300,000 Seats in Six Months — How India's IT Giants Dragged Copilot Out of Pilot Purgatory

For two years, the most honest word in enterprise AI was "pilot." Companies bought a few thousand seats of an AI assistant, ran a proof of concept, generated a slide deck full of cautious optimism, and quietly let most of those licenses gather dust. The industry even had a name for it: pilot purgatory - the place where promising AI deployments go to stall. So when a number lands that suggests the purgatory might be ending, it is worth checking carefully and then sitting with what it means.

Here is the number. On June 3, Microsoft said that India's three largest IT services firms - Infosys, Tata Consultancy Services, and Wipro - have each expanded Microsoft 365 Copilot to more than 100,000 employees, pushing their combined total past 300,000 seats. Six months earlier, in December 2025, Microsoft was highlighting roughly 50,000-seat deployments at the same firms. That is a sixfold jump in half a year, and it is one of the largest enterprise AI rollouts anywhere on the planet.

The number behind the number

Seat counts are easy to inflate and easy to misread. A license is not the same as a habit. The figure that actually tells you whether something is working is not how many people could use the tool, but how many of them open it on a regular basis - and this is where the story gets genuinely interesting.

According to Microsoft, Infosys reports more than 91% monthly active usage among its licensed employees. TCS reports roughly 86%. Wipro reports over 95% monthly active usage and - the detail that brings it down to earth - about 7.5 million Copilot prompts every month. Those are not pilot-purgatory numbers. Most enterprise software would be thrilled with 40% active usage. When you are seeing 86 to 95 percent of a six-figure workforce returning to a tool month after month, you are no longer looking at an experiment. You are looking at infrastructure.

That distinction - experiment versus infrastructure - is the whole point. The skeptical case against enterprise AI was never that the technology didn't work in a demo. It was that real organizations couldn't get it past the demo. High, sustained activation across hundreds of thousands of people is the first large-scale public counterevidence.

Why these three companies are the canary

It matters enormously that this is happening at Infosys, TCS, and Wipro specifically, and not at, say, three random multinationals. India's big-three IT services firms occupy a unique position in the global economy: they are simultaneously among the heaviest internal users of enterprise software and the people who implement that software for thousands of other companies worldwide. When they adopt a tool at workforce scale, they are doing two things at once. They are running the largest possible internal stress test, and they are training the workforce that will go on to deploy the same tool inside banks, insurers, retailers, and manufacturers across the globe.

In other words, these firms are a leading indicator. What they internalize this year tends to become what their clients deploy next year. A sixfold scaling at the implementers is a strong hint about where the broader enterprise market is heading.

The boring reasons it finally worked

The temptation is to credit a model upgrade - some leap in raw capability that suddenly made Copilot irresistible. The more likely explanation is far less glamorous, and far more instructive: the blockers that kept these deployments small were never mostly about intelligence. They were about governance, data residency, and trust.

Microsoft spent the last several months addressing exactly those friction points - broadening compliance certifications across the Copilot portfolio and adding connectors that let the assistant work against live enterprise data instead of stale snapshots. That is unglamorous plumbing. But unglamorous plumbing is usually what stands between a pilot and a rollout. The lesson buried in the 300,000 number is that enterprise AI scaling is gated less by how smart the model is and more by whether a compliance officer will sign off and whether the tool can safely see the data people actually work with.

The caveats worth keeping

A responsible read requires a few asterisks. These figures come from Microsoft and from the firms themselves, all of whom have an obvious interest in a triumphant adoption story; we have verified the headline numbers against Microsoft's own newsroom, but they remain self-reported. "Monthly active" is a generous bar - opening Copilot once in a month counts - and tells us about engagement, not about productivity or return on investment, which are much harder to measure and conspicuously absent from the announcement. And what scales inside a tech-forward IT services firm, where employees are already fluent in software, may not scale as cleanly inside a hospital system or a regional bank.

Still, even with the asterisks, the direction is hard to dismiss. The question hanging over enterprise AI for two years was whether anyone could get it past the pilot. Three of the largest workforces in the industry just answered that they can. The next question - the one worth watching through the back half of 2026 - is whether activation turns into measurable output, and whether the companies these firms serve follow them through the same door.

The seats are filled. Now we find out what the people in them actually get done.