Microsoft launches Frontier Company for AI delivery

Microsoft launched Frontier Company to build enterprise AI systems while protecting customer data and IP.

Microsoft is putting $2.5 billion behind a new business called Microsoft Frontier Company, and it says 6,000 industry and engineering experts will work directly with customers. The pitch is simple: help companies move from AI experiments to measurable business outcomes without handing over their proprietary intelligence.

Microsoft is turning AI consulting into an operating business

The announcement comes from Judson Althoff, CEO of Microsoft Commercial Business, in a post on the Microsoft Official Blog. Frontier Company is not being framed as another services arm. Microsoft describes it as a new operating business built around “Frontier Transformation,” a term the company uses for AI programs that keep improving after deployment.

That matters because a lot of enterprise AI today still gets stuck in pilot mode. Teams test copilots, chatbots, and document tools, but the systems often stop short of changing how work gets done. Microsoft is saying the next step is deeper: embed engineers and domain specialists inside customer organizations, then keep tuning the system against real business results.

The company says this model combines three things that usually live in separate teams: industry knowledge, change management, and enterprise AI engineering. In practice, that means Microsoft wants to be involved not just in model selection, but in workflow redesign, rollout, and continuous improvement.

• $2.5 billion is the headline investment.
• 6,000 experts will be embedded with customers.
• The company says the work will focus on measurable outcomes.
• Microsoft says the business will operate globally through partner channels.

Microsoft’s core promise is protection, not extraction

The most interesting part of the post is not the size of the investment. It is the line Microsoft keeps repeating: a customer’s IQ is protected. In Microsoft’s framing, that means a company’s proprietary data, workflows, expertise, and decision-making patterns should improve over time without being used to train models in a way that weakens its edge.

That is a direct response to a concern every serious enterprise buyer has now. If AI systems learn too much from a customer’s internal data, who owns the upside? Microsoft’s answer is that the customer does. The company says Frontier Company will sit on top of a model-diverse AI platform so organizations can choose the right model for each task without locking themselves into one vendor.

Judson Althoff also ties this to two ideas he says matter most in enterprise AI: intelligence and trust. Intelligence comes from the customer’s own data and processes. Trust comes from governance, security, observability, and FinOps controls that help teams measure return on investment across the stack.

“There is no societal permission for an AI future that eats the intelligence of the companies it’s deployed inside.” — Satya Nadella

That quote, which Althoff cites in the post, is doing a lot of work. It draws a bright line between AI that helps a company compound its own expertise and AI that quietly turns that expertise into a vendor’s training fuel.

The first customer examples show what Microsoft wants to sell

Microsoft points to early work with LSEG, where engineers and industry experts embedded AI into LSEG Workspace. The goal is to let finance professionals ask complex questions and get fast answers across structured and unstructured financial content. Microsoft says the system is refined through client feedback and real-time testing, which shortens iteration cycles and improves quality over time.

The company also names Land O’Lakes, Unilever, and Novo Nordisk as customers already seeing measurable outcomes. Microsoft does not give public numbers for those projects in the post, but the pattern is clear: it wants Frontier Company to be the team that turns AI from a demo into a working business process.

That is also why the company is leaning on partners. Microsoft says it already has FDE partnerships with Accenture, Capgemini, EY, KPMG, and PwC, plus others. That partner layer matters because Microsoft is clearly trying to scale a labor-heavy model without making every deployment a custom one-off.

• LSEG is the clearest public example in the post.
• The use case centers on finance professionals and mixed financial content.
• Microsoft says the system improves through feedback and live testing.
• Named customers also include Land O’Lakes, Unilever, and Novo Nordisk.

What this means for enterprise AI buyers

Frontier Company tells us where Microsoft thinks the enterprise AI market is headed. The easy money is gone. Buyers want systems that cut cycle time, improve decisions, and show a return. They also want to know that their own data will not become a hidden subsidy for the vendor’s model training.

Microsoft is betting that the winning pitch is an end-to-end package: people who know the industry, engineers who can build agentic workflows, and a platform that lets customers swap models without losing control. That is a stronger story than “we have a chatbot.” It is also a harder business to run, which is why the 6,000-person investment matters.

Rodrigo Kede Lima, who is taking over as president of Frontier Company, brings 30 years of industry experience and six years at Microsoft leading enterprise transformations in the Americas and Asia. That choice signals that Microsoft wants a business leader who can sell AI outcomes, not just technical capability.

The real test is whether Frontier Company can keep its promise when deployments get messy. If Microsoft can show repeatable gains in finance, manufacturing, healthcare, and consumer goods while keeping customer IQ protected, other vendors will have to answer the same question: can your AI improve my business without taking ownership of my advantage?

For now, the best takeaway is practical. Enterprise teams should ask every AI vendor the same two questions: what measurable outcome will this system improve, and who owns the intelligence it learns from? Microsoft just made those questions harder to dodge.