A late-April 2026 McKinsey study crystallised a growing unease: by the end of 2025, nearly nine in ten companies had deployed AI in at least one business function, yet 94% reported no significant benefit from those expenditures. The finding, viewed from Washington and Silicon Valley, has accelerated a strategic pivot among the largest cloud providers away from simply renting out models and towards embedding engineering teams deep inside customer operations.

Microsoft announced on Thursday the creation of Microsoft Frontier Company, a new operating entity backed by $2.5 billion and staffed by 6,000 industry and AI engineering experts. The unit will help clients such as Unilever and Novo Nordisk select, integrate and fine-tune AI tools—from Microsoft and third parties—with their proprietary data. Critically, customers retain the resulting intellectual property rather than feeding it back to the vendor. Judson Althoff, CEO of Microsoft Commercial Business, told Reuters the move was partly born from the company’s own experience: “Three years ago, when we built Copilot, we made a mistake by binding it to OpenAI models only.” The new organisation explicitly enables model-swapping across frontier systems from Google, DeepSeek and others.

The initiative mirrors a $1 billion Forward Deployed Engineering unit launched days earlier by Amazon Web Services, which will also dispatch thousands of engineers to client sites. Both moves revive a model pioneered over a decade ago by data analytics firm Palantir Technologies and recently adopted by AI labs OpenAI and Anthropic, which began placing their own engineers with enterprise customers this spring. Analysts in North America note that large corporations increasingly suspect that relying on a single frontier lab could eventually equip those labs to compete with them in fields such as coding and law.

The push comes as Microsoft’s shares have shed nearly a quarter of their value since January, despite strong cloud growth, and the company cut roughly 15,000 jobs in 2025 with further reductions expected. The embedded-engineering strategy is designed to shorten the time to measurable return on AI investment by co-designing systems that compound a client’s proprietary intelligence over time. The next factual milestone will be the operational rollout of these units and the subsequent enterprise AI adoption metrics reported by Microsoft and AWS in coming quarters.