Between January and September 2025, monthly job listings for forward-deployed engineers rose by more than 800 percent, a figure the Financial Times reported and one that tells you the model has stopped being a Palantir oddity and become a category that several very different kinds of company now compete in. An operator weighing an AI delivery firm in 2026 is no longer choosing between a consultancy and a software vendor; they are choosing among firms whose models differ so much that the word "forward-deployed" now hides more than it tells you. This post sorts those firms, not to relitigate what the third layer is, which has been written, but to locate the specific kind of firm an operator running a mid-market business should be looking for, and to set out who else is competing for the same work.
The incumbent who wrote the role
Palantir invented the forward-deployed engineer, and for most of the company's history it had more of them than it had software engineers, which is the single fact that explains why everyone else is now copying the playbook rather than treating it as a curiosity. The role was always a way to push a product into an environment too institutionally complicated for a sales engineer and a slide deck, and the customers were governments and the largest enterprises on earth. That heritage matters because it set the default assumption that forward-deployed delivery is something you do at the top of the market, against contracts large enough to absorb a resident engineering team for a year. The model works there, and the open question the rest of the post answers is whether it works anywhere else. The answer turned out to be yes, which is what made this a contested category rather than one company's staffing quirk.
The model labs growing their own arms
The most visible new entrants are the foundation-model labs themselves. OpenAI, Anthropic, and Cohere are all standing up forward-deployed teams, per the Financial Times, because the labs have discovered the same thing every infrastructure company eventually discovers, which is that the model is worth little without the deployment, and a customer cannot capture the value of a frontier model without engineers who will sit in the environment and make it run. Crunchbase News called 2025 the year the role spread well past Palantir, and the lab-owned teams are the most prominent part of that spread. For an operator, though, the lab teams are the wrong fit, since they exist to drive consumption of one vendor's model, they are staffed against the large-enterprise accounts that justify a lab's attention, and they are structurally incapable of recommending against their own model when the workflow would be better served some other way. They are forward-deployed, and they are not the firm a mid-market business should call first.
The vertical AI-native players
A second group is made of AI-native startups using what a16z has named services-led growth, where a company ships a product but wraps it in enough hands-on delivery to actually get the first customers into production. Emergence Capital, whose July 2025 essay carried the title that captured the thesis, points to portfolio names like Federato in insurance and Bland in voice as examples of the pattern working inside a vertical. Emergence also cites the sobering figure that roughly one percent of companies are at full AI maturity, which is the demand-side reason the services wrap exists at all. These firms are real and many are excellent, but each one is a point solution: Federato is built for insurance underwriting, not for a distributor's month-end close or a manufacturer's accounts-payable queue, and the operator whose problem does not match the vertical the startup was built around gets no use from it. The vertical players win where the problem is exactly their problem, which leaves the horizontal middle of the market, where most operating businesses sit, unserved.
The services firms and the capital backing them
The third group is the one closest to flowscope, and it is the one venture capital has started writing about directly. 8VC's essay on the AI services wave describes a cohort of tech-enabled services firms, many founded by Palantir alumni, that combine private-equity-style inorganic growth with the delivery model rather than scaling purely through software. Dedicated capital has arrived to fund exactly this: Forward Deployed VC, a 2025 fund led by Mark Scianna and backed by alumni of Palantir, SpaceX, and Anduril, exists specifically to bet on companies that put engineers inside customer environments. A fund named after the model is the clearest signal that the category has settled. What the 8VC cohort and flowscope share is the conviction that the deliverable is a running agent rather than a recommendation, and where they diverge is the customer, since the roll-up firms often chase the same large or roll-up-adjacent accounts that suit a private-equity growth engine, which is a different thing from sitting horizontally across the mid-market.
Where the independent mid-market implementer sits
Put the three groups together and the gap becomes obvious. The lab arms are vendor-owned and aimed at large enterprise, the vertical startups are independent but narrow and useful only when your problem is their problem, and the services-and-capital cohort is independent and horizontal in ambition but often pulled toward the large accounts that fit a private-equity thesis. The position that stays open is the independent firm, loyal to no single model, working horizontally across the kinds of operating businesses that run on QuickBooks and legacy ERPs and people doing manual work every month, sized to engagements that a mid-market company can actually buy. That is the position flowscope holds. Being independent means we can recommend against a model when a workflow is better served another way, which a lab arm cannot; being horizontal means we are not waiting for your problem to match a vertical we happened to build; and being mid-market by design means the production-grade reliability work gets done for companies the rest of the field passes over. The scaling economics that let a services-software firm reuse agents across customers are what make that position viable rather than charitable.
A reasonable counter is that the map is converging, that the labs will move down-market, the vertical players will broaden, and the services firms will eventually cover the middle, so the open position is temporary. There is something to it, and convergence will happen at the edges, but the structural incentives run the other way. A lab arm is paid to sell its parent's model and cannot become model-agnostic without undermining the business that funds it; a vertical startup that broadens horizontally loses the domain depth that won it the vertical in the first place; a private-equity-shaped services firm answers to a growth engine that rewards larger accounts. Each of those incentives holds a firm in place, so the independent, horizontal, mid-market position is not a gap waiting to close but a distinct shape of firm, and the operator who understands these categories can tell which one they are actually talking to.