Amazon this week launched a $1 billion Field Deployment Engineering organization, embedding engineers inside client companies to deploy purpose-built AI agents. This follows identical moves by OpenAI and Anthropic. The framing is always the same: fast deployment, customer self-sufficiency. The reality is more interesting: the largest AI companies have decided the bottleneck is not the model, it is the organizational tissue around it. So they are buying their way inside.
The Consultant Is Dead, Long Live the Agent
McKinsey and Accenture built trillion-dollar businesses on the premise that companies need outside expertise embedded temporarily to solve structural problems. The FDE model is that, but with a subscription to the infrastructure underneath. When an Amazon engineer embeds in your logistics operation and builds agents that run on AWS, the dependency relationship is not temporary. It is architectural. A 2026 arXiv paper by Zhang et al., Internalizing the Future: A Unified Agentic Training Paradigm, captures the technical ambition: agents that internalize world models to plan sequentially, not just respond. The embedded engineer is deploying agents that are meant to outlast the engineer's visit.
Self-Sufficiency as Product Strategy
Amazon's stated goal of customer self-sufficiency is the tell. In consulting, self-sufficiency is the enemy of repeat revenue. In infrastructure, it is the product: once the agents are running on your cloud, maintaining them requires more cloud. TurboFund's live VC intelligence has been tracking the surge in enterprise AI deployment startups trying to carve a niche before the hyperscalers absorb the category entirely. The race is on, and the prize is not the model. It is the org chart.