Microsoft Just Bet $2.5 Billion That Software Alone Isn't Enough

2026-07-21 — marketing implementation consulting

I watched a client open their Copilot license. They didn't know what to do with it. Within three months, nobody was using it.

This is the thing Microsoft can't say out loud, so they announced a $2.5 billion company instead.

On July 2, 2026, Microsoft created Frontier Co. The idea: deploy 6,000 engineers, trainers, and consultants directly into enterprise client organizations for six to twelve months. They work as virtual employees. They sit in your meetings. They redesign your workflows. They train your staff. They stay until the work actually sticks.

The subtext is brutal.

MIT's Project NANDA found that 95% of enterprise AI pilots show zero measurable impact on the bottom line. Microsoft's own products aren't immune. Copilot for 365 never became the standard tool it was supposed to be. GitHub Copilot has been losing ground to newer coding tools, more or less steadily. The company realized something: selling the software was the easy part. Getting people to actually change how they work is the hard part.

Rodrigo Kede Lima, formerly president of Microsoft Asia, is running the subsidiary. The model itself has a name—forward deployed engineering. Palantir did this for government and defense contracts. Microsoft is now running it at a scale nobody else has attempted. Six thousand people. Embedded inside client organizations. For up to a year.

Accenture, Capgemini, EY, KPMG, and PwC are all supposedly partners in this. They're listed as global systems integrators who will work alongside Frontier Co. Accenture even launched a Microsoft-focused forward deployment practice back in March. But the dynamic has fundamentally changed, and that's what matters if you work in this world.

Microsoft isn't handing off implementation anymore. They're not collecting the software revenue while a consulting firm handles the hard part. They're sitting inside your operations. They're accountable. They have skin in the game.

Actually, that's not quite right—they always had financial skin in the game. What they're doing now is different. They have operational skin in the game. They can't blame the consultants. They can't blame your team for not adopting fast enough. They're in the room, redesigning workflows around their own tools, which means the pitch consultants have been making—"Let us manage your Microsoft relationship"—gets a lot harder to land.

What Frontier Co. is actually selling isn't engineering capacity.

It's accountability. It's the thing enterprise clients have been asking third-party consultants for repeatedly, and not always getting. It's the admission that buying software has never automatically changed how a company operates. Not once. Not ever.

Amazon, Anthropic, and OpenAI all launched similar programs in 2026. Amazon. Anthropic. OpenAI. All of them. That tells you something about where the real bottleneck is. It's not the AI models. The models work fine. The bottleneck is implementation. It's the stubborn fact that software licenses don't change organizational behavior on their own. Microsoft is making an expensive, public bet that the only way to close that gap is to show up and actually do the work yourself.

Can they pull it off? That's separate question entirely. Embedding 6,000 people across hundreds of enterprise clients simultaneously is a logistics problem that has nothing to do with how sophisticated the AI is. And running a quasi-competing relationship with five different systems integrators while also trying to partner with them—that requires discipline. Large organizations don't always sustain that kind of discipline.

But the diagnosis underneath is correct. The reason enterprise AI adoption stalled isn't that the tools don't work. It's that most companies don't know how to use them. And a thirty-minute onboarding webinar was never going to fix that.

Microsoft probably should have made this bet two years ago.