The Vendors Just Named Our Category
§1. Two structural bets in two weeks
In the same fortnight of May 2026 the AI category committed somewhere north of five billion dollars to one specific delivery shape.
On 4 May, Anthropic announced a joint venture worth more than 1.5 billion dollars with Blackstone, Hellman & Friedman, and Goldman Sachs to embed Anthropic engineering resources inside a standalone enterprise services firm. One week later, on 11 May, OpenAI launched OpenAI Deployment Company with more than four billion dollars in committed capital led by TPG, with Advent, Bain Capital, and Brookfield as co-lead founding partners. As part of the launch OpenAI acquired Tomoro, an Edinburgh-based AI consulting and engineering firm, bringing roughly 150 experienced Forward Deployed Engineers and deployment specialists to the new entity on day one.
Eight days later, on 19 May, KPMG announced a global alliance with Anthropic framed around portfolio transformation and the design of AI operating models across PE-backed holdings. The Big Four were the visible bridge from the vendor’s engineering capacity to the buyer’s fund-level transformation program.
The delivery shape these bets are funding has a name. Forward Deployed Engineer at Palantir, Applied AI Engineer at Anthropic, Forward Deployed Engineer at OpenAI. Same function in all three cases. An engineer embedded inside a single enterprise account, writing the customer-specific integration code, designing the evaluations against the customer’s compliance bar, and operating the production system through change-management. Top-end loaded compensation runs $300,000 to $600,000 per engineer per account, according to publicly listed roles. Across thirty-nine AI companies there are now 224 open FDE roles.
This is the shape that costs five billion dollars to fund. It is also the shape that has a structural ceiling, and the ceiling is exactly where the portfolio operating partner spends most of their time.
§2. The ceiling is the labour-arbitrage line
An Applied AI Engineer embedded full-time inside a single Fortune-500 account costs the vendor between $300,000 and $600,000 in loaded compensation each year. The account is paying somewhere between several million and several tens of millions of dollars in annual program spend. The engineer’s compensation amortises. The unit economics work because the deal size makes them work.
Below that deal size the math inverts. A fifty-million-EBITDA dermatology roll-up cannot absorb a half-million-dollar embedded engineer at the marketing layer. Neither can a hundred-and-fifty-million-EBITDA dental DSO at the patient-acquisition layer, or a two-hundred-million-EBITDA behavioural-health platform at the intake-triage layer. The engineer cannot be billed against the deal. The model cannot stretch.
Anthropic has been clear about where the line is. In the public positioning of its Applied AI Engineer function, the company states that the team focuses on financial services, healthcare, legal services, and government as the four regulated big-enterprise verticals where the compliance perimeter requires the embedded engineer. The same positioning continues with one sentence that is structurally consequential for every operating partner reading this.
“Less regulated verticals like e-commerce or media tend to be served by partner networks and self-serve API tooling instead.”
Read it as a sentence the vendor wrote about its own roadmap. Below the Fortune-500 regulated tier the vendor will not field engineers directly. The customer base it cannot reach gets routed to partner networks and to self-serve APIs.
The PE-backed mid-market is on the wrong side of that line by deal size, but on the right side by regulatory perimeter. Healthcare MSOs carry HIPAA on every patient touch. Dental DSOs add state board ownership rules. Behavioural-health platforms add 42 CFR Part 2 on top of HIPAA. Legal boutiques and boutique financial advisory carry conflict-of-interest and KYC perimeters at their own deal sizes. The compliance discipline is the same as the Fortune-500 regulated tier; the deal size is not. The vendor knows it cannot serve this segment with its own embedded engineers and has written that admission into its public-facing positioning.
§3. The Big Four are the visible bridge, not the substitute
The KPMG-Anthropic alliance is what the bridge looks like at fund level. The Big Four sells the fund a transformation program. Design of the operating model, sequencing of capability roll-outs across the portfolio, change-management, and a multi-year service contract that scales with the portfolio. The program is priced at fund level and sits at the strategy altitude. The deliverable is a program.
The program is the right work for the fund to commission. It tells the portfolio what operating model to run. It does not staff Monday’s work at the first portfolio company.
Across the portfolio of multi-location roll-ups, the staffing problem is real and named by the operating partner data. The FTI 2026 Private Equity AI Radar, published 19 May 2026 from a survey of 200 fund and operating leaders, reports that ninety-five percent of funds say AI initiatives are meeting or exceeding their original business case criteria, and that talent remains the primary constraint to scaling adoption, cited by thirty-five percent of respondents. The constraint is not whether AI works. The constraint is who staffs and governs it at each portfolio company.
A fund that has signed a Big Four transformation program has the answer to “what should be done.” It still does not have the answer to “who is doing it at the first portfolio company on Monday.” The vendor’s own FDE team is structurally not the answer at PE-portco deal size. The partner-network sentence from Anthropic’s own positioning is the answer.
§4. What the partner-network half of the sentence looks like in practice
Across twenty-four operator and buyer subreddits sampled in the parent research arc, one category was empty. The managed operations partner that delivers governed marketing-and-operations outcomes per portfolio company, on the architecture the fund underwrites once and deploys many times, did not have a vendor occupying the square.
The shape that fits the empty square has four properties.
First, it is sold as outcomes consumed rather than as engineers assigned. The portfolio company does not learn a new platform. The operator approves the work the system surfaces. The operating partner gets a portfolio-wide view of what shipped and what was governed.
Second, it runs one operating standard across the book. Same per-tenant scoping. Same named-approver gates on anything that publishes, sends, or touches PII. Same replayable audit trail across messaging channels. Same posture across every portfolio company that adopts it. The audit trail that defends each business to its CFO is the artifact the LP and the diligence team need at exit.
Third, the labour-arbitrage is in the governance layer, not in the engineer-per-account model. One operator runs three to five portfolio companies at portfolio-marketing-operations scope because the recipes, the approval gates, and the channel chassis are templated and operated centrally. The unit economics are governed by the leverage of the governance layer, not by the compensation of the embedded engineer.
Fourth, it complements the Big Four program by occupying the execution altitude the program does not. The program designs the operating model. The Operations Partner staffs the seat at the first portfolio company on Monday and produces the first month of governed work against the requirement the program implied.
§5. The first portfolio engagement is the same shape as the open requisition
Every portfolio company has a marketing requisition that has been open between sixty and a hundred and twenty days. The hiring manager has lost a candidate and the search is moving slowly because the brief is broad. The responsibilities section is a list of the work the company expects from the seat. The salary band is a pre-approved budget envelope, with benefits and on-costs already loaded in. The level is the seniority gate, signed off by Finance and HR. The hiring manager is the named sponsor.
The requisition is a Statement of Work, written, with a budget envelope already cleared by everyone who has to clear it. The full mechanics are in the companion essay published yesterday, The Job Description Is the SOW. The relevant detail for this essay is that the trial mechanic and the open requisition are the on-ramp into the empty square. The fund picks one portfolio company. The Operations Partner staffs the seat on Monday. The first two weeks are at no cost. The hiring manager grades the work against the requisition’s own responsibilities. The Big Four program upstream tells the fund what model to run; the Operations Partner ships the first month against it.
§6. What this means for the value-creation plan
When acquisition velocity stops working, integration is the value-creation lever left, and integration is an operating capability rather than a transaction. The earlier essay in this arc, Integration Is the Moat, traced that argument in detail. The May 2026 vendor-side moves do not change the argument; they confirm it at the largest possible scale. The frontier vendors have committed nine-to-ten figures of capital to making managed deployment the durable value layer above the model. They have committed at the Fortune-500 regulated tier and stated publicly that they will not deploy that team into the mid-market.
For the portfolio operating partner the read is straightforward.
The transformation program commissioned with the Big Four partner sits at strategy altitude and is the right work for that altitude. The frontier vendor’s own engineering team is structurally not deployable into the portfolio companies. The execution layer the program implies is the layer the Operations Partner is built to occupy. The unit economics work because the labour-arbitrage sits in the governance and operations layer rather than in an embedded engineer per portfolio company.
The diligence frame is the same one Integration Is the Moat named. At exit the question the diligence team will ask is whether the portfolio’s AI operations are evidenced and governed in a form that survives review. A portfolio that ran governed marketing-and-operations outcomes across its holdings through a single operating partner, with one audit trail and one approval posture, answers that question. A portfolio that re-funded a separate integration project at every acquisition and ended up with N fragmented stacks does not.
§7. The category just got named
For eighteen months the phrase “AI Operations Partner” described a category that was not legible at the operating-partner level of vocabulary. As of the second week of May 2026 it is. The frontier vendors have named the delivery shape, capitalised it, and admitted publicly the segment they cannot serve directly. The partner-network half of that admission is the empty square the Operations Partner fills.
The first portfolio engagement is the open requisition. The trial mechanic and the offer detail are in the Job Description Is the SOW essay and its one-page operator brief at jiegou.ai/downloads/job-description-is-the-sow-v1.pdf. Pick the portfolio company. Send the JD. The work starts Monday.