Bloomberg Tech

Judson Althoff on Selling Outcomes, Not AI Adoption

Judson Althoff· CEO of Microsoft's commercial business at Microsoft
·~8 min·English·Bloomberg
AI CompanyBusiness StrategyAgentsAI Infrastructure
TL;DR

Microsoft is standing up a 6,000-person, ~$2.5B Frontier Company unit to push enterprise AI past adoption pilots into outcome-tied deployments, staffed with industry veterans and built on a model-diverse platform that spans 11,000 models.

01Core Mental Model

AI Has to Serve the Business

Althoff's whole pitch reframes enterprise AI from a usage metric into a business result: the job is not to get PhDs adopted, it is to make AI move the outcomes a company already cares about.

AI has to serve the business. It has to serve business outcomes. It's less about deploying PhDs to just simply drive AI adoption, but rather infusing the right level of skill around industry expertise, around change management and continuous improvement.

— Judson Althoff, Bloomberg Tech
Key Insight
The reframe is a shot at rivals: Microsoft is betting the enterprise buyer no longer wants to be sold headcount or model access, but a P&L result. Whoever can price AI against outcomes instead of seats or tokens gets to own the relationship.
02The Bet

The $2.5B Frontier Company Bet

To sell outcomes at scale, Microsoft is standing up a dedicated 6,000-person unit built on three deliberate pillars — the right skill, the right scale, and the right platform.

So we felt it was necessary to assemble a world class team with the right skill, the right scale, and the right platform to drive these outcomes.

— Judson Althoff, Bloomberg Tech
Key Insight
At a stated ~$2.5B for 6,000 people, Microsoft is matching Palantir and the consultancies on bodies while insisting the platform, not the bodies, is the durable edge. The size signals this is a land grab, not a pilot.
03The Reframe

Forward-Deployed: Verb, Not Noun

The interviewer drew the line and Althoff took it: a forward-deployed engineer is a person you place, but forward-deployed engineering is the outcome you methodically drive and infuse into how the work happens.

Sure, we're going to put a lot of great engineers on the ground at customers to help them with AI, but we're first going to be methodical about making sure that the AI solutions that they're building are really driving business outcomes and that are infused into the way they work.

— Judson Althoff, Bloomberg Tech
Key Insight
By splitting the noun from the verb, Althoff quietly concedes that 'forward-deployed engineer' has become a commodity label. The defensible thing is the engineering discipline — evals, change management, a continuous-improvement loop — that a competitor cannot poach one PhD at a time.
04The Differentiator

What's Left Behind Belongs to the Customer

Althoff's answer to enterprise lock-in fears is a hard promise: every artifact the engagement produces — the IP, the data, the semantic context, the evaluation thinking — belongs to the customer, not to Microsoft.

All of that belongs to the customer at the end of the engagement, which is fairly differentiated here.

— Judson Althoff, Bloomberg Tech
Key Insight
Handing back all IP and context directly answers the enterprise's deepest fear of vendor lock-in — but it also binds the customer tighter to Microsoft's platform, because that compounding 'unique intelligence' lives on the platform and keeps improving there.
05The Talent Play

20-Year Veterans, Not Just PhDs

The scarce hire is not the AI researcher but the domain veteran: Frontier Company is staffed with people who spent decades in banking, retail, energy, and life sciences and know which business process is worth automating.

We've got folks that have been in banking for 20 years in retail for 20 years, energy, life sciences, getting to the meat of what customers actually need to achieve with their business.

— Judson Althoff, Bloomberg Tech
Key Insight
Sourcing 20-year industry veterans — and spooking the consultancies who employ them — reveals where the real scarcity sits. AI skill is the easy half to hire; the hard half is someone who understands a business process well enough to know which part is worth automating.
06The Technical Core

Model Diversity: 11,000 Models, One Right Answer

Instead of betting on one frontier model, Microsoft routes each task across 11,000 models — start frontier, optimize, swap to a fine-tuned open-source model — to land the right outcome at the right price and dodge a token-cost explosion.

We support over 11,000 models so that you can start with a frontier model, optimize it, maybe use an open source model, fine tune it, get it down to the right outcome at the right price point to avoid the cost explosion around token yield.

— Judson Althoff, Bloomberg Tech
Key Insight
'Model diversity' recasts the frontier-model arms race as a cost-engineering problem: the win is not the smartest model but routing each task to the cheapest one that still hits the outcome — which conveniently makes Microsoft the indispensable neutral router across 11,000 models rather than a bet on any single one.
07The Moat

More Than FTE: The Platform Is the Moat

Althoff applauds Palantir for the forward-deployed-engineer playbook, then argues Microsoft's edge is 'more than FTE' — a left-to-right platform stacking Copilot, Teams agents, model routing, and closed-loop observability on top of the people.

You have to have this left to right platform that allows AI to empower human ambition and do it in a model diverse, open and heterogeneous way across every layer of the stack.

— Judson Althoff, Bloomberg Tech
Key Insight
Applauding Palantir while insisting on 'more than FTE' is the tell: Microsoft concedes it cannot out-Palantir Palantir on forward-deployed engineers, so it changes the game to platform breadth — Copilot, Teams agents, model routing, observability — where its incumbency, not its headcount, is the moat.