based debt advisory firm helping tech and high-growth companies structure the right debt facilities from seed to IPO. With a network of 400+ lenders and investors across four continents, Linus and his team have become a go-to partner for founders navigating the complexity of debt capital markets.

Sahil Patwa is a General Partner at Tenet, Europe’s first inception-stage investment firm dedicated to AI-Powered Roll-Outs (AIPRO). He is also the creator of AI Roll-up Nexus, a public resource hub for AI Roll-up founders and investors.

AI Roll-ups had their moments in 2025 — in tech media, on Twitter, and in pitch decks landing on investor desks across the world. But amidst all the noise, a basic question remained unanswered: does the hype translate to real capital?

At Tenet, we wanted to move past the anecdotes. So we surveyed over a 100 investors spanning VC, Private Equity, and Family Offices across Europe, the US, and globally to build a data-driven picture of where sentiment actually stands. Respondents included capital allocators from firms like Accel, Bessemer, General Catalyst, Insight Partners, EQT, BlackRock, Atomico, and many others, covering every stage from pre-seed to growth.

Here’s what we found, and what we think it means.

The Capital Is Real

The headline finding:

Over 90% of respondents expressed willingness to explore an AI Roll-up investment, with at least 40% actively looking to deploy capital into the category

Return expectations are equally striking. More than 60% of investors believe AI Roll-ups can deliver better IRRs than both Public Equities (~10%) and Private Equity (~15%). And 46% see this as a bigger VC opportunity than Crypto; a category that attracted over $120 billion in deployed capital.

Perhaps most telling: over 90% of respondents expect AI Roll-ups to receive significant investor attention in the next 12-18 months. This is no longer an obscure strategy.

A New Asset Class?

More than half of the respondents investors believe AI Roll-ups represent a new asset class entirely, while 25% see them as “just PE with an AI wrapper”.

You buy your distribution (reducing GTM risk), deliver the same service with a lower cost structure (reducing PMF risk), but with the potential for breakout margin expansion that looks much more like a technology company. Interestingly, PE investors themselves are the most bullish on the “new category” framing — 70% see AI Roll-ups as a new asset class, compared to 47% of VCs. Perhaps those closest to the traditional roll-up playbook appreciate just how different the AI version is.

The Funding Gap Nobody’s Talking About

Here’s the finding that should matter the most to founders:

Investor appetite peaks after the first 1-2 proof-of-concept acquisitions are complete (46% of respondents prefer this stage). But appetite drops at the pre-acquisition stage, when a company has built its product but hasn’t yet proven the transformation thesis through a real deal.

This creates a structural gap. A typical $1-3M seed cheque often isn’t enough to fund both the initial product build and the first acquisitions needed to prove the playbook. Founders risk building great technology but running out of runway before they can demonstrate the EBITDA uplift that unlocks the next wave of capital.

We believe the inception and pre-acquisition stages should be combined into a slightly larger founding round — giving teams enough firepower to reach the proof points that open the floodgates.

Sectors, Risks, and Exits

Professional Services and Financial Services are the clear investor favourites, while IT Services, Real Estate, and Trade & Logistics look underpenetrated relative to demand.

On value creation, AI-led margin improvement dominates (86% of respondents), and a 2x EBITDA uplift would cross the bar for 90% of respondents (Note: this is far higher than the 10-20% uplift that traditional PE synergies typically deliver).

Integration and change management complexity is top-of-mind for investors (~80% ranked it as top 3); but on the contrary, it sits lower than expected in what respondents considered ‘must have’ founder skills.

The other key risks are overhyped AI value-creation (68%), and poor M&A discipline (62%). We share these concerns, with a nuance: the real risk isn’t that AI won’t be able to automate service workflows — it’s that teams try to do too much too quickly. This is a “systems thinking” problem more than an AI capability risk.

On exits, ~80% see PE consolidators as the primary path, while a third also consider IPO realistic for breakout cases. The coexistence of both reinforces the “PE Floor with a VC Ceiling” view.

What Happens Next

As one investor in the survey put it: “Many will do it and few will execute it well.” Another noted: “There are a ton of legacy markets which have not changed in 20+ years where this strategy makes sense.” Both are true — and the tension between them is exactly what makes this space exciting.

Read the full report for all the data, cross-cuts by investor type and geography, and founder quotes from the field.

An interactive report is available here, or you can download a pdf copy here.

Subscribe now

People who know me know that I’m obsessed with AI Roll-ups.

I truly believe that AI-enabled services will define one of the largest global opportunities over the next decade - and one that Europe, in particular, is uniquely positioned to embrace. Not just economically, but structurally. AI Roll-ups might be the single most powerful lever to drive real digital modernization across European industries that have been underserved by technology for decades.

When I launched AI Roll-up Nexus in early 2025, the thesis already felt exciting and logical. But after spending time with the community of founders building in this category, this excitement turned to overwhelming optimism! These were AI-native founders, wanting to build real businesses, and scale them to ultimately transform entire industries!

The Core Thesis

At the heart of it, the thesis is simple:

AI can automate 50%+ of knowledge work. And when applied to the right services-heavy, low-margin industries, this AI-automation can transform sleepy businesses into machines that scale with software-like margins and pace.

We’ve written extensively about AI-native services and AI Roll-ups across topics like industry selection, how they scale, pivoting from SaaS to AI Roll-ups, and several practical lessons from the field (we’ve curated these resources on AI Roll-up Nexus). As we continued our study of this space, several elements eventually fell into place.

But as the ecosystem matured, one gap became impossible to ignore.

The Capital Mismatch Problem

At inception, AI Roll-ups don’t fit cleanly into existing buckets:

• Pre-seed / Seed Venture Capital (typically €1-3m) is well-suited for classic SaaS companies where the capital is primarily deployed towards initial product development and early traction with the explicit objective of raising follow-on capital in 18-24 months. However it is structurally misaligned with the needs of AI Roll-ups.
  ~€2m is generally insufficient to simultaneously build meaningful AI capabilities, acquire initial operating assets, and progress towards cash-flow sustainability or “default-alive” status. As a result, founders become highly dependent on near-term follow-on financing without sufficient proof points.
  Moreover, traditional VC value creation is optimised for exponential organic growth, often codified through aggressive scaling benchmarks. Failure to achieve breakout growth materially reduces the probability of follow-on funding, creating strong (and unwarranted) pressure to prioritise rapid scaling over sustainable business development.
  In addition, classic VC funds operate under a “Power Law’ return model. Portfolio construction assumes that a small number of investments drive the majority of fund returns. This typically implies underwriting against multi-billion-dollar exits. Where such “fund returner” potential does not manifest early, companies risk being deprioritised within the portfolio, reinforcing a scale-or-bust dynamic.

  Pre-Seed Venture Capital is intentionally structured to deliver high-upside but low-predictability outcomes. This creates a structural mismatch: it treats AI Roll-up companies as binary bets, whereas we view them as durable assets that retain the optionality to become either venture-scale successes (the “VC way”) or cash-generative consolidation platforms (the “PE++ way”).

• Micro-PE and Search Funds sit at the other end of the spectrum. They lean heavily on debt, optimize early for cash flows, and often lack the flexibility to make transformational investments in product, AI, and change management - precisely where the biggest long-term value for AI Roll-ups lies.

The uncomfortable truth: at inception, it’s almost impossible to predict whether an AI Roll-up will scale the “VC way” or the “PE++ way”. The critical unknown is not ambition or execution, but how powerful AI capabilities will be 2-3 years from now, which no one can credibly predict.

And hence, forcing rigid capital structures too early is often the wrong answer for AI Roll-up founders.

We built Tenet to fill this gap.

Going Back to First Principles

The word Tenet means a core belief — something accepted as fundamentally true.

In many ways, this is about going back to the core tenets of investing:

• capital should serve the business,

• structure should follow strategy,

• and flexibility is a feature, not a bug

Our capital is purpose-built for AIPRO founders at day zero, and uniquely flexible to support both VC-style and PE-style capital as the business scales. We put business first, and design capital around the company - not the other way around.

Europe is at the “perfect storm” of world-class tech operator and AI-application talent on the one end; and massive under-digitization, a looming SMB succession crisis, and severe shortage of a qualified workforce on the other hand. This makes Europe the most attractive region for AIPRO - it is also where we see the largest opportunity, and we’ve chosen to focus exclusively on this region.

I’m incredibly lucky to be building Tenet alongside my partners Alex, Martin, Simon, and Igor. Interestingly, we all come from very different worlds (early-stage VC, growth equity, and micro-PE), and yet independently converged on the same conclusion: AI Roll-ups are the biggest opportunity of the coming decades, and AIPRO founders need a new kind of inception-stage partner.

Specialist support

As a specialist fund we will always endeavour to provide support that goes far beyond board meetings and hiring intros. We are purpose-built to help with challenges that are unique to AIPRO founders - like M&A execution, integration and change management, raising and managing debt, and navigating TopCo/MidCo/OpCo company structures.

Through my experiences driving large-scale digital transformations at BCG, and convincing grocery store owners to adopt technology at Swiggy, I’ve seen firsthand how brutally hard change management actually is. Founders who can drive it are special - and rare! And so, we’re proud to be guided by European AI Roll-up pioneers from companies like Arbio, Buena, and Zinco, who have lived these challenges in the real world.

Hello, World!

We’ve already made our first investment, and are open for business! We’re very grateful that tech.eu has written about Tenet - you can read more about us here.

If you’re building (or thinking about building) an AI Roll-up or AI-native services company in Europe, we would love to talk.

And yes, for those paying close attention: I did tease this in my last post. 🙂

Last week I heard the news that a tech-enabled Roll-up in the Roofing space was raising capital through an Initial Coin Offering (ICO) 🤯.

This was certainly not on my AI Roll-ups Bingo Card for 2025, but it got me thinking - what are my expectations of this space in 2026? So here goes …

Subscribe now

My 10 predictions for AI Roll-ups in 2026

1. At least 400 companies on the AI Roll-up Nexus

Let’s start with an easy one. We have ~150 companies on the list right now, and we had just started in the second quarter of 2025. I expect ~250 more companies to get added to the database next year.

2. Total capital invested in AI Roll-ups in 2026 crosses $3b

Yes, this sounds like a large number, especially when compared to the total funding raised by AI Roll-ups in 2025 so far. But it still represents only around 1 percent of total VC, and an even smaller fraction of PE capital deployed this year. My view is that AI Roll-ups will continue to grow while remaining a non-consensus category for some time. Against that backdrop, $3b feels like a reasonable first milestone for the space.

3. PE investors get into the game: at least 5 deals where PE provides follow-on capital

While AI Roll-ups have been the domain of Venture Capital so far, I feel 2026 is finally the year when PE will start figuring out a structure to partner with them. Given the low risk-appetite of PE, I predict they will get involved with AI Roll-ups during their “Growth Phase” when they have already done several acquisitions and proven that a scalable playbook exists.

4. At least 1 Venture Debt fund launches(/markets) a dedicated product for AI Roll-ups

Venture Debt funds have been very receptive to this opportunity, and I’ve been impressed at how quickly they have adopted this into their repertoire of playbooks that they fund. Still, lending to AI Roll-ups requires different guardrails and expectations compared to traditional Roll-ups. I believe at least a few Venture Debt funds will close the gap, and at least one will openly market a tailored product for AI Roll-ups

5. Launch of at least one $100m+ specialist fund dedicated to AI Roll-ups

After hundreds of conversations with founders, and investors (including the Investor Sentiment Survey), I am convinced that AI Roll-up founders need a specialist capital provider.

AI Roll-ups represent a fundamentally new value proposition for entrepreneurs seeking to transform industries through technology. The model enables sustainable scaling rather than growth at all costs, or an overreliance on financial engineering. While they may eventually be compatible with either venture or private equity capital, neither is a one-size-fits-all solution at inception. What’s missing is a dedicated fund that can provide capital purpose-built for this strategy, especially at the earliest stages.

I predict 2026 will be the year someone raises the first truly dedicated AI Roll-up fund (and I envy them!).

6. AI Roll-ups enter the physical world, with at least 5 Roll-ups heavily leveraging Drones/Robotics and Computer Vision

While we all have our eyes on LLMs, there is another powerful revolution happening in the physical world. I am looking for early-signs of new AI Roll-ups that are built on the back of these physical world technologies in 2026!

7. At least 1 AI Roll-up on each continent (except Antarctica) that has raised $50m+ capital

I strongly believe that certain geographies lend themselves especially better to AI Roll-ups, compared to others (my top pick is Europe - by a margin). Having said that, I still believe AI Roll-ups will travel far and wide, and we will have at least 1 sizable company (based on $$ raised) on every continent where humans outnumber penguins.

8. Standardized AI Roll-up Metrics become industry norms

SaaS has a Rule of 40 and the SaaS funding napkin, Crypto has a 1000 true fans, consumer/marketplaces have a virality coefficient. I believe 2026 might see a focal AI Roll-up Metric appear in a lot of fundraise decks. One would hope it’s something to do with sustainable operational efficiency/margin uplift, and not multiples expansion!

9. At least 1 video on AI Roll-ups goes viral on Instagram/TikTok with 1m+ views

I expect a Codie Sanchez-meets-Andrej Karpathy moment in 2026, where some early-to-the-game ‘Fin’fluencers start discussing AI Roll-ups; and one of them creates a breakout video!

10a. A bestseller book gets published on AI Roll-ups

OR

10b. A top-tier business school publication writes a case study on AI Roll-ups

I would assume it’s not long before a major publication like Harvard Business Review picks up on the topic. While the strategy does not exactly lend itself to mass-appeal (and we are still in the earliest stages of its evolution), another manifestation of my prediction is that a book on AI Roll-ups becomes a NYT bestseller.

So far, I’m encouraged that this category has largely been approached with a sensible growth lens rather than treated as a hype driven gold rush, and these predictions reflect that view. That said, a few of them are deliberately a bit “out there” to keep things interesting 🙂

I’m looking forward to seeing how many we can check off when I revisit this list next year. If you have thoughts on the list or believe I’ve missed an important box, feel free to DM me on X or LinkedIn.

AI roll-ups are getting a lot of buzz—but how do investors really feel about them?

I’ve launched a short survey to capture perspectives on market appetite, risks, and opportunities. The goal is to share aggregated insights back with the builder + investor community.

✅ Already, 40+ organizations have taken the survey.

👉 Take the survey here

If you’re a VC, Family Office, or Growth investor tracking AI, I’d love your input.

Be part of the signal 📡

Over the last few months, we've been dissecting AI roll-ups from every possible angle: industry selection, financing options, financial modeling, go-to-market paths, and even how SaaS incumbents might pivot into this play. But theory only gets you so far.

In recent conversations with founders actively building in this space, several hard-won lessons have emerged — messy, nuanced, invaluable. These aren't strategic frameworks or Excel models. These are the insights you only gain when you start implementing the model in the real world.

Below, I've organized these field lessons into four critical areas: Upfront Choices, Change Management, Tech Platform, and Growth & Scaling.

Subscribe now

🧭 Before You Start

Critical decisions to make upfront

Figure out whether this is a VC or PE play

Before raising even a single dollar, get clear on what you're actually building. Are you relying on fundamental AI transformation, or just incremental efficiencies and multiple arbitrage? This determines everything: your investor base, team, timeline, success metrics.

If you're banking on incremental efficiency unlock and multiple expansion, you're playing a PE game. Here's the catch though: PE is increasingly moving away from industries vulnerable to AI disruption, which means that multiple expansion might or might not materialize at exit. If your model depends on buying at 4x and selling at 8x, you could be building on shaky ground.

If you're underwriting fundamental AI transformation (proving you can drive 3-4x margin improvement through technology), you're in VC territory. However, not all VCs are comfortable with this motion; you need investors who truly believe in this model.

Build first, then Buy

Should you build and then buy, or buy and then build? Experiences of successful founders hint strongly at building first.

You need an opinionated tech-driven transformation playbook before you acquire assets. Without a proven product, your ability to drive meaningful change remains theoretical. If you can't confidently model a 3-4x EBITDA uplift through your tech platform, the unit economics might not work, and your debt service costs might pile up.

Even if you lack domain expertise, work with design partners in your target industry and build as a software provider first. This approach lets you validate that meaningful transformation is actually achievable before you commit acquisition capital.

Do as much building upfront as possible, while maintaining flexibility — you'll inevitably need to rebuild components once you understand the operational realities of your acquisitions. Building is a continuous process anyway, but starting with a proven foundation makes all the difference in your ability to execute successful integrations.

Obscure markets might turn out to be the best

People often ask what industries are great for AI roll-ups. While some are objectively promising (see), recent conversations have highlighted the virtues of the more obscure ones.

Why? Less competition for assets means better pricing. Fewer ready-made tools means more defensible margins. When everyone's chasing the same obvious markets, differentiation gets harder. If I can list the 30 "best" industries on a whiteboard, I'd rather go after the 31st. The best play might be one where you have a unique edge: niche knowledge, access, or insight. That often makes the path to success much smoother.

🔄 Change Management

Integrating people is the hardest part — and the most important

Your first acquisition sets the tone

The platform company becomes your testing ground for transformation, integration, and playbook refinement. Multiple founders have emphasized prioritizing "forgiving targets" for your first deal: companies with open-minded teams, forward-leaning leadership, and ideally, founders who are genuinely excited about the AI transformation rather than just looking for an exit. As one founder put it,
"Your first acquisition will teach you everything about what works and what doesn't. Pick a target that won't punish you for the inevitable mistakes."

Plan your integration strategy around the seller's motivation

As you scale beyond your first deal, you'll encounter sellers with vastly different motivations; from retirement-seeking exits to growth-hungry partnerships. Recognizing these motivations during diligence, not after closing, should fundamentally shape how you structure both the deal terms and integration approach. If the target company's founder is nearing retirement and wants out, plan for a graceful handover that preserves operations without expecting ongoing involvement. But if they're younger and energized by the AI opportunity, structure the deal to bring them into the fold — they might become your most valuable operator and future transformation leader.

Design Change Management for a heterogeneous team

You'll be integrating people across a wide spectrum of AI-nativeness: from 25-year-olds who think in prompts to 55-year-olds who still consider Zoom cutting-edge. This isn't a bug; it's a feature if you plan for it properly.

Standardizing tools is straightforward. Harmonizing people requires deliberate strategy. Design differentiated training programs, roll-out timelines, and support systems that account for varying comfort levels with AI. Be ruthless about moving fast where you can, but patient where the learning curve demands it, as long as there is intentionality from the team member to adapt to this new way of working. The goal isn't just uniformity — it's getting everyone to effective productivity with your new AI-enhanced workflows, regardless of their starting point.

🧱 Tech Platform

Build a system that scales with your ambition

Build your core product in-house

While AI tools are proliferating rapidly, most founders warn against the temptation to outsource your core product offering. Off-the-shelf solutions create real risks that compound over time:

• Flexibility constraints: Will third-party tools provide the extensibility you need as your requirements evolve? Most won't.

• Integration complexity: Can you maintain reliable integrations between external products you don't control and your in-house features? The more dependencies, the more fragile your system becomes.

• Vendor risk: Most AI tools come from startups, not established companies. What's the survival rate? If that startup crashes, you're suddenly left with a critical gap in your tech stack.

• Loss of differentiation: Your core product is your competitive moat. Outsourcing it means outsourcing your advantage.

You can't build everything from day one, but structure your architecture as an orchestration layer with swappable components. Start with off-the-shelf tools for non-core functions, but always design with a clear path to replace them with in-house alternatives as you scale. Your core product should be yours from the beginning.

Go Outside-In: Phasing Your Platform Rebuild

One of the most pragmatic implementation strategies emerging from the field is to rebuild your platform from the outside in — starting with user-facing layers and gradually moving inward to the core infrastructure.

Why this works:

• Automation wins are user-facing: The user layer often offers the quickest automation gains. These not only deliver EBITDA impact, but also build credibility by giving teams tools that make them 10x more effective (and actually work), setting the tone for broader transformation efforts.

• Credibility compounds: Early wins help establish belief. When employees see automation making their lives easier, they’re far more likely to engage with deeper, more structural changes downstream. It’s change management by way of proof, not mandate.

• The core is full of unknowns: Legacy systems often hide a web of edge cases, tribal knowledge, and implicit dependencies that only reveal themselves over time. Attempting a full-scale core rebuild upfront is a recipe for surprises, delays, and avoidable risk. A gradual approach gives you space to learn and adapt before tackling the complex stuff.

• Fail-safe fallback: If your outside-in rebuild strategy hits a snag, it’s much easier to fall back to a stable legacy core layer while you regroup. This creates a natural safety net — a buffer against catastrophic platform failure during integration.

Think of it as iterative system replacement with concentric expansion. Build trust and functionality at the edge. Let usage, feedback, and operational exposure guide your march inward. Over time, you’ll earn the right to retire and rebuild the core; on your own terms.

Opinionated (Software) beats Flexible (Software)

One of the key advantages of building an AI roll-up is that you can enforce a single, opinionated way of working across every acquisition. Unlike SaaS companies that must accommodate countless edge cases and different workflows, you have the power of consolidation.

Use this ruthlessly. Build software around your specific playbook, not around flexibility. During integration, unify all people, processes, and systems under one standardized approach. Don't build for multiple ways of achieving the same outcome; build for your way. This opinionated approach makes integration cleaner, scaling faster, and operations more predictable across the entire portfolio.

Think “Systems”, not just “Automation”

Your tech platform isn't about finding small automation opportunities here and there. It needs to be a thoughtful system that integrates platform, people, and processes to balance automation with accuracy of output.

Design the platform with validation layers where models check each other's work. For critical steps, implement risk scoring that determines when to escalate to human review. Build tracking mechanisms to ensure exceptions are flagged for human attention and nothing slips through the cracks unnoticed. The share of human involvement would naturally be higher at the start and then eventually tapers off as models become smarter and errors decrease; the process playbook should be nimble enough to accommodate this transition.

This Systems approach to human-AI collaboration (not just pure automation) determines your ability to scale while maintaining quality and managing risk.

🚀 Growth & Scaling

M&A is just the beginning

Think beyond M&A for growth

It's crucial to be thoughtful about how go-to-market and scaling will evolve for your platform. While you might still rely on M&A after 8-10 acquisitions, if you've executed well in your first few deals, other growth opportunities will likely emerge.

• Organic growth through reputation: When you deliver AI-native services that are meaningfully superior to traditional competitors, the market notices. Your reputation as a best-in-class provider can unlock organic growth opportunities that didn't exist when you started: customers will seek you out rather than the other way around. You can also use your enlarged margins to invest more in organic growth, or be more price competitive if the market allows it.

• Technology product expansion: You'll inevitably build valuable technological components (payments infrastructure, data analytics products, workflow tools). These can become standalone software offerings that generate additional (high margin) revenue streams beyond your core service business.

• Franchising your platform: If you've created something truly unique and defensible as a technology product, you might grow through franchising. Others can build service businesses using your technology as the underlying infrastructure, handling the distribution challenges while you focus on platform development.

Consider selective acquisition strategies

You don't always need the entire company. Sometimes you only want customer relationships, not operational complexity. In industries where distribution separates from execution, consider acquiring just the book of business rather than the full entity. This approach is often cleaner, faster, and cheaper.

Prepare for Disintermediation Risk

One long-term threat in AI roll-ups is customer bypass of the service layer entirely, especially as technology becomes "good enough" for direct interaction (or customers get “savvy enough”). Think SMBs purchasing insurance or legal advice from AI chatbots without human intermediation.

This risk reinforces smart industry selection but also demands ongoing introspection. Regularly ask: Is there a better way to combine our technology and distribution? Sometimes the next growth phase requires evolving (or even disrupting) your own model. (Hat Tip: Clayton Christensen’s Innovator’s Dilemma). The key question becomes: Can the same team that builds opinionated internal tooling also create consumer-facing product experiences?

Final Thoughts

The AI roll-up model has moved beyond theory. It's being built, tested, and refined in real time. These insights represent part of that evolution, and I’m sure there are many more learnings to be distilled over time.

If you're working on AI roll-ups or thinking about entering this space, I'd love to hear from you and learn about your experiences in the field. Feel free to reach out via Twitter or LinkedIn.

Introduction

The idea that AI will transform services is close to consensus now. AI has the potential to invigorate previously sleepy industries, which have long resisted meaningful change, even as SaaS tools flooded the landscape.

But in many such markets, the winning path isn’t to sell more AI software. It’s to go full-stack and operate as an AI-native service provider.

I quite like this 2x2 (I’ve recreated a cleaner version below) to visualize where AI-native services make sense, but I’d tweak it slightly. I would add a third axis around willingness to change in industries where both supply (service providers) and demand (end-customers) resist AI adoption, AI-native services have a distinct edge. I’ve seen multiple AI-enabled or agentic software companies stall around the $5–8M ARR mark — not due to product issues, but due to this type of market inertia.

In this piece, I break down two key questions for founders building in this space:

1. How to win distribution

2. How to build sustained advantage

A. How to win distribution?

While AI-native service companies have a strong technological advantage in delivering the service, incumbent service providers have the advantage of already owning customer relationships. In order to create a successful (scalable, fast-growing, profitable) AI-enabled services business, founders will need to be very thoughtful about how to build distribution. I see four different levers for this:

1. Turbocharge GTM with AI

In transactional spaces where contracts aren’t deeply recurring, AI can power GTM by ensuring you're (a) the fastest to respond, and (b) the one delivering the best quality response.

While AI SDRs are transforming SaaS sales, many traditional service providers — say insurance brokers — have never heard of 11x or Landbase. They certainly wouldn’t know how to build a voice-based lead qualifier using ElevenLabs. That’s a real edge for AI-native GTM teams.

Of course, this edge won’t last forever (AI-native competitors will adopt the playbook, incumbents will reluctantly adopt vertical-specific GTM tools) - but still, there’s probably at least a 12–18 month window where AI-powered GTM can reign free.

2. nX superior offering

The obvious way for AI-enabled service businesses to win customers is through a better offering compared to incumbents - not incrementally better, but at least 2-3x better - sufficient to convince someone to switch providers. This could take many forms, but will roughly be on one/more of these dimensions:

1. Volume (do more)

2. Speed (do faster)

3. Cost (do cheaper)

4. Accuracy (do measurably better)

The true differentiator comes when AI-native companies can unlock something which wasn’t possible before. For instance, AI may allow you to analyze 100% of receipts instead of sampling 10%, or deliver same-day quotes instead of waiting five business days. Tegus is a great example. They used AI to unlock something no expert network had offered before — not just access to individual experts, but also the ability to leverage a rich library of past interview transcripts across customers. This turned previously siloed insights into a searchable, reusable knowledge base.

3. Ride a demand-supply gap tailwind

Industries where there has been a sudden surge in demand (typically caused by some big macro/regulatory dislocation) are great opportunity-spaces for this GTM play. Existing suppliers are unable to catch up to this demand peak, and this creates an “easy mode” entry point for AI-native players.

A case in point: Accounting and Tax Advisory in the UAE. This is the first year for mandatory VAT filings, and tens of thousands of businesses are scrambling to find service providers.

4. Acquire distribution inorganically

We have written about this motion (dubbed AI-powered Roll-ups) extensively — you can read about it here (parts I, II, III, IV, V) and check out this public database of companies already scaling through Roll-ups.

B. How to create a sustained advantage

While solid GTM could help win initial distribution, sustained value creation would need something more - a strategic advantage that the business accumulates over time. From my vantage point, I see 5 potential sources of advantage:

1. Network Effects

The OG (and my favourite) - very hard to build, but even harder to dismantle. Several categories of service businesses benefit greatly from network effects (e.g. 2-sided marketplaces like recruiting/staffing, expert networks, sell-side IB). AI-native companies could leverage technology to outperform (in volume/speed/quality) the incumbents and get the nfx flywheel to start spinning

2. Reputation/ Marquee Logos

In several industries, like advisory services, who your customers are often matter a lot more than how many they are, or how much revenue they give you. In such industries, acquiring logos of marquee customers early on is going to be vital. Companies that have the right logos, might win against even scaled businesses that don’t. This is where publicly positioning yourself as an AI-native player might help you capitalize on the boardroom AI hype and get your foot into the door.

3. Access to Human Capital Networks

AI may reduce headcount, but it also raises the bar for remaining humans-in-the-loop. Companies that have an unfair access to the right AI-fluent talent pool will be able to build enduring value in this space. This is especially true for industries where service providers need to be certified and hence the overall pool of candidates is finite and difficult to expand.

This is a big one. AI-native services firms will look very different from traditional services organizations — new org structures, new mindsets, new cultures. Being able to create a space where AI-fluent talent wants to work will be a “trade secret” hiding in plain sight.

4. Proprietary Workflows/ Process Design IP

This is the "Operational Excellence" play on steroids. AI-native services are essentially a complex interplay of platforms, processes and people. Great execution here requires thousands of small design decisions that collectively result in superior service. The resulting workflows could not only provide organizations with the ability to outperform competitors on service delivery, but they’re also difficult to reverse engineer, because unlike software not all parts of the offering are visible externally. Piecemeal copying doesn’t work unless it’s aligned with the rest of the playbook.

Some people are starting to refer to this as “taste” (see also).

5. Data flywheels

The notion of a data flywheel — where more data improves the product and attracts more users, creating even more data — is seductive, and valid.

But in practice, it’s often hard. More data only helps if it’s the right data, used the right way. In agentic applications, you’re not building foundational models; you’re using those foundational models to essentially build workflows. Sure, more data does help finetune the models where relevant, but the reality is that you often need orders of magnitude more data to get a meaningful delta in output (I see finetuning output as a log function on input data).

That said, RLHF data and good old A/B testing results still matter. They help shape the “taste” of your system.

Conclusion

Having a deep understanding of the industry dynamics, and the resultant strategic imperatives early on, is going to be vital to enduring success for founders (and enduring alpha for investors).

Those who misread the terrain risk chasing the wrong levers.

For instance, focusing on growing revenue through sub-scale customers may backfire in markets where marquee logos matter more. Or, overindexing on scale & margins early on might come at the expense of internal culture and might distort access to human capital. Similarly for investors, investing based on scale of revenues might be a poor indicator of success if “taste” is what would drive long-term value creation.

It's also worth noting that barring data flywheels, most of these advantages resemble traditional services moats. That’s a healthy reminder: at the end of the day, we are still building services businesses.

As the AI-native services playbook is still being written, I am excited to discover more avenues for building strategic advantage, and more GTM hacks. I’m closely following this space and genuinely excited about what’s to come. If you're building an AI-native services business, I’d love to exchange ideas. Feel free to reach out via Twitter, or LinkedIn.

As AI continues to reshape traditional service industries - from property management to logistics to insurance - entrepreneurs and investors face a critical question:

Should you build a new AI-powered services business from scratch, or buy and transform existing ones through roll-ups?

The right answer depends on the underlying characteristics of the market. Through experience and observation, four key criteria have emerged that can guide the decision:

1️⃣ Cyclicality

2️⃣ Distribution stickiness

3️⃣ Service standardization

4️⃣ Regulatory complexity

Let’s walk through each and explore when it points to organic builds vs roll-ups - with examples.

1. Cyclicality: Predictable Cash Flows Are Roll-Up Fuel

Why it matters:
Roll-ups rely on stable cash flows to service debt, fund acquisitions, and invest in platform R&D. Highly cyclical markets make this risky. In volatile industries, organic builds are safer and more adaptable.

Example: Recruitment Agencies (Organic Build)
Recruiting was on fire in 2021 - firms were turning away business due to a hiring boom. But by late 2022, the pendulum had swung: mass layoffs, hiring freezes, and dried-up demand. This kind of volatility makes it hard to build a leveraged roll-up. However, an AI-native recruitment platform - with variable infrastructure costs (like GPU inference tokens instead of headcount) - can flex up or down as demand shifts, giving it a structural advantage in cyclical markets.

See: Mercor

2. Stickiness of Distribution: Buy It if You Can’t Break In

Why it matters:
If customers rarely switch providers or think about alternatives, distribution becomes a moat. In sticky markets, roll-ups help you skip the brutal GTM grind - and the good news is, once you're in, you're likely to stay in. The same inertia that made it hard to win customers now makes them hard to lose. But if switching is easy, a better, cheaper AI-native service can grow organically.

Think of this almost as a novel GTM approach - you're growing through acquisitions because the LTV/CAC makes more sense that way than through traditional organic methods.

Example: Property Management (Roll-Up)
Most landlords don’t proactively look to switch property managers unless there’s a serious issue. Contracts are annual, but landlords typically stay with a manager for 4–5 years. Even when they're open to switching, the window is narrow - just weeks before renewal. This makes cold-start distribution brutally hard. Rolling up existing players and improving their efficiency with AI is often the only viable path at the beginning.

See: Dwelly

3. Standardization of Service Delivery: AI Loves Consistency

Why it matters:
AI thrives in environments with structured, repeatable workflows. Roll-ups only work when acquired companies can be brought onto a common operating model. In highly bespoke industries, standardization is difficult - better to start fresh.

Example: Investment Banking (Organic Build)
M&A advisory services are deeply personalized. Each deal is unique, strategies are tailored, and execution relies on hard-won client relationships. No two boutique banks operate alike - making post-acquisition integration a mess. Rather than roll up a fragmented landscape, a better approach is to build a focused, AI-native IB platform around repeatable tasks - like automating deal sourcing or first-draft materials - and expand from there.

See: Offdeal

4. Regulatory Complexity: Roll-Ups Can Fast-Track Compliance

Why it matters:
In regulated markets, going from zero to compliant is slow and expensive. Acquiring already-compliant firms can provide licenses, regulatory infrastructure, and domain expertise - saving years of effort.

Example: Financial Advisory (Roll-Up)
Providing financial advice isn’t just about trust and performance - it’s also about regulation. Advisors must hold certifications (like Series 7, 63, or CFP), follow fiduciary standards, and maintain compliance infrastructure for audits and disclosures. Starting an AI-native advisory firm from scratch means building all of that up, often over several years. Rolling up boutique advisory firms gives you immediate market access and regulatory coverage. From there, AI can be layered in to offer smarter portfolio analytics, faster onboarding, and scalable client communication - modernizing the experience without reinventing the compliance wheel.

See: AMCA

While not an AI Roll-up, AMCA illustrates this point very well in the context of defense/ aerospace (thanks for flagging this Evan!). Every legacy supplier they acquire already holds hard-won aerospace certifications (AS9100, FAA PMA, ITAR, etc.). Buying these businesses lets AMCA inherit decades of audit history and approved-vendor status, skipping the costly 'test- and-certify' cycle that a greenfield parts maker would face.

A Framework, Not a Final Answer

This isn’t a hard-and-fast playbook - think of it as a starting framework. Many companies will evolve past the initial conditions that defined their early strategy:

• A roll-up that succeeds in building a reputation for quality and navigates regulatory hurdles might find it easier to grow organically, expanding its offering and reach without further acquisitions

• Conversely, an AI+services business that starts with an organic wedge may find itself layering on selective, strategic acquisitions to accelerate scale, unlock new geographies, or enter regulated verticals

• Or, they might follow completely novel methods to scale further
  
  

In other words: how you start matters, but it doesn’t lock in how you scale.

Build or buy - but start with clarity on the terrain.

After sharing my framework for industry selection a few weeks back, I started hearing two recurring questions from founders:

1. “If I’m already exploring an industry, is there a structured way to quickly assess the relevant parameters?”

2. “If I’m starting from scratch, how do I efficiently filter out the noise and identify high-potential industries to explore further?”

Great questions — and ones I’ve wrestled with myself.

So I built a thing.

As a starting point, I’ve put together a lightweight Industry Screener. For several Industry × Geography combinations, it maps against the key parameters I laid out in Part I of my essay series — think marketing fragmentation, EBITDA margins, intensity of M&A activity, etc.

It’s not exhaustive. It’s not perfect. But it’s designed to help you:

• Save time on early-stage diligence

• Eliminate obvious non-starters

• Spot promising paths for deeper research
  

⚠️ A Quick Caveat Before You Dive In

This screener is stitched together from a mix of inputs:

• Conversations with industry operators

• Materials shared by founders actively building in these spaces

• Desk research (with generous help from Perplexity + ChatGPT)
  

While every entry passes a basic smell test, some of the data is directionally useful—not definitive. Treat this as a compass, not a GPS.

If you're using this to guide investment or build decisions, bring your own judgment (and apply a healthy margin of safety).

📈 Where We Go From Here

This is a living, breathing database. I’ll be adding to it over time as new insights come in and the landscape evolves. If you’ve got an Industry x Geo combination you think deserves a look — or if spot anything off — I’d love to hear from you [LinkedIn/Twitter]

🗃️ Link to the screener: https://www.ai-rollup.fyi/industryscreening

I hope across Parts I, II, and III I have done enough to convey my belief that for the right industry, team and capital structure, AI-powered Roll-ups could be a fantastic business idea for someone starting up, and a great investment opportunity for VC investors.

But what about people who’re not starting from scratch?

I've been having conversations with founders who are already running traditional SaaS or AI-application businesses—and who are beginning to explore a new direction. Many of them are considering “buying distribution” by acquiring services businesses. A few have already pulled the trigger on their first acquisition.

In this essay, I’ll explore why this shift—from product to services, from SaaS to hybrid—can make sense, but why it’s far from trivial.

From SaaS to Services: A Growing Playbook

Metropolis’ successful(?) acquisition of SP Plus is a canonical reference for this approach. And I do believe it can work. But it’s not an easy journey. There’s a reason you don’t see this playbook used more often—it’s operationally complex, culturally intense, and capital-sensitive.

Pulling it off requires getting several ducks in a row - here are a few considerations for founders exploring this shift.

1. Investor Expectations and Valuation Math

Existing investors who had funded your company based on SaaS multiples (c.10x ARR) are not going to be pumped at the prospect of you becoming a services company (valued at c.5-10x of EBITDA). This could also likely mean an immediate valuation haircut when raising your next round (yes, narratives wildly affect valuations in our world).

It is essential to thoughtfully manage this transition for your existing investors and bring them along for the journey: Show them the long-term perspective, and give them comfort that there is light at the end of the tunnel. And most importantly, have an honest conversation around the different paths forward for the business. Accept that some of your past assumptions around growth did not pan out that way, and work with the investors to build more robust assumptions for the next stage of your business. You’ll need to get them excited about the long-term narrative and comfortable with the short/medium-term numbers. (Now is also a good time to remind them that they promised to be “founder friendly” 😉)

2. Is This the Team to Lead the Next Chapter?

Shifting from SaaS to services—and layering in a roll-up strategy—is not a minor evolution. It’s a fundamental change in how the business operates.

Building great software and scaling a services org are different skill sets. Leading M&A on top of that is a third. It’s worth asking, candidly: do you have the team for this?

• Do you have operators who know how to run high-quality, margin-sensitive services?

• Can someone build a repeatable M&A playbook—and avoid expensive rookie mistakes?

• Is your finance lead ready to handle deal structures, integration, and debt?
  

These aren’t easy questions. But they’re necessary ones. Sometimes the answer is yes—but with some support. Other times, you may need to level up the team, or even rethink your own role.

That doesn’t mean stepping aside. But it might mean stepping into a different seat—or bringing in people who’ve walked this path before. The sooner you get real about the gaps, the faster you can fill them.

This is still your company. But to lead the next chapter well, you may need to lead it differently.

You don’t need all the answers now. But you do need to acknowledge the open questions—and accept that there will be some hard answers along the way.

3. The HQ Right-Sizing Dilemma

You no longer have the luxury of 90% gross margins.

That means no room for a bloated HQ. Every dollar saved needs to go toward acquisitions, contingency, critical R&D, and debt service. You probably don’t need a large sales and marketing team anymore. What you do need: lean, focused corporate functions that can support integration, financial planning, and ops.

Right-sizing HQ while simultaneously building new capabilities is one of the trickiest balancing acts in this transition. But it’s necessary. Think deeply about how you will achieve this, before you make the plunge.

4. Customer Strategy and Communication

One of the trickier parts of this shift is figuring out what to do with your existing customer base.

Some might willingly shift from buying software to services. Some might even become competitors once you enter the services market; while others might be great candidates to acquire. The equation is complex, but with some thoughtful structure will help. Some questions to answer:

• You might not have capital to acquire all of your services customers on Day1, what is your phasing strategy?

• Would you prioritize acquiring new players (net new revenue) or existing customers (familiarity)?

• How do you prioritize the product roadmap when internal and external needs diverge?

There are no easy answers, but what matters most is having a definitive plan—and a communication strategy that keeps customers in the loop and reassures them about what’s changing (and what’s not).

5. Rebuilding Culture from Product to Services

This pivot is not just financial or strategic—it’s also cultural. Be deliberate about how you will build a services culture - and define a new set of north star metrics, incentives, and values so the entire organization embodies it.

What does success look like in a services org? What does operational excellence mean? How do you balance responsiveness with profitability?

One of the most overlooked aspects of this transition is change management (platform + people + processes). Aligning the team on new KPIs, behaviors, and expectations takes time—and intentional effort.

Conclusion: A Narrow Bridge, But a Real One

If you’re at this crossroads, know that you’re not alone. I believe there’s a real opportunity here—especially in sectors where the tech/services boundary is blurry, and where AI capabilities can be a force multiplier for delivery.

But don’t underestimate the complexity of this path. It’s not a simple extension of your current business—it’s a reimagination.

If you’re contemplating this shift—or already deep in it—I’d love to hear from you. Reach out on LinkedIn or Twitter. Always happy to chat.

Also, do check-out this public database of AI-powered Roll-ups (and fill the form/ reach out to me if you’d like to get featured in there)!

Excited to announce the launch of our open database for AI-powered Roll-ups!

AI Roll-ups represent what I truly believe will become one of the leading modes of value creation and capture using AI in the coming few years. Right now, it's a niche community of believers—small, but growing and incredibly passionate.

We've created AI Roll-up Nexus as a public directory to help facilitate real-time sharing of information, ideas, and connections within the ecosystem.

If you're building in this exciting space, we'd love to have you featured. Please submit your details via the form linked below, and we'll make sure you're added promptly. Prefer staying under the radar? No worries—feel free to reach out to me directly on LinkedIn/Twitter. Happy to chat offline!

Let's build the future of AI-powered Roll-ups together.

Link to database: https://www.ai-rollup.fyi/

Definitions and Approach

In Parts I & II of this series, we built a framework to identify industries best suited for AI-powered roll-ups and established that the venture debt market is primed to support such endeavours. But what about equity markets? What types of equity investors are best suited to fund these ventures?

Specifically, I'm trying to determine whether AI-powered roll-ups could qualify as VC-scale investment opportunities. Before we begin, how do we define VC-scale? For this article, we'll define it as:

"if things go according to plan, is there a reasonable chance of achieving a 10x return?"

I'm mindful that this definition might stir debate, with VC purists insisting only potential 100x returns qualify (a logic I fully understand and respect). But let’s stick to the 10x definition for now.

For me, the simplest approach is to simulate how such an opportunity could unfold. By modeling various scenarios, we can identify which could realistically yield that 10x return. It’s a simplistic model designed to provide directionally accurate insights—not precise calculations down to the nth decimal.

You can download the model here, and run your own assumptions.

In the remainder of this article, I'll walk you through my findings and discuss the implications for both investors and builders in this space—crucial to set realistic expectations for both parties.

But first the tl;dr answer:

Yes, achieving 10x returns is possible—but *Conditions apply

Forecasting the financials of an AI-powered roll-up

Let's start by exploring one likely scenario to familiarize you with the model.

You can download the full model here

Here are the key assumptions and rationale. While there are numerous smaller assumptions, these are the pivotal ones:

Building these assumptions out over ten years, you can achieve a $500M+ ARR, ~$70M EBITDA business generating very healthy cash—impressive! Remarkably, if executed correctly, this can be accomplished with less than $50M in equity investments.

The Seed, Series A, and Series B investment amounts were calculated to provide sufficient runway (1.5-2 years, with buffer) covering HQ, debt servicing, and acquisition expenses.

As evident, this scenario offers a 10x opportunity for Seed investors, though less so for Series A and B. I ran numerous such simulations to determine if realistically achieving 10x returns is even feasible - and if yes, what does it take?

What does it take to achieve 10x+ for Seed, Series A & Series B investors?

1. Potential for 4-5x EBITDA uplift in 2-4 years:

As shown below, a 4x+ EBITDA uplift is nearly essential for targeting a 10x outcome. Even at a 2-4x uplift, the business case remains compelling, though returns for Series A investors onwards may range closer to 5-8x.

The speed of achieving this uplift significantly impacts returns—the quicker the uplift, the sooner cash becomes available for subsequent acquisitions, thereby amplifying investor leverage.

Naturally, achieving this uplift is easier if the initial EBITDA percentage is low (3-8%). Nonetheless, it's crucial to validate upfront that a 4x+ uplift is feasible. And this is not fiction, I have seen those kinds of AI-powered bottomline boosts happen at several companies (if you’re curious to know the details, ping me separately and I’ll tell you 😉).

Also, remember - these uplifts are what justify R&D spending at the HQ level!

Another implication of this, is that when faced with a choice, focus on achieving EBITDA uplift (even at the cost of delayed next acquisition) - and prioritize acquiring good quality assets where path to uplift is relatively clear over assets that you might get at a cheaper multiple.

2. Fewer equity rounds but higher dilution per round than typical VC deals:

Plan for 3-4 financing rounds through exit—no more. Eventually, cash flows, additional debt capacity, and seller financing/earn-outs should fund new acquisitions. Using venture funding for early R&D and initial proof-of-concept acquisitions makes sense; funding stagnant companies at 5% EBITDA does not.

Given significant early equity goes into acquisitions, valuations shouldn't follow typical VC multiples (10x revenue = fail). Expect 25-30% dilution per round for economics to work for VC investors (still leaving founders and management 30-50% equity).

3. HQ costs must stay nimble:

This is crucial, especially early, when capital must stretch between R&D and acquisitions. This balance is delicate; Series A investors seek proof that founders can acquire companies reasonably and achieve substantial EBITDA uplifts.

As the company scales, HQ overhead can easily balloon—this can undermine strategy. Aim for HQ costs to grow sub-linearly relative to revenue. Staying nimble accelerates profitability, unlocking cheaper debt financing (see Part II). If there's ever a place to stretch a dollar, this is it.

4. Founder profiles are critical:

A founding team with strong industry, operational, tech, and M&A expertise can secure larger seed rounds, enabling greater early investment in R&D or more significant initial acquisitions.

As detailed in Part II, strong founder profiles also unlock venture debt earlier, adding non-dilutive capital cushion in early stages. The underlying assumption: strong teams will execute effectively, turning Excel models into actual returns.

5. Don't rely on multiple-expansion:

I have started seeing financial models assuming acquisitions at 3-4x EV/EBITDA, exiting at 6-7x, even 10-12x multiples. While scale can lead to higher multiples, remember: multiple expansion is the cherry on top—not the sundae itself! As a general rule, I never factor any multiple expansion into my projections.

And there’s a practical explanation for this: acquisition multiples naturally rise as assets grow larger, with significant revenue coming from later, higher-multiple acquisitions. The scenarios below—constant multiples versus expanding multiples—show nearly identical outcomes (numbers sometimes align beautifully!).

Another subtle but important point to consider here, is the longevity of EBITDA uplifts. The simulation assumes consistent EBITDA uplifts even in later acquisitions (6-7 years down the road). If competitors catch up technologically, price competition may erode EBITDA gains. Hence, choosing industries where competitor catch-up potential is low (covered in Part I) is essential.

Finally, I'd like to re-emphasize that numerous assumptions underpin this analysis. A huge thanks to everyone who shared their valuable insights - the model’s strength depends entirely on the quality of these inputs. And if you have any additional perspectives or suggestions, please reach out via Twitter/LinkedIn—I look forward to incorporating your feedback in future updates.

In Part 1, we laid out a framework for evaluating which industries are best suited for an AI-powered roll-up. From market fragmentation and automation potential to margin profiles and macro resilience, the goal was clear: pick a category where AI can be a true economic multiplier — not just a bolt-on feature.

Now let’s shift gears.

Once you’ve identified the right market, the next big question is: how do you finance the roll-up? What capital structure supports this model best in the early stages, and how does that evolve over time?

We’ve spoken with venture debt firms, bank-backed credit arms, and corporate finance experts to understand how this new model is viewed from the lender’s side of the table. Here’s what we’ve learned:

1. The First Acquisition Is (Almost) Always All Equity

Let’s start with reality: your first acquisition will almost always be financed entirely with capital raised from VCs. Lenders are understandably hesitant to underwrite execution risk on an unproven roll-up model.

In rare cases — say, when the founding team has an exceptional pedigree — you might get up to 30% of the acquisition value financed through debt. But that's the exception, not the norm.

That means early equity rounds need to be sized with this in mind: enough to fund both the first acquisition and the group-level infrastructure (R&D, HQ functions) that supports it.

2. Venture Debt First, Corporate Debt Later

Once the first acquisition is in place and you’ve got some operating history, venture debt becomes the most viable source of non-dilutive capital. Both pure-play VD firms and bank-affiliated VD arms are watching the space — cautiously optimistic, but still conservative.

The inflection point comes when the roll-up hits $10M+ in revenue and becomes EBITDA breakeven at the group level.

Note: Group-level EBITDA includes all acquired assets plus shared services like platform R&D, M&A team, and HQ overhead. It’s not enough for individual companies to be profitable — the entire structure, inclusive of holdco burn, needs to be in the black.

At that stage, more traditional corporate credit options become accessible — including structured debt, mezzanine financing, and potentially even private credit — all at more favorable terms.

3. Once Proven, Debt Becomes a Bigger Lever

Once the model is validated — meaning you’ve shown that AI can reliably drive 2–3x EBITDA improvements post-acquisition — lenders start to view the roll-up as a repeatable engine rather than a one-off experiment.

At that point, 50-70% of the value of subsequent acquisitions could be debt-financed. This significantly lowers the equity requirement per deal and improves capital efficiency and returns.

In short: the better the model performs, the more leverage you can responsibly apply.

4. Lenders Are Still Writing the Playbook — So Guardrails Are Tight

AI-powered roll-ups are still a new category. Unlike SaaS aggregators, there’s no long-established benchmark for how much leverage these models can responsibly take on. As a result, lenders are still writing the playbook — and that translates into tighter guardrails early on.

Common guardrails include:

• Debt to Total Equity Raised = 0.3-1x

• Debt to Gross Margin ≤ 1x

• Debt to Total Revenue ≤ 0.5x

• Acquisition at ≤ 4-5x EBITDA (criteria could be relaxed for very high quality assets with sticky revenue and a clear path to margin improvement)
  

Until a clear performance pattern emerges, lenders are likely to remain conservative on leverage.

5. What Typical Venture Debt Terms Look Like

Separate from the guardrails, here’s what a typical venture debt package looks like in practice:

• Loan Structure: Term loans with 6–18 months interest-only, and a 48-month typical tenor. Lines of Credit (LoC) and Loans with Bullet Repayment terms usually come after scale and group-level profitability (though in exceptional cases, LoCs are possible even for scaled but cash-burning businesses)

• Interest Rates: 11–12% is a good rate currently; could be higher across a broader spectrum of lenders

• Warrants: 7-10% warrant coverage, pegged to the prior or next equity round

These terms reflect how lenders are balancing enthusiasm for the AI roll-up model with a healthy dose of caution.

6. Lenders Expect PE-Style Structuring from Day One

As venture debt enters the mix, lenders will typically expect the business to adopt private equity-style structuring — not just to cleanly separate operating entities, but to optimize their security position across the group.

That means setting up a legal architecture that includes:

• HoldCo / TopCo – the primary equity-holding and governance layer

• MidCo / PlatformCo – where shared services like AI, engineering, and G&A live

• AcquiCos – subsidiaries used to execute and hold each individual acquisition
  

This structure allows lenders to secure their loans against specific cash-generating assets while protecting against cross-contamination of liabilities. It also makes enforcement easier in downside scenarios.

In many ways, this mirrors standard PE platform architecture — and for good reason: venture debt providers want the same protections. Founders should factor this into their legal and capital planning from the outset.

7. The Playbook Matters — But So Does the Player

Lenders aren’t just evaluating the strategy — they’re evaluating the founding team and the cap table.

The definition of a “top-tier” founding team is still evolving, but lenders are typically looking for:

• A strong AI/tech background, ideally with hands-on R&D experience

• Proven operational execution chops

• M&A experience is helpful, though not essential

• Backing from credible VCs with follow-on capacity and a track record in roll-up plays

8. Permitted Acquisitions: Guardrails from Debt, Vetoes from Equity

While venture debt providers don’t typically have full veto rights over each acquisition, they do include “permitted acquisition” clauses in their loan agreements — placing clear boundaries around what kinds of companies can be acquired using debt.

Typical constraints include:

• A minimum number of years of operating history (e.g., 5+ years)

• Topline growth for at least past three years

• EBITDA positive for at least two years

• Limits to value of earn-outs

• Not in the process of insolvency / cashflow positive in the most recent year/ appropriate cash runway after acquisition

In short, VD providers set the guardrails — but they won’t dictate every move.

Equity investors should and typically will push for full veto rights on acquisitions — especially early on. Until the strategy is proven, they need tighter control over what gets added to the platform. Venture debt sets the guardrails, but equity should hold the steering wheel.

In Part 3, I’m looking to unpack the big question: do the economics of an AI-powered roll-up actually stack up for venture investors? Where does the alpha come from? What drives multiple expansion? Can this model deliver true venture-scale returns (>10x) — or is it better suited to a different kind of capital? Still early thinking here, but that’s where I’m headed next.

If you have any inputs on this topic, I would love to hear from you— please reach out on Twitter/ LinkedIn.

A big thanks to Nico, Jack, Mridul for their invaluable inputs on the draft!

Roll-ups, traditionally the domain of Private Equity (PE), are increasingly catching the eye of Venture Capital (VC) firms. The (not-so-)secret ingredient? AI.

A lot has already been said (both for, and against) the intersection of AI and roll-ups (Tidemark, Slow Ventures, Vin Iyengar, Euclid Ventures, Andrew Ziperski), but as a community, I believe we’ve only just scratched the surface.

The basic premise

AI and software can drive significant margin expansion across a range of industries. But in many of these sectors, selling software alone might not be the most effective route to value capture, for a variety of reasons. The better play might be to go full-stack—offering the service itself, powered by proprietary software. That means stepping into the arena as an operator, not just a vendor, and competing directly with incumbents.

The problem? Building distribution from scratch for these services is a long, painful process. The shortcut is to buy distribution—i.e., scale through acquisitions—and then apply AI-driven automation to unlock 2–3x improvements in EBITDA.

While this structurally resembles a PE roll-up (acquire, streamline, scale), the potential to fundamentally transform the margin profile of underlying businesses and industries provides the additional promise of software-like scaling. Furthermore, AI Roll-ups eliminate PMF and GTM risks at the outset (you are automating an “existing service” and “buying distribution”). This presents founders with an attractive paradox—AI Roll-ups are more ambitious than traditional PE and less risky than traditional SaaS—providing “a PE floor with a VC ceiling”.

Questions Galore

As a growth-stage VC, I find this space fascinating and full of potential. There are a bunch of open questions I’m thinking through:

1. Which industries are well-suited for an AI-powered roll-up strategy?

2. Could this model deliver true venture-scale outcomes—say, 10x+ returns?

3. Are debt markets mature and deep enough to finance such a strategy at scale?

4. What would great founding teams look like in this space?

5. What could be the common traps in this strategy?

…and many more

Let’s start with the first one: what makes an industry a good fit for an AI-powered roll-up?

An AI Roll-Up Needs the Right Terrain

Not all sectors are equally suited to this model. The right industry can amplify returns; the wrong one can make scaling a grind. Here’s my wishlist of industry characteristics:

A. Software Aversion
This strategy is best suited for sectors where players are reluctant to adopt software, or where selling software as a standalone product is difficult due to the heavy change management involved. For industries that are more open to buying software, the vertical SaaS route likely makes more sense.

B. Margins Matter
Low-margin, high-revenue sectors are often overlooked—but they’re ideal for this type of a play. Even modest efficiency gains could lead to a 2-3x EBITDA lift.

C. AI and Core Margin Expansion
The best plays are those where AI can automate the core service—not just admin or back-office tasks. If tech only improves support functions, I see it more as a PE play. But when AI enhances or replaces the core delivery, you get true gross margin expansion—and potentially, a VC play.

D. Market Fragmentation
If the top 10–15 players control 80%+ of the market (high concentration), one will likely need to acquire a large incumbent early—high complexity and high risk. Markets with high fragmentation, where there is a large number of potential assets to buy across different sizes, are more suitable. Ideally, the market consists of at least 1000 players, presenting a large enough buyable universe of targets.

E. Level of Automation Among Incumbents
Sectors still running on manual processes offer the greatest potential for value creation—automation here can be a step-change, not just a tweak. But if incumbents are already partially automated, the window may be narrower. They’re more likely to catch up once AI tools go mainstream, limiting the durability of any EBITDA lift. And even if you get there first, the product needs to be complex enough to be defensible. If the tech is easy to replicate, expect margin compression—fast.

F. Knowledge Intensity
Some services—law, consulting, etc.—rely on specialized expertise that tech can’t easily replace. These are tough to automate. Moreover, the “alpha” in these industries comes from specific personnel (Partners) and there is a real risk of losing topline post-acquisition, if a few of them exit. Focus instead on industries with repeatable, process-driven workflows where software can scale the service without sacrificing quality.

G. Macro Resilience
Prefer industries with stable, recurring demand and low exposure to economic cycles. Areas like insurance TPAs, freight audit and payments, or accounting tend to be more predictable than cyclical ones like mortgage brokering—making scale and forecasting far easier.

H. Limited Roll-Up Activity
Avoid sectors where roll-ups have already driven up valuations. You want room to acquire at 2-5x EBITDA, leaving space for multiple expansion and operational upside. Less competition also means better deal terms and more attractive targets.

I. Product Expansion Potential
If you can layer on adjacent products—payments, financing, insurance—it’s the cherry on top. These add-ons boost margins, extend LTV, and deepen customer lock-in, turning a strong services business into a full-stack platform.

Of course, finding industries that check most of these boxes is hard, and doesn’t guarantee success—but it could significantly increase the odds. These characteristics create the foundation for scalable, tech-enabled roll-ups that can deliver both operational leverage and venture-scale outcomes. If you have opinions on which sectors are especially interesting for this play, or if you disagree with any of the points above, I’d love to hear from you—feel free to reach out on Twitter/ LinkedIn.

Traditionally, cybersecurity solutions have catered to the enterprise and mid-market sector, leaving individuals and SMBs underserved and vulnerable. This has not become a major crisis so far because the return on investment (ROI) of conducting an attack on an individual did not justify a cyberattacker’s time. But this is now changing — fast!

The rapid rise of mobile-first banking and payments makes individuals more susceptible to financial losses due to cyberattacks. As our lives become increasingly digital, new attack vectors emerge daily. Already, phishing attacks on individuals are doubling every year, and the average loss from such incidents is now over $2,000 in some geographies. Generative AI is further lowering the cost of launching sophisticated attacks, leading to an exponential increase in both the volume and severity of threats.

We need a new type of solution to protect the digital population of the world, and we believe Whalebone is leading the way.

Whalebone — the new paradigm of individual cybersecurity

Antivirus (AV) software was once the frontline defense for individuals and SMBs, but it falls short in protecting them at scale. The problem? AVs only guard against malware that has already infiltrated a device — while modern threats like phishing and identity theft strike online, without ever installing malicious software. On top of that, AV adoption is riddled with friction. Buying, installing, updating, and renewing is a hassle for users. And because acquiring customers is so expensive, AV providers are forced to price their products high — putting protection out of reach for many. Protecting individuals and SMBs is as much a distribution problem as it is a technology problem.

Whalebone’s offerings solve all the above problems:

Whalebone partners with telecom operators (telcos) to block malicious URLs at the network level. Security is embedded directly into the network infrastructure. Customers simply opt in through their telco provider, and protection is activated with a single click. For a small monthly fee — a fraction of traditional AV costs — users receive seamless, always-on protection. Whalebone earns a revenue share for every protected subscriber.

This model not only lowers CAC but also removes friction entirely:

• No software installation

• Automatic updates managed by the telco

• Integrated billing with monthly telco invoices

This innovative distribution model is backed by world-class technology and threat intelligence developed by their team in Brno, Czech Republic — a global hub of cybersecurity talent (also home to AVG).

Through continuous R&D and collaborations with CERTs, CSIRTs, universities, and government agencies, Whalebone has achieved best-in-class detection rates (what % of cyberattacks are detected and blocked) and industry-leading low false positives (what % of blocked URLs were actually not malicious). Their solution is as precise as it is comprehensive!

Embedded Cybersecurity in action

We see Whalebone pioneering a new paradigm — what we like to call “Embedded Cybersecurity”. This approach integrates security into the very product/service that the customer is using, such as telco networks, ensuring frictionless protection for individuals and SMBs.

This model achieves the objectives of cost-effective distribution, seamless buying journeys, and hassle-free installation/maintenance through an embedded product and B2B2C/B2B2B GTM model.

Just as embedded finance revolutionized payments, we believe Embedded Cybersecurity will extend to other cybersecurity domains, redefining digital safety for individuals and SMBs against a wide range of threats.

Our investment in Whalebone

Richard and Robert founded Whalebone in 2016 with a clear mission: to make cybersecurity accessible to individuals and SMBs. When we first met them 15 months ago in 2023, their rare blend of bold ambition and disciplined, frugal execution stood out. Over the course of time, our conviction grew — Richard and Robert were exactly the kind of founders we wanted to back. What impressed us even more was how this culture of ambition and operational excellence extended across their entire team.

During this time, Whalebone forged partnerships with leading telcos across Europe, Asia, Africa, and the Americas — all while running near burn-neutral. Our conversations with key players in our telco network further reinforced Whalebone’s reputation for delivering results.

Fast forward to 2025: Unbound is thrilled to lead Whalebone’s Series B, along with participation from the majority of their existing investors. This funding will accelerate their global expansion and fuel cutting-edge R&D to extend their leadership position. Beyond telco-led cybersecurity, Whalebone is already leveraging its core technology to protect end-users of ISPs, enterprise networks, and even government agencies.

Many exciting product launches are on the horizon, so stay tuned for updates. Whalebone is also growing its team across multiple functions. If you’re passionate about reshaping cybersecurity for individuals and SMBs, or eager to write the next chapter in the Embedded Cybersecurity playbook, explore their Careers page.

Here’s to 1 Billion protected users!

The rise of Embedded Cybersecurity: why we invested in Whalebone was originally published in The Thesis on Medium, where people are continuing the conversation by highlighting and responding to this story.

Update on seat vs usage-based pricing metrics

A couple of quarters ago, I made a comment that it is useful to view public cloud companies through a seat vs usage-based pricing lens. The assumption is that companies in these two categories would see different growth-profiles depending on a complex set of forces that affect buyer/user behaviour (budgets, buying patterns, urgency of use-case and need to bypass making a business case, impact of AI, persistence of demand, …).

Sharing two charts based on this basis the latest (Q3 ’24) earnings.

On an average, it seems like growth of usage-based companies has picked up at a faster pace compared to seat-based pricing companies.

On the flip side, FCF numbers indicate that usage-based companies have also invested more in growth (here I use declining FCF as a proxy for investments into growth).

The phantom expense

Much has been said about the hidden costs of excessive stock based compensation (SBC) by SaaS companies. I particularly like Janelle Teng’s summary of all points of view on this topic (read Part I and Part II of her posts).

I believe this is an interesting metric to track going forward, as it could have a pretty significant impact on a shareholder’s returns. So as part of my quarterly analysis, I’m also putting together a few charts on SBC.

Median SBC as % of revenue is at 17% — lower as compared to late ’22 / early ’23 (where it was ~21%) but still quite significant. For some companies like data infrastructure players Snowflake and Confluent, this figure is as high as 35–40%!

Interestingly, this SBC as % of revenue metric did not change significantly in Q3 ’24 compared to Q2 ’24. Given all the investor attention to this metric, one would expect this change to be more rapid.

What is even more surprising, is that the median public cloud company spends more than their free cash flow on SBC! The median SBC/FCF ratio is 107%, and there are companies in the 200%+ range as well

Do we need a SBC-adjusted rule of 40?

So my conclusion is that public cloud companies are indeed trying to reign in their SBC expenditure, but not fast enough. But surely the investor community would catch on to this, and penalize companies with very high SBCs?

To figure this out, I ran a regression on the EV/LTM Revenue multiple, with independent variables as (i) SBC/Revenue, and the 2 rule of 40 metrics, (ii) YoY Revenue Growth % and (iii)FCF margin %. All figures were based on the latest Q3 ’24 quarter

Here’s the regression equation:

EV/LTM Revenue = 35* Revenue Growth % + 14* FCF % - 1* SBC/Revenue + 0.3

Note: R-squared = 0.54 which is pretty decent and in line with Bessemer’s rule of X

The first two coefficients are unsurprising. There has been a lot of literature around the fact that growth is given 2–3x more weightage as compared to free cash flow, while valuing a company. (see: Bessemer’s Rule of X, Meritech Rule of 40). Our regression also shows that Growth gets ~2.5x weightage (35/14) compared to FCF.

What I did not expect, was that SBC/Revenue matters less than <0.03 times as compared to Revenue growth. The charts below show that SBC/Revenue hardly affects the EV/LTM Revenue multiple.

Or if you like statistics, another way to look at it is, the P-value for SBC/Revenue is very high at 0.85. Now, a P-value of <0.05 indicates that a variable is statistically significant — at 0.85, our variable is highly insignificant.

In other words, the public markets are unaffected by the amount of SBC spent by the management. It really does seem like SBC is a phantom expense 🙂

The good news, we can continue using the rule of X as a useful heuristic without having to worry about other adjusted FCF, or GAAP operating profits!

The bad news, the multiples of public cloud stocks could be on a ticking time-bomb — what will be the trigger? I hope to not find out the hard way

Stock Based Compensation, and more on Seat vs Usage based pricing was originally published in The Thesis on Medium, where people are continuing the conversation by highlighting and responding to this story.

We ❤ B2B Software

B2B software has historically been a very exciting space for us at Unbound. It is already a dominant category in private markets, accounting for 25%+ of all VC funding value in 2023; and in public markets with 78 public cloud B2B software companies commanding a total market cap of more than a trillion dollars! Part of the allure for B2B software has been that if done right, it can be a very resilient category with potential for strong moats and consistently high NRRs.

During all the noise about software being dead, we remain quite bullish on this category. One of the key reasons is the capability of this space to continuously increase TAMs. Over the past decade, B2B software has seamlessly transcended beyond enterprise to mid-market, and SMBs creating many winners in the process. Now new advances especially in AI/ML are further accelerating this movement, expanding old markets and creating several new ones which were earlier not feasible.

For the purposes of this post, when I say B2B software I mean all solutions designed to streamline and automate various processes of an organization’s operations.This also includes the platforms, tools and primitives used by organizations to build their own software applications. While some segments of FinTech also fall within the above definition (after all, “every company will be a fintech company” right?), this article excludes anything that primarily does what most would consider bread and butter fintech (lending, payments, banking, wealth management, insurance).

Breaking it down further

We slice B2B software into 3 key categories:

Software Instructure & Tooling: Software Infrastructure refers to the foundational technology and systems that support the development, deployment, and operation of software applications. Software tooling enable creation, testing, deploying, and maintaining software applications

Horizontal SaaS: Software solutions designed to serve a wide range of industries and business functions, and delivered on the cloud.Horizontal SaaS are general-purpose applications that can be applied across various sectors and are not limited to any specific industry

Vertical SaaS: Cloud-based software solutions tailored to meet the specific needs of a particular industry or sector. They provide specialized functionalities and tools that address the unique requirements and challenges of a specific industry.

Currently Horizontal SaaS is clearly the dominant segment, commanding a >50% share across both public and private companies in this space. But we strongly believe this dominance is coming to an end, and that a higher share of new software budgets will now be spent on the other categories.

The great exodus

Which of the other two categories is this erstwhile horizontal SaaS spend moving to: Software Infrastructure & Tooling, or Vertical SaaS? Our prediction says BOTH.We feel the Horizontal SaaS category could be split in the middle, and each side is getting attacked by a different category.

Horizontal ⇒ Vertical SaaS:

A growing number of (especially mid-market/ SMB) organizations are now realizing that sector-agnostic ERPs, CRMs, and Project Management tools are not working for them anymore. To begin with, they need to be heavily customized/ configured to meet their desired workflows, often relying on Solution Consultants/ System Integrators. In fact the TAM for ERP SIs is more than a third of the TAM for ERP solutions! And even after the cost and effort spent in configuration, the solutions still lack the flexibility that is needed to seamlessly integrate into the organization’s workflows. Vertical SaaS solutions also trump their horizontal peers in being quick to release vertical-specific functionality (often related to specific requirements for reporting, compliance and keeping pace with changing industry regulations).

Horizontal SaaS also struggles to compete with pricing and GTM motions which are tailormade for a particular industry, often ending up being both the costly and ineffective option for buyers. This does not mean Vertical SaaS has lower ACVs compared to their horizontal peers, in fact quite the opposite. For the same type of customer, there are enough instances of Vertical solutions being able to charge a higher price due to the significantly higher value unlocked.

As software copilots (and LLMs in general) shrink the cost of building applications, the advantages of Vertical Solutions (deep understanding of customer problems, focused GTM motions) will widen the ROI-gap between Vertical and Horizontal solutions.

Horizontal SaaS ⇒ Software Infrastructure & Tooling:

On the other side, more sophisticated (often enterprise/ scale-up) organizations are now relying on their internal platform teams to build critical new applications in-house, and diverting their spend from Applications to Infra & Tools. Interestingly, the proliferation of Data Warehouse/ Lakehouse could be one of the key enablers of this shift — if you think about it tools like CRMs, HRMS etc are nothing but hardened data apps built on top of a warehouse! While an extreme example is Klarna replacing Salesforce & Workday with their in-house CRM and HCM solutions; I expect this change to happen for more “missing-critical” new applications from flexibility and speed of execution reasons, and not solely for cost reasons.

In addition to this, as adoption of cloud, microservices, and AI/ML grows, the need for robust infrastructure and effective tooling will continue to grow. This rapid digitalization and taking control of own stack will also expand the attack-surface and make organizations open to new vulnerabilities — strong tailwinds for cybersecurity demand.

Average revenue growth of public companies in these categories provides some additional early validation for the above exodus. To be clear, our point of view if not that companies will replace existing horizontal solutions with vertical/ in-house solutions en masse (we are not that naive), rather our claim is that incremental $$ will likely go more towards Vertical SaaS & Software Infra and these categories will create a much higher proportion of large outcomes.

Our B2B software thesis

Some thoughts on Vertical SaaS companies

1. We like solutions in industries where there is sufficient customer fragmentation (ideally a potential customer base of 10,000+, but at least 5,000+). This is important to avoid customer concentration risk where 50–100 customers pretty much control the entire market

2. The problem space also should be sufficiently complex to warrant investments in building a sophisticated product that can be distributed to several customers. If 2 copilot-powered 10x engineers can build a solution for the problem statement in 6 months, then the problem is not complex enough to lend itself well to long-term defensibility

3. Very likely the vertical solution will have some element of AI (currently on the roadmap), and fit-for-purpose solutions in today’s AI age mean seamlessly integrated into my exact workflow. Achieving this fit would likely require some service component during onboarding and even steady state — Vertical Software founders should not shy away from this, and rather, figure out a way to do this with healthy gross margins and without an the services manpower being a rate limiter for growth

4. Given problem discovery, solution creation and sales are so tightly interlinked for these extremely fit-for-purpose applications, having a GTM wedge and strong access to customers from Day0 is extremely vital (even more so than usual)

5. Scaling vertical solutions on a hockey stick trajectory does not happen without all critical ingredients in place, and consequently the right path for many companies could be slow(er) but near-breakeven scaling (compared to T2D3) till the stage is set for explosive growth (likely at Series B)

6. We still struggle to find Verticals where it looks possible to scale to $100m+ ARR without having a presence in the USA — our bias therefore remains for companies solving problems which are either truly-global, or else USA-centric

Some exciting themes and opportunities for Software Infrastructure companies

1. As the costs of conducting a cyberattack decline due to AI, SMBs and individuals become feasible targets. But keeping them secure is less of a technical problem, and more of a distribution problem. We see an opportunity for innovative business models like embedded agentless cybersecurity to cater to these segments

2. Enterprises will accelerate their pace of data stack modernization driven by a need to implement Generative AI, which will also likely drive on-prem to cloud and monolith to microservices architecture migrations. This creates a need for durable execution/ workflow orchestration solutions

3. We foresee increased scrutiny on both Data & ML inference spends, leading to room for new architectures and/or additional layers in the data & MLOps stack that address this. The challenge here is going to be around how to enable organizations to adopt them with minimal disruptions, and how to ensure the juice is worth the squeeze

4. Enterprise automation could look very different in the agentic systems era, and we are excited for what the new paradigm (beyond task-mining, process-mining and RPA) looks like

We are at a pivotal moment in the evolution of B2B software, and I am excited to see which ones of our theses play out, and where do we receive huge surprises. If you are building in similar areas as ones outlined above, we would love to exchange notes! And if you’re building in a completely different area that we’ve not thought about, would love for you to educate me :)

I can be reached at sahil@un-bound.com, or @sahilpatwa.

Our B2B Software thesis was originally published in The Thesis on Medium, where people are continuing the conversation by highlighting and responding to this story.

1. “what’s the point — things rarely go as per projections anyway”

2. “I feel like this is a total distraction from my main task of building/hiring/selling”

3. “it is too early to start projecting, there are so many unknowns”

While I do not disagree with any of the above (things WON’T go as per projections, it WILL take time away from building/hiring/selling, there ARE several known and unknown unknowns), I still strongly believe that building a financial budget is an essential task for a Founder/CEO.

The biggest reason for this disconnect is that founders sometimes see budgets as a score-card for judging whether or not they did well, and I see them as a map that tells you all the ways to arrive at your destination, and whether you need to accelerate/change the route to get there on time and with the fuel left in your tank.

Why is a budget important?

• A budget forces you to objectively articulate your financial goal ($XXm in ARR while burning <$XXm of cash), and keeps reminding you of it every time you review it (you’ll be surprised how many Seed/Series A founders cannot tell me what their goalposts are before their next round of funding)

• It then shows you one possible path to achieving that goal, including what it would take (growth rate needed, headcount needed, S&M budget needed, etc.). You will of course never hit the budget “exactly” this way, but you will never hit it “at all” if you don’t have one to begin with

• A budget also provides starting point answers to questions like “how much can we afford to pay our head of marketing” and “can we afford to do 5 free pilots this quarter?”, instead of a founder having to answer them from scratch as and when the need arises

• The biggest value add of a budget is, when things go wrong (and they most definitely will), you can pin-point to which assumptions were wrong — this helps you understand which areas to focus on going forward, and also what parts of the budget need to change for the reminder of your cash runway. It will also tell you when “seemingly drastic” actions are premature, or overdue!

• It would also elevate the quality of your Board conversations and make them much richer. Board members (especially VCs) sit on multiple boards, and could share specific insights or action steps based on their experience, if presented with numbers against a budget. This way the discussion can quickly move from “what is the problem” to “how can we fix the ship” when early alarm bells go off

• Especially for founders who are not financially savvy, this one-time pain will actually help you better understand your business financially (and whether you consider yourself financially savvy or not, you NEED to understand how your actions will affect the financial outcomes of your business — it’s not optional)

Convinced? OK, then let’s move to creating a financial plan.

How to create a budget:

I will not go into the nitty-gritty of actually building one in a spreadsheet — there are plenty of blogs/ videos out there that will help you do that. Rather, I want to talk about the broad framework for building one. (even if you feel like you’re already quite savvy at building one, I would still urge you to give the below a quick read!)

Step1: Knowing the goal. Your goal might well be hitting break-even before you run out of cash (which is quite sensible in itself), but chances are, if you’re a VC funded business, your immediate goalpost is raising your next round of funding. So a likely goal will be “Hitting $Xm ARR in Y months with a maximum burn of $Zm”. This is not a line you say lightly — I strongly advise you to be well informed about what are reasonable values of X, Y and Z. Speak to your current investors/ board if you have one, other VC firms, read some market insights (like @Chrija’s famous SaaS funding Napkins, SaaStr, Craft Ventures’ posts) to validate these numbers. Also note that your timeline will be dependent on your cash-out date. Since you will be starting fundraising at least 6 months before your cash-out date, your target ARR needs to be achieved by month 18 (if runway is till month 24), and hence your growth rate needs to be adjusted accordingly.

Step2: Build roadmap to revenues. Now start breaking down this revenue figure into # Customers * ACV (all best-guess assumptions), and then ramp this up from current traction to goalpost traction. Here, avoid the temptation to apply a month-on-month/quarter-on-quarter growth rate, and instead try to even out the revenue growth over the entire duration. A good way to do this is to estimate new ACV added every quarter (and you can feel free to apply a growth rate there)! Read another one of @Chrija’s excellent posts on the pitfalls of incorrectly analysing your ARR/Revenue.

Step3: Identify what is needed to achieve this. Next work backwards from this to identify what you would need to achieve this — (Sales & Marketing: how many customers do I need to onboard every month? What conversion rate should I assume? Hence what qualified pipeline do I need? Hence how many BDRs do I need? And how much money should I spend on marketing to achieve this qualified pipeline?); (R&D: what should my product roadmap look like to close type 1, 2, 3 leads in my pipeline? Hence what kind of an R&D team should I be looking for?)

Step4: Putting it all together. Bring all revenues & costs together to see what your burn looks like — will you run out of cash before achieving your goalpost traction & product maturity? Go back to your assumptions and figure out what you need to change — can I hire less? can I improve ROI on my marketing? can I cut any unnecessary spends without materially affecting my outcomes? (An easy solution founders turn to is just being more optimistic about your growth assumptions — avoid this trap at all costs! Revenue is less under your control than expenditures, so if you need to push, better to push on cost assumptions. Also, by being optimistic about revenue, should things go wrong you will face the double whammy of lower revenue contribution to bottomline, and higher cost which you had not reduced during planning and your burn will shoot up)

Step5: Sanity check. Lastly, go through the entire budget and check if it is “reasonable” — will you realistically be able to hire as per the headcount plan, or should you consider a longer lead time? Does that mean product velocity will be slower than assumed? Will that have an impact on new customer onboarding?

A few tips while you build the model:

1. Only include metrics/assumptions to a granularity that you can track — the idea is to make this a practical plan which you will come back to every month. That goal is not achieved if you can’t track your performance against those assumptions every month!

2. Create the model in such a way that you as a founder are comfortable playing with the assumptions, and understanding the interplay between them. A complex model is no good if you can’t understand it and work with it

3. Be REALISTIC about your assumptions. Achieving your goal on a spreadsheet is scarcely going to help if you can’t replicate that in real life! It is better to realise that it is not possible to achieve your desired goal with realistic assumptions, and hence go back and understand if it’s possible to revise the goals (e.g. aim for a smaller Series B, or a slower more capital efficient growth). E.g. don’t expect conversion rate of 70% if your past conversion has been 20%

4. Don’t spend too much time on building this — remember the REAL work is building/hiring/selling as several founders have rightly pointed out to me ;)

OK, so now we have a budget. Job done. I don’t have to think about it for another year right? Wrong.

How do I use the budget?

1. First of all, use the budget to establish hard constraints on your expenditures. Remember, your mission (should you choose to accept it 😎) is to not just to hit the new ACV goal, but rather to hit the new ACV AND burn goal. I have often seen founders overspend on marketing vs budget because that’s what it takes to achieve the budgeted ACV. This is a dangerous attitude — minor deviations are fine, but any major departure from budgeted expenses should factor into the plan, and should be discussed internally (with your co-founder, and board/investors)

2. Review the budget vs actuals at a set frequency — ideally every month, but at the very least every quarter. I find it helpful to not just measure budget vs actuals for outcomes, but also for assumptions — which hypotheses seem to be on point? Which ones did not pan out as expected? What have we learned about the market/ business/ operations because of this?

3. Use this to revise any assumptions, and build an updated view of the future. This is also where you start taking important calls about the next few months

4. First question to ask is, has my runway materially changed? If yes, I need to change something. This means plugging growth-related expenses (marketing/ sales/ hiring) till I get comfortable with the growth<>runway equation

5. My rule of thumb is, 1 month of underperformance on budget can be dismissed as an anomaly, but 2 straight months of underperformance means you need to start thinking about course correction. A full quarter of underperformance means I will likely not hit my goal at my current pace — I need to control my costs, step back and figure out what to fix so that my growth flywheels kick-in again. Once I am convinced that my growth machinery is working (achieving expected conversion %, quota attainment, TOFU build), I can start spending on sales & marketing again.

Note 1: It is important to recognize from the above that flexibility of cost-base is usually helpful as it helps absorb some short-term fluctuations/ deviations from plan. This is also why I suggest tying Hiring goals with Traction milestones, rather than just months — if your growth is not up to the mark, sooner or later you will need to cut expenses, and headcount is the most (mentally, emotionally, financially) expensive cost item to cut.

Note 2: Conversely, if you see better than expected returns on your investment in S&M/ R&D (translating to higher revenue growth/ lower burn), you could consider spending more and growing faster (what a delightful situation to be in!)

Using the budget as a tool to take better decisions for the business is clearly a craft, one that can be honed over time (and after suffering many deviations from the plan). If you have thoughts on how to do this better, or if you agree/disagree with any of the conclusions above, I would love to hear from you on X (@sahilpatwa) or sahil@un-bound.com

Big thanks to Vikram & Mridul for their feedback on early drafts of this post!

Budgeting for Success (and Runway) was originally published in The Thesis on Medium, where people are continuing the conversation by highlighting and responding to this story.