The consulting market has had two layers for fifty years. The first layer is strategy: McKinsey, BCG, Bain, the firms that sell architecture and recommendation to senior executives at large enterprises. The second layer is industrial-scale integration: Accenture, Deloitte, IBM, Capgemini, the firms that handle large standardized deployments and the implementation work that follows the strategy recommendation. The handoff between the two layers, and the second handoff from the integrator to the customer's internal team, is where these large change programs have always landed. It is also where they tend to fail. Harvard Business School's Michael Beer and Nitin Nohria were already putting the failure rate of corporate change initiatives at about 70 percent in 2000, and the handoff structure is part of the reason: nobody in the chain owns the question of whether the recommended system is actually running in production a year later.
The third layer of consulting that has not been named
A third layer has been forming as AI spreads, and it does not yet have a category name, a standard business model, or a talent pipeline. The sharpest description of it so far is Diogo Santos's April 2026 analysis of the Palantir forward-deployed engineering model, which names no firms but pins down the work. The third layer, in his framing, is the set of teams that "wire AI into live systems, govern it in production, and remain accountable for what happens six months after the platform vendor has moved on." He also draws the boundary: the layer is the only one of the three "willing to operate inside the institutional complexity that neither the strategists nor the integrators are prepared to enter."
Why the third layer is forming now, under AI
This third layer is forming now, after fifty years of stable two-layer market structure, for a specific reason. The eighty-to-ninety-nine-percent problem created the demand, and foundation models created the supply. The institutional gap between strategy and integration finally has a discipline to fill it, because the work is now mostly engineering rather than analysis (the foundation models do the analysis), and the engineering has to happen in the customer's environment because the unwritten rules and the edge cases only show up there. Neither existing layer can do this work, because the strategy firms have the wrong people for it and the integrators have the wrong access.
What a third-layer firm looks like structurally
Structurally, a third-layer firm is a team of three to ten engineers, on the customer's stack, with permission to write to the customer's systems. Outcome-aligned pricing, because the engagement is metered against units of work processed rather than against time billed. A delivery model that compresses discovery into days rather than months, because the engineers can install observation infrastructure on day one rather than spending twelve weeks interviewing stakeholders. A go-to-market motion that does not look like a strategy firm's sales process, because the customer is buying delivery rather than analysis and the sales conversation is about which workflow to ship first rather than about which strategic framework to apply.
How the FDE role spread from Palantir to OpenAI to Ramp
The Pragmatic Engineer documented the spread of the FDE role through 2024 and 2025: from Palantir's internal Delta role (created in the early 2010s) to OpenAI, Ramp, Commure, Matta, and Gecko Robotics. Until about 2016, Palantir had more forward-deployed engineers than software engineers. The model has been quietly proven at scale by Palantir for fifteen years, and is now being adopted explicitly by the venture-scale companies building the new layer of enterprise AI delivery. CRV's guide to the role makes the operationally important point: the model "connects to a land-and-expand sales motion," and the "progression from custom work to core product capability is exactly what the model is designed to produce." Until the core product can absorb what the deployed engineers build, the engineers are the product.
Why boutiques default to the first or second layer
Most boutique consulting firms sit in the first or second layer by default, and the reason explains why the third layer is a critique of the strategy firms and the integrators and, equally, of the boutiques that present themselves as their alternatives. The compensation structure of a traditional consulting firm pays partners and senior consultants for selling and analyzing, not for engineering. The org chart is built around the pyramid (one partner, three principals, ten consultants, twenty analysts), which assumes the work is fundamentally analytical with engineering as a downstream activity that can be offshored. The sales motion is built around producing recommendations that the customer's internal team will execute. Each of these structural commitments maps cleanly to the first or second layer of Diogo Santos's framing. None of them is what the third layer requires.
Building the third layer requires building a different kind of firm. Engineering DNA, where the senior staff are engineers who have shipped production systems rather than consultants who have produced decks. Embedded delivery, where the team works in the customer's environment rather than against an offshore delivery center. Outcome-aligned pricing, where the firm only gets paid when the customer captures the value. Each of these amounts to a different model entirely rather than a minor adjustment to the consulting one. They are hard to bolt onto a traditional consulting structure, because the legacy P&L pulls the firm back into the pyramid that the old structure rewards.
A reasonable counter is that the FDE role is a Palantir-specific staffing model that does not generalize into a category. The argument has a historical version (Palantir was unique, the role does not transfer) and a present version (it has not generalized yet, so it might not). The historical version is being refuted in real time. OpenAI and Ramp both run the same model, ElevenLabs says its forward-deployed engineers have helped hundreds of enterprises launch AI agents, The Pragmatic Engineer documents the spread, and Foundation Capital's services-as-software thesis holds that "forward-deployed engineers have become one of the most strategic assets in enterprise AI companies." The model is generalizing fast, and the firms that recognize this earliest will hold the positions that the firms recognizing it later will not be able to win back.
The third layer of consulting is where the next generation of services companies is forming. It is also the layer that the existing consulting industry cannot reach into, because reaching it requires a different commercial structure, a different talent profile, and a different sales motion. The incumbents have noticed: Accenture merged strategy, consulting, Song, technology, and operations into a single Reinvention Services unit in 2025, a restructuring Julie Sweet framed as making the firm "an even stronger engine of reinvention that more rapidly delivers the power of Gen AI." The question for an operator looking at AI delivery in 2026 has changed. Rather than asking which of the incumbent delivery models to buy, they should ask whether the firm they are talking to is structurally in the third layer or structurally in the first or second. What an engagement in the third layer looks like end to end is its own subject, and a direct description of one is worth more than another definition by contrast.