Why Clients Want This Now: Spending Needs to Pay Off — and Experience Can't Be Siphoned Off by AI

Microsoft's read is that enterprises are long past the "just trying it out" phase. What clients want now is for their AI spending to translate into measurable business results — proof the investment was worth it.

At the same time, enterprises carry another worry. The proprietary data, workflows, and industry experience they've built up over years are what set them apart from competitors. They worry that once this gets absorbed into general-purpose models, it turns into an ordinary capability anyone can call on — flattening their own moat.

Microsoft distills these two concerns into two words: Intelligence (amplifying your intelligence) and Trust (making it trustworthy). Judson Althoff has previously written that these are the two most important pieces of any AI solution, and they form the foundation the entire organization below is built on.

Two Pieces of the Puzzle, Taken Separately: One Banks Your Proprietary Know-How, the Other Keeps Watch Over Your AI System

To deliver both "amplified intelligence" and "trustworthiness" at once, Microsoft says it takes two independent platforms underneath. They have different jobs — let's look at each separately first.

What the 6,000 Experts Actually Do: Keep These Two Pieces Turning

What the 6,000 experts do at their core is keep information flowing in a constant loop between the two platforms above. Data and experience get fed in, the system produces results, those results go back to the experts for optimization, and the cycle rolls forward round after round.

Agentic business processes here means workflows that can complete multiple steps on their own, deciding for themselves which tools to call to get the job done, without a human clicking through every step. What the experts do is make these processes more accurate with every round of feedback.

The Promise, Written Into the Terms: Your Data Won't Be Used to Feed Models Your Competitors Can Also Use

For enterprises, this promise is what determines whether they're willing to hand over their crown jewels. Microsoft lays out the difference between the two approaches clearly.

No societal consensus would allow an AI's future to consume the very intelligence of the companies deploying it.Satya Nadella, CEO of Microsoft (as quoted by Judson Althoff)

Building on this principle, Microsoft's answer is an open, heterogeneous multi-model platform: enterprises shouldn't be locked into a single model, just as they shouldn't be locked into a single technology vendor. Within the same system, each scenario can run the model best suited for it, without handing control to any single one.

Who's Already Using It: London Stock Exchange Lets Analysts Ask AI Directly for Answers

Among the deployment examples Microsoft gives, only London Stock Exchange Group (LSEG) explains the actual mechanics — the rest are just mentioned by name.

Microsoft's engineers and industry experts worked with LSEG to embed AI into its LSEG Workspace, letting finance professionals ask complex questions directly against structured and unstructured financial content and get fast answers. Underneath, it's continuously refined through client feedback and live user testing, with each iteration getting faster and the model's quality and coverage improving bit by bit.

Besides LSEG, Microsoft also names clients including Land O'Lakes, Unilever, and Novo Nordisk, along with global consulting partners such as Accenture, Capgemini, EY, KPMG, and PwC, saying it will lean on these partners to roll this model out across markets and industry verticals worldwide. The official blog doesn't go into specifics for these cases.

Who's Steering — the Numbers at a Glance

Steering this new organization is Rodrigo Kede Lima, who takes on the role of President. He brings 30 years of industry experience and has spent the past 6 years as a sales leader at Microsoft, running enterprise-scale transformations across the Americas and Asia, and helping clients and partners turn technology shifts into business results over the long term.

Microsoft stood up a dedicated team, Microsoft Frontier Company, putting in $2.5 billion and embedding 6,000 engineering and industry experts directly inside client companies. Against the two problems above, it delivers two moves: keeping AI effective over time (through embedded experts continuously refining it — more in the next section), and writing "your data won't be used to feed models your competitors can also use" into a non-negotiable principle, while letting you freely switch models by scenario.

That's what makes it safe to hand over your crown jewels. But how does it actually deliver "lasting results"? See the animated diagram in the next section.

This service's signature move is keeping two things constantly turning against each other: one platform banks your company's proprietary data, experience, and processes (growing thicker the more it's used), while the other keeps watch on how the AI system is performing and whether it's worth it. The 6,000 experts sit embedded in the middle, forming a continuous improvement loop between the two.

Microsoft says it already has results on the ground, but only one case actually explains the specifics: London Stock Exchange (LSEG), where finance analysts no longer have to dig through piles of content themselves and can ask the AI directly for answers. The other clients are named but not detailed. What's left is a string of numbers.