What is an AI rollup?+
An AI rollup (also written AI roll-up, or AI-enabled roll-up in General Catalyst's coinage) is a roll-up strategy with an AI operating layer. A classic roll-up, also called buy-and-build, buys a platform company in a fragmented services industry, bolts on smaller add-ons at lower EBITDA multiples, and exits the combined business at a higher multiple. An AI rollup adds a reusable AI and engineering layer that gets deployed into every acquired business to expand margins and capacity, and it usually holds the companies rather than flipping them. General Catalyst coined the term in its 2025 essay The Future of Services; Long Lake, Beacon Software, and Multiplier Holdings are executing versions of it.
Is an AI rollup just private equity with an AI wrapper?+
Sometimes, and the evidence bar on this page exists to tell the difference. General Catalyst's position is that the model is not PE: it adds engineering cost, commits $100M or more per platform, and holds for the long term instead of a three-to-five-year fund clock. Skeptics such as Fortune's 'AI rollup mirage' op-ed and Equal Ventures reply that savings from AI tools every competitor can rent are ephemeral, and that the result is 'at best, tech-enabled private equity.' What settles it is third-party-verified margin and growth data from acquired businesses. As of this update, only conventional PE benchmarks (Vista, Apollo) carry grade-A outcomes; every new-category firm is still at B or C.
Is an AI rollup the same thing as services as software?+
Related, not identical. 'Services as software' (Sequoia's 'sell the work, not the tool'; HFS Research's trademarked Services-as-Software) describes AI delivering a service outcome directly, often from a firm built AI-first. General Catalyst's Madhu Namburi calls the AI rollup 'service as software' because the end state is the same: a services business priced and run like software. The difference is the starting point. Every firm on this map acquires businesses that existed before the AI. AI-native services startups that build from scratch are out of scope here.
What is the difference between an AI holdco and an AI rollup?+
Mostly capital structure. A permanent-capital holdco (Thrive Holdings, Long Lake, Multiplier Holdings, Sequence Holdings, Circeus) has no fund life and no forced exit, so it can hold what it buys indefinitely. A venture-backed platform (Crescendo, Eudia, Dwelly) sits on a fund clock and is expected to return capital on a schedule. Both run the same acquire-and-transform playbook. 'AI holdco' also gets used for AI-product studios and, separately, for advertising holding companies; neither is what this map covers.
What is AI-native private equity?+
The phrase has two meanings. In the fund sense, an AI-native PE firm runs its own sourcing, diligence, and monitoring on AI; EQT's Motherbrain is the best-known example. In the owner sense, which is what this map grades, it is a fund or holding company whose value-creation model depends on an in-house AI transformation capability applied to every company it owns, rather than on capital and governance alone. Conventional PE firms running AI programs across an existing portfolio sit in the benchmark lane.
Do AI rollups actually work?+
The public record cannot prove it yet, which is why this map grades evidence instead of ranking firms. The bull case has dated, sourced facts: Long Lake's 30+ acquisitions and its $6.3B Amex GBT take-private, Beacon's $225M Series C, Current's roughly 7,000 AI-drafted tax returns. The bear case has base rates: most classic roll-ups fail to create value, services businesses trade at a fraction of software multiples, and savings from tools every competitor can rent are hard to defend. Every operating number from the new-category firms is still company- or investor-reported. A firm moves to grade A when a customer, auditor, or counterparty confirms the outcome.
An AI rollup wants to buy my business. What should I ask?+
Ask for audited or client-confirmed before-and-after numbers from at least two businesses they already own: EBITDA margin at close versus today, and the average time to margin improvement. Ask what their headline AI number measures and who verified it. Ask what happens to your team, brand, and client relationships in year one. If equity rollover is part of the offer, ask what the second bite is worth and on what timeline. And ask who keeps the software and the data if you part ways. The grades on this map tell you how much of each firm's pitch survives contact with the public record.
Which private equity firms have verified AI results?+
On this map, Vista Equity Partners and Apollo carry grade-A outcomes: their portfolio AI results are documented by third parties, not only by the firms. Hg, Blackstone, EQT, and Permira publish substantial programs whose headline numbers are firm-reported (grade B). KKR, Carlyle, Thoma Bravo, TPG, and Ardian publish AI programs but, as of this update, no third-party-verified portfolio outcome we could find, so they are not profiled.
Is a transformation partner the same as an AI operating partner?+
Functionally, yes, without the payroll. An AI operating partner is a fund-level seat, distinct from a portfolio company's Chief AI Officer and from the technology operating partner, that owns AI value creation across a portfolio; Korn Ferry and Heidrick & Struggles both describe the role, and it can be filled in-house, fractionally, or by an outside firm. The transformation partners on this map, Caritas included, act as that outsourced AI operating arm: they embed engineers in businesses they do not own and are paid in fees, outcomes, or equity.
Which verticals have the most substantiated AI rollup results?+
Accounting and IT services (Thrive Holdings' Current and Shield, Multiplier Holdings), property and HOA management (Long Lake, Dwelly, Buena), corporate travel (Long Lake's Amex GBT deal), and parking (Metropolis) have the most reported operating results, all company- or investor-reported so far. Legal (Eudia), managed IT (Titan), insurance, healthcare, and dental have serious theses and funding but little public proof of AI outcomes yet.
How were firms selected and graded?+
We reviewed 80+ owner-operators, rollup platforms, transformation partners, and PE benchmark programs, then verified every claim against primary sources: press coverage, case studies, funding announcements, SEC filings, and counterparty confirmations. Grade A means a named AI outcome verified by a credible third party. Grade B means a client-named or company-reported result without independent attribution. Grade C means a real thesis with no publicly verifiable outcome yet. Corporate facts such as a closed acquisition or a funding round are cited but do not raise the grade on their own. Demos, projected margins, and branded 'AI OS' claims did not count. The wider universe is larger: the AI Roll-up Nexus directory lists 209 companies; this map profiles the firms that had enough public record to grade.
Why is Caritas Venture Co. on its own map?+
Because we operate in the transformation-partner lane and it would be misleading to publish a category map that omits us or quietly ranks us first. Our entry is graded B under the same standard as everyone else: client-named, vendor-published cases with stated methodology and no third-party audit. The disclosure section explains exactly how to read that.
Are these firms ranked by performance?+
No. Grades measure the quality of public evidence, which is different from actual performance. A grade-C firm may be performing brilliantly in private; a grade-B firm's numbers may not survive an audit. The grade tells you how much weight the public record can bear today.
How often is this map updated?+
Continuously as primary sources change, with every edit logged in the changelog. Corrections are welcome: email a primary source and we will fix the record. Firms marked 'profile reviewed' have confirmed or corrected their own entry.
