Nearly $15B just piled into putting AI to work

Incumbents and AI-native firms are rewriting B2B services. Each week: one deep dive into what they’re actually building, plus a few headlines worth your time.

Hi, I’m Jimmy Bijlani, founder and CEO of AI Momentum Partners (AMP). Welcome to the first issue. This week, we follow the money betting that AI’s hardest problem is no longer the model. We’ll try to keep it shorter than your last board deck.

Every AI services pitch this year comes with the same slide. An arrow pointing up and to the right, the word “agentic” in 44-point font, and a stock photo of three people pointing at a laptop they are clearly not using.

So we put the decks down and read the funding announcements, SEC filings and press coverage instead.

We tracked the companies raising serious money to close AI’s last mile, the gap between what AI can do in a demo and work actually getting done inside a company. 38 cleared our bar: founded or rebuilt around AI since 2022, at least $10M in disclosed capital, and in the business of getting AI working inside companies or doing the work with it, rather than selling software for one function. Together they’ve raised or committed about $14.7B. Of the money we could date, only about $210M was announced before 2025.

Before anyone hits reply to correct us, “or committed” is doing some work there. About $4B is pledged to OpenAI’s venture and paid in over five years, about $1.5B is a reported figure Anthropic hasn’t confirmed, and $3.1B is one AI rollup’s equity check to buy a corporate travel business. We counted it all because it’s all capital lined up behind the same bet, and we’ll flag the asterisks as we go.

Some of you saw the map on LinkedIn. Here’s what it’s actually telling us.

More than 85% of it was announced in 2026, and more than 60% in a single quarter, mostly in one week in May, when OpenAI and Anthropic each announced a PE-backed deployment business and Long Lake agreed to buy Amex GBT.

The cloud providers were moving too. Google put up a $750M partner fund in April, AWS committed $1B at the end of June and Microsoft $2.5B two days later, each aimed at least partly at putting engineers inside customers. None of that is in our $14.7B, since it’s internal budget rather than outside capital.

So in about ten weeks, OpenAI, Anthropic, Google, AWS and Microsoft all put money behind getting AI installed. Nobody pays to install something that installs itself. The bottleneck has moved out of the model and into the work.

Nearly 80% of the money, about $11.6B, sits with four vehicles. The other 34 companies have raised about $3.1B combined, roughly what Thrive Holdings has raised on its own.

Look at how the four are financed. OpenAI says 19 partners have committed more than $4B to its venture, which the FT reported is paid in over five years with a targeted minimum annual return of about 17.5%. Thrive Holdings and Long Lake are venture-backed but run the private equity playbook: permanent holding companies that buy services businesses and keep them, with no fund clock ticking. And actual private equity is all over the lab ventures, where 11 PE firms, from TPG and Bain Capital to Blackstone and Apollo, have committed money.

That makes for a tidy loop. Those same PE firms own plenty of the companies the lab ventures want to sell to, and OpenAI’s venture counts its partners’ 2,000-plus businesses as target clients. For a sponsor, that’s a stake in the vendor and a ready pipeline in the portfolio. If you run one of those portfolio companies, the likeliest first pitch comes through your sponsor.

Every company on the map is betting on the same gap. They disagree about who keeps the savings once AI closes it, and the money splits roughly in half.

The selling side has about $7.0B. Lab ventures, independent implementers and platforms get paid to make AI run inside a client’s work. If it works, the client keeps the savings, minus the invoice.

The owning side has about $7.7B. AI-native service firms and rollups do the work themselves and keep the savings as margin. Long Lake’s one deal is $3.1B of that, but even without it, about $4.5B is betting the savings end up as the owner’s margin.

On the selling side, the labs dominate. The $4B+ committed to OpenAI’s venture alone is nearly 12 times the $342M raised by the five independent implementers in our guide combined, and Ode was reported at about $1.5B. Bain & Company, Capgemini and McKinsey all invested in OpenAI’s venture, which is one way to handle a new competitor.

We doubt the labs are in it for services margins. A deployment that reaches production becomes years of model usage, and owning the deployment arm keeps that usage on the lab’s own models. If you sell AI implementation for a living, your newest competitor is the company whose API bill you’ve been expensing.

Thrive Holdings has raised more than $3B since April 2025 to buy accounting and IT services firms and rebuild them around AI. It owns more than 70 businesses, OpenAI is an owner, and its latest round valued it at $12B. Thrive doesn’t disclose revenue, so we can’t put a multiple on that, but nobody values a group of accounting firms and IT shops at $12B expecting them to stay accounting firms and IT shops.

Long Lake committed up to $3.1B of equity to buy Amex GBT, the corporate travel company, in a deal that closed Sept 29. The people booking your flight to the AI strategy offsite now work for an AI rollup. And Sequence Holdings, which started with a $75M seed, has agreed with Dell’s family office to take insurance broker The Baldwin Group private at an enterprise value of about $7.7B. That one isn’t in our total, since we haven’t seen Sequence’s share of the equity disclosed.

So don’t read our $14.7B as a ceiling. Owners borrow and bring partners. Long Lake’s purchase came with up to $1.25B from Koch and $2.75B of committed bank debt on top of its own equity.

Then there are 18 AI-native service firms with about $1.2B between them. They sell the result instead of the hours: a security alert handled at 3 a.m., an H-1B petition for a published flat fee, an estate settled for a percentage of its value. Six sell legal work. Lawyers, we hear, are thrilled. They compete for security, legal, accounting and outsourcing budgets, not AI budgets, and as AI does more of the work, the gain lands in their margin rather than a lower invoice, at least until a competitor cuts prices.

For all the noise, $14.7B is about 2.5% of the $576.5B Gartner expects companies to spend on AI services in 2026 alone. And that’s only what the selling side is chasing. The owning side is after something bigger: what companies already spend on accounting, IT, legal and outsourcing, much of it with incumbents.

Accenture alone pledged $3B for AI back in 2023 and now counts nearly 110,000 AI and data staff. Nobody on our map can out-hire that. They’re betting they won’t need to, because they can serve the same client with far fewer people, or buy the firm that does.

None of these bets is safe. Lab ventures are tied to one lab’s models. Implementers have services margins and a hiring problem. Platforms are software companies running services teams, a sentence that makes software investors itchy. Service firms own the liability when the AI gets it wrong. Rollups have to integrate a lot of acquisitions at once, and every operating partner reading this just winced.

Our bet: the winners own the work and the context around it. The data, the workflows, the judgment calls nobody wrote down. Models get more interchangeable every quarter, so what you own around the model is the part nobody can copy over a weekend. In a typical portco, that context lives in a few people’s heads. Get it out of there before a rollup hires them.

If you’re a PE operator: your value creation plan probably has a slide called “AI adoption.” It needs one called “Who’s about to undercut us.” At the next portco review, ask what your biggest competitor’s cost to serve would be if a rollup bought it tomorrow and ran it on AI. If nobody in the room can answer, you’ve found your next slide.

If you run a PE-backed company: before an AI vendor bake-off, find out whether your sponsor is a partner in one of the lab ventures, so you know who’s on which side of the table. And if you’re in services, the firms selling outcomes at a fixed price are telling you, very politely, where your pricing is headed.

If you’re a consultant: you’re on the selling side, and the model makers just walked in with PE money behind them. The consultants who do fine from here will get paid like owners, with a share of the savings or equity next to the sponsor. (Hi. We know you’re reading this.)

All 38 companies, every number sourced, sorted by how each one makes money, plus who we left off the map and why: https://aimomentum.co/lastmile/. Fair warning, it has footnotes.

Next week: revenue per head rose at 20 of the 26 biggest consulting and IT services firms we tracked. Before anyone thanks AI, we read the fine print.

Hit reply and tell me who’s about to undercut you. The best answers, anonymized, will show up in a future issue.

Warmly,

Jimmy Bijlani

P.S. If someone forwarded this to you, you can subscribe at theservicesrewrite.com.