Microsoft launched a new business on July 2, 2026 called Microsoft Frontier Company, committing $2.5 billion to place its own engineers, industry specialists and salespeople inside customer organizations to build, deploy and continuously improve AI systems. The bet behind it: the value of enterprise AI is won not in the model, but in the messy work of deploying it.

What Frontier does

About 6,000 experts — drawn mostly from existing Microsoft teams — will embed at customer sites under a "forward-deployed engineering" model, co-designing and running AI tied to measurable business outcomes. Microsoft frames it around two pillars it calls "Intelligence + Trust": amplify a customer's capabilities with AI while protecting their data and IP so "none of it is used to train models in ways that commoditize what differentiates them." The platform is model-agnostic and multi-cloud, supporting OpenAI, Anthropic, Microsoft and open-source models. Named early customers include LSEG, Unilever, Land O'Lakes and Novo Nordisk; delivery partners include Accenture, Capgemini, EY, KPMG and PwC.

Why now

The pitch targets a stark statistic: an MIT study widely cited in coverage found that roughly 95% of enterprise generative-AI pilots deliver no measurable profit impact. Microsoft says customers have moved "well beyond experimentation" and want returns. The unit is led by Rodrigo Kede Lima — a roughly 30-year veteran, most recently president of Microsoft Asia — as president, reporting to Judson Althoff, CEO of Microsoft's commercial business.

An industry land-grab

All four major AI vendors moved into forward-deployed engineering within weeks. Amazon committed about $1 billion two days earlier, on June 30; OpenAI launched a "Deployment" company and bought the consultancy Tomoro; and Anthropic struck a roughly $1.5 billion partnership with Goldman Sachs, Blackstone and Hellman & Friedman. Microsoft's $2.5 billion is the largest of the group. The model itself was pioneered about two decades ago by Palantir.

The skeptics

Critics call it rebranded consulting: firms like Accenture and EY have embedded enterprise transformation for decades, and Microsoft is now partnering with the very firms whose model it is copying. Other doubts: services carry lower margins than software and could dilute Microsoft's economics; the 95% failure rate may be a data-foundation problem that bodies on-site won't fix; and the plan risks channel conflict with the system integrators Microsoft depends on, while squeezing venture-backed AI-implementation startups now competing with a bundled incumbent. "This goes beyond what has been labeled as Forward-Deployed Engineering, and will be the largest, most capable, outcome-driven engineering organization in the industry," Althoff said.