AI Is Coming for Professional Services – OpenAI Just Put Money on It

01 /The Concept

On December 1st, OpenAI took an equity stake in Thrive Holdings – a private equity fund that invests in professional services companies, currently focused on accounting and IT services businesses.

Thrive Holdings was created by Thrive Capital, the venture firm that backed OpenAI in 2023 when its valuation was $27 billion, then led the $6.6 billion round that pushed that valuation to $157 billion. OpenAI is now valued at roughly $500 billion.

A quick distinction worth having in mind:

- Venture funds invest in startups, aiming to sell their stake at a profit when the company goes public or gets acquired. The goal is valuation growth.

- Private equity funds buy stakes in (or entire) established companies and earn returns from dividends over time. The goal is rapid profitability improvement – recover the investment quickly, then maximize ongoing earnings.

Thrive Holdings is targeting accounting and IT services companies – a market running into hundreds of billions of dollars annually. These businesses operate on processes designed decades ago. Employees spend significant time on manual, repetitive work. Clients pay premium rates for slow turnarounds.

This is broadly true across the "real economy" – non-digital service industries generally. They're large, they're fragmented, and they're running on outdated workflows. Introducing modern AI into that environment produces dramatic efficiency gains – and dramatic profitability improvement, which is exactly what private equity is designed to capture.

But applying AI to real-world business processes is harder than it sounds. Research from MIT this year found that 95% of enterprise AI pilots fail.

The partnership's logic follows from that: OpenAI deploys its own engineers into Thrive Holdings' portfolio companies to ensure AI implementation actually works. OpenAI gets equity in Thrive Holdings; Thrive's portfolio companies get implementation support that most organizations can't assemble internally.

The financial terms aren't public, but analysts suggest OpenAI most likely received its equity stake not for cash but in exchange for the engineering labor – a meaningful contribution given how scarce hands-on AI implementation talent is.

02 /Why It Matters

This looks like a high-level play between large institutions. It's actually a signal about a trend large enough to accommodate thousands of startups of any size – because the target market is enormous and highly fragmented.

The framing from one of the deal announcements says it plainly: "OpenAI takes stake in Thrive Holdings to remake the services industry with AI."

The services sector is projected to exceed $17 trillion in 2025 and grow to $34.2 trillion by 2033. It's one of the least digitized corners of the economy, even by pre-AI standards. Introducing AI into that environment isn't incremental improvement – it's closer to a sector-wide transformation.

Several startups are already executing against this thesis, each with a slightly different model:

Multiplier ([related review](/review/unikalnyj-moment-kogda-dlja-masshtabirovanija)) raised $27.5 million in June. Rather than selling software to accounting, audit, and financial consulting firms, it acquires them – then deploys its own AI tools to increase their efficiency from the inside. Its first acquisition happened without any cash changing hands: the deal was structured entirely around a revenue and profit growth commitment. Within 8 months, the acquired firm's revenue grew 2.5x.

Platform Accounting Group ([related review](/review/a-umnye-vidjat-vozmozhnost)) raised $85 million in February last year. It provides a shared digital platform and centralized back-office services to small local accounting firms, which continue operating under their own names but share economics with the parent. Strong performers can be acquired outright.

Pipedreams ([related review](/review/makdonalds-dlja-uslug)) took the same acquisition approach but for home services – HVAC, plumbing, maintenance companies. It raised $25.5 million in February last year, then announced it would be its last equity round: the returns from buying, improving, and holding service businesses are predictable enough to fund subsequent acquisitions entirely through debt.

The OpenAI/Thrive story surfaces another important data point. The a16z partner recently wrote that "forward deployed engineers" – people who embed with clients and turn technical prototypes into working business solutions – have become the hottest role in AI startups.

This model trades margin for retention. You pay more (you're deploying expensive human talent), but the integration runs deep into the client's operations. Switching costs are enormous. LTV expands. Future revenue becomes more predictable.

03 /Opportunities

The headline opportunity is straightforward: build specialized AI solutions for specific service industry verticals.

The route can follow the traditional startup path – build a platform, sell it to clients. But the more interesting models in this space involve acquiring the client businesses themselves, deploying the platform internally, and capturing a share of the efficiency gains rather than selling software licenses.

In the most conservative variant of this structure: acquired firms grow their revenue and profit significantly. The original owners retain a smaller stake but earn more in absolute terms. The acquirer recoups investment and earns ongoing returns. Employees get meaningful bonuses tied to performance. Everyone wins.

In the more aggressive variant – Pipedreams' approach – you borrow to buy, improve, repay, and then operate profitably at scale.

The model details matter less than the underlying logic: AI applied well to a previously manual service business creates a step-change in profitability. That delta is the economic engine. What matters most is choosing the right service vertical, building deep enough domain expertise to make the AI genuinely useful, and executing the implementation well enough to avoid being part of that 95% failure rate.

Which service businesses in your domain are still operating the way they did twenty years ago – where AI could make the work dramatically faster, cheaper, and more consistent? That gap is still very much available.