The Engineers Are Already Inside. Where Is Your CHRO?

There is a quiet restructuring happening inside large enterprises right now, and the people who should be leading it are not in the meeting. OpenAI, AWS, and Anthropic have begun racing to embed forward-deployed engineers directly inside organizations, redesigning HR and operational systems from the inside out — and doing so, in most cases, without the CHRO involved. HR Executive reported this month that even the Pentagon is part of this pattern. When the U.S. Department of Defense is restructuring its workforce systems with vendor engineers in the room and its own people executives are not, you have a signal worth taking seriously.

This is not a technology story. It is a power and governance story. The firms winning the embedded-engineer race are not selling software — they are acquiring institutional knowledge, redesigning workflows, and setting the architectural defaults that will govern how your workforce operates for the next decade. Once those defaults are set by an outside engineering team optimizing for their platform's capabilities, reversing them is expensive and slow. The question for every senior HR or operations leader is not whether this is happening in your industry. It is whether it has already happened in your company without your knowledge.

The performance data emerging around AI deployment makes this governance gap more urgent, not less. A study from KPMG and the University of Texas at Austin found that workers who actively direct and refine AI output significantly outperform peers who simply delegate tasks to it — even when all other skill sets were identical. That distinction — director versus delegator — is now a measurable performance variable. It is also, critically, a design variable. The systems those embedded engineers are building will determine whether your workforce is trained and incentivized to direct AI or merely hand off to it. That is a workforce design decision. It belongs in your lane, not theirs.

The pressure to cede that lane is real. Oracle, Amazon, Cloudflare, and Block have all cited AI in 2026 workforce cuts, and as HR Executive noted, the reasoning behind each is different — some are genuine automation displacement, some are cover for cost-cutting, and some are competitive repositioning. The problem is that when layoff announcements all arrive draped in AI language, it becomes easy for boards and CFOs to treat workforce restructuring as a technology function rather than a human capital function. That framing shift has consequences. It is precisely the context in which vendor engineers get invited in and HR leaders get left out.

McKinsey's research on the human advantage in an AI economy is relevant here, though perhaps not in the way it is usually cited. The report argues that tech investment alone will not deliver competitive edge — organizations need systems that strengthen both cognitive capacity and AI-era skills simultaneously. The implication that tends to get lost: building those systems requires someone with workforce design authority to own them. If that person is not in the room when the AI architecture is being built, the cognitive scaffolding gets designed by default, not by intent.

Meanwhile, the talent pipeline feeding these decisions is shifting in ways that compound the problem. In London alone, 46% of workers — approximately 2.4 million people — are already in occupations exposed to generative AI, with 313,000 roles at direct risk, according to a 2026 Greater London Authority analysis. The Financial Times has separately argued that universities need to arm graduates with AI evaluation skills — the ability to test models against real organizational needs — rather than just AI fluency. Both data points point to the same gap: the workforce entering and inhabiting these systems needs a fundamentally different kind of preparation than most organizations are designing for.

Which brings the argument back to its sharpest point. Embedded vendor engineers are not malicious actors. They are doing what they are incentivized to do: deploy their technology effectively and deeply. The problem is structural. When workforce architecture decisions get made without workforce leadership in the room, the result is systems optimized for platform stickiness rather than human performance. The defaults get set. The contracts get signed. The organizational muscle memory forms around someone else's design choices.

The window to establish governance over how AI systems are designed into your workforce is not infinite. If your organization has a forward-deployed engineering relationship with any major AI vendor, the right question for this week's leadership meeting is simple: who from the people function is in that room, and what authority do they actually hold?