Three AI vendors have independently settled on the same answer to the enterprise AI integration problem — sending their own engineers to embed inside customer companies.
Amazon, OpenAI, and Anthropic are all building the same thing from different angles: a corps of engineers embedded inside customer companies to install AI systems and leave. AWS committed $1 billion to the idea on Tuesday, two months after OpenAI and Anthropic made similar moves through private-equity-backed joint ventures. What looks like a race is actually three different structural bets on who bears the cost of the AI integration gap.
The model has a name. Forward-deployed engineers, or FDEs, are senior technical staff from a vendor who temporarily move into a customer's organization to build, install, and hand off a working system. The model traces back to Palantir, which used FDEs to install data and analytics systems inside government and corporate clients over the past decade. In practice, an FDE engagement looks like a small strike team spending weeks or months on-site: mapping the customer's data and workflows, building the first version of the system, training the customer's team, and leaving the customer with skills and patterns to run it themselves.
Amazon is betting that enterprise customers cannot do this alone. On June 30, 2026, AWS announced a new internal organization of AI-focused FDEs dedicated to "purpose-built agents" and customer self-sufficiency, authored by Francessca Vasquez, AWS's VP of Professional Services and Agentic AI. The $1 billion is a commitment of internal Amazon resources: headcount, training, infrastructure. It is not an outside investment vehicle or fund, which makes it structurally different from what its rivals are doing.
OpenAI and Anthropic went a different route. In May 2026, both launched enterprise services through joint ventures with private equity partners: OpenAI at roughly $4 billion in committed capital and Anthropic at $1.5 billion. Those structures let the labs share the financial and operational load of running embedded engineering corps while tapping customer relationships and capital that the labs themselves do not have. AWS is keeping it in-house, paying for the engineers and the engagements directly out of Amazon's own books.
Three companies, three capital structures, one customer problem. The convergence is the news, not any one announcement. Enterprise buyers keep telling AI vendors that they cannot integrate frontier models into their own systems. The data is fragmented, the workflows are bespoke, and the regulatory environment is uneven. The vendors have collectively decided that the only credible answer is to send their own people in to do it. A Computerworld Q&A on AWS's FDE practice describes the engagement shape explicitly: embed, build, hand off, then leave the customer running it.
That bet carries two criticisms worth preserving. First, an FDE corps is labor-intensive in a way that does not scale cleanly. A vendor must keep a full traveling bench of senior engineers on staff even between deployments, which means the unit economics only work if each engagement is large enough to keep those engineers billable most of the time. Second, the existence of this market is an implicit admission that current AI tools and APIs are not yet drop-in: if they were, customers would not need vendor engineers living inside their companies for months at a time. Both criticisms are real constraints on the model, not growing pains.
What to watch is how each structure performs. AWS is betting that owning the whole deployment relationship inside Amazon is worth the capital intensity. OpenAI and Anthropic are betting that PE capital and JV structures can carry the load and let the labs stay focused on models. If PE-backed deployments scale faster or convert more customers, expect more frontier labs to copy that shape. If internal orgs scale more cleanly under direct corporate control, expect AWS's structure to spread. Either way, a procurement pattern that did not exist two months ago is now part of the standard enterprise AI offering.