The Vendors Just Told You Where AI ROI Is Won. Now Measure It.

When the largest software vendor on earth spins up a $2.5 billion subsidiary to sit inside your building and prove AI works, that is a signal worth reading carefully. On 2 July 2026, Microsoft announced Microsoft Frontier Co. The company said it is investing $2.5 billion and moving 6,000 employees into the unit to be embedded with clients, in a practice known as forward deployed engineering (Source: CNBC, cnbc.com). It is led by Rodrigo Kede Lima, who previously ran Microsoft’s Asia business (Source: CNBC).

The timing tells the story. Microsoft’s stock is down 21% this year, and the Microsoft 365 Copilot assistant has not achieved broad enterprise adoption (Source: CNBC via Technobezz). The move landed two days after Amazon committed $1 billion to its own Forward Deployed Engineering organization on 30 June (Source: CNBC, cnbc.com). OpenAI and Anthropic launched comparable ventures earlier in the year, reported at roughly $4 billion and $1.5 billion (Source: American Bazaar, americanbazaaronline.com). Four of the biggest names in AI are now spending real money on the same admission: model access was never the bottleneck. Deployment is.

What the vendors are actually conceding

The honest subtext here is that buying the tools was the easy part. Amazon’s own framing is that the constraint is deployment capacity, with an initial pod of five or six engineers embedded in a customer to get systems into production (Source: CNBC, cnbc.com). Microsoft’s Judson Althoff, CEO of its commercial business, described the most success coming from a methodical approach to building out an intelligence platform (Source: CNBC).

This matches the evidence that started the reckoning. MIT’s State of AI in Business 2025 found that 95% of enterprise generative AI pilots delivered no measurable impact on profit and loss, and only 5% created significant value, against $30 to $40 billion in spend (Source: MIT via Forbes, forbes.com). The same body of research found that over 80% of organizations have piloted tools while only a fraction reach production at scale (Source: MIT via Forbes).

So the vendors are correct that outcomes get won in deployment. What they are quieter about is that outcomes also get proven in deployment, and proof is a measurement problem, not an engineering one. Embedded engineers can build the system. They cannot tell you, on their own, whether your people actually use it, whether it saves time, or whether you are paying for seats nobody touches.

What to measure once the engineers arrive

If you are about to host a forward deployed team, or you already run a stack of AI tools, here is the short list worth tracking. None of it requires waiting nine months for a case study.

Active adoption, not licences bought. Count the share of provisioned seats used in a meaningful way in the last two weeks, by team. Licences purchased is a spend number. Weekly active use is an outcome number. Good looks like adoption climbing and holding above 60% within a quarter for a tool you are paying full price for. Flat or falling use after month one is your early warning that a rollout is stalling.

Time saved, self reported and specific. Ask people how much time a tool gives back per week, and on which tasks. Self reported time is imperfect, but tracked consistently across a whole team it shows direction and size. Watch for the gap the MIT work flagged, where employees double check outputs so heavily that the promised gains never land. If reported time saved is falling while usage rises, you have a trust problem, not a productivity win.

Wasted spend. Multiply unused or barely used seats by their cost. This is the single number that turns a vague worry into a defensible line for a CFO. Good looks like waste under 10% of your AI budget. Anything higher is a renewal conversation waiting to happen.

Shadow usage. The MIT research described a shadow AI economy where people use personal tools even when official pilots fail (Source: MIT via Forbes). If your team leans on unsanctioned tools, that tells you where the real value sits and where your sanctioned rollout is missing.

Read these together, by team, over time. One tool can be thriving in finance and dead in legal. A blended company average hides exactly the detail you need to act on.

The measurement-first point stands whether or not you ever hire a forward deployed team. When a vendor arrives to prove outcomes, you want your own baseline in hand first, gathered independently, so the scorecard is not marked entirely by the people being paid to deliver a good grade. That is the difference between a number you present to your board and a number you hope is true.

The vendors have told you where the game is played. The next move is having numbers of your own. Start a GAiGE trial and get a real read on adoption, time saved and wasted spend across your AI tools in a couple of weeks.