In April 2025, Mustafa Suleyman stood in front of an enormous video screen that read "Your AI Companion" and told the world what Copilot was going to be. Seventeen months later, the executive who now oversees Copilot's core workplace features put it differently to Bloomberg. "We're not going to build a Copilot that's like your personal companion," Charles Lamanna said. "That's just not what people want from Microsoft," he added. "We help you get stuff done."
The occasion was an event in Seattle where a few dozen business and technology leaders got a preview of what Microsoft calls the new Copilot. Bloomberg's Matt Day reported on 25 September that the product, the result of a six-month engineering effort, has absorbed the separate version built for home use, kept some of its design, and dropped the marketing that cast Microsoft as a builder of personal AI. The company is ceding the personal chatbot market to OpenAI, Google and now Meta.
A bet that did not pay
Two years ago Microsoft had set up separate teams for a home assistant and a work assistant, Bloomberg recalls, and put Suleyman, a co-founder of DeepMind, in charge of the consumer push. Having helped OpenAI bring ChatGPT to the world, the company bet its own assistant could win a new set of customers.
It did not happen. ChatGPT sprinted to a billion users. Google pushed Gemini across the Android ecosystem. "Comparatively few people took Microsoft up on its offer of a personal chatbot," Bloomberg writes. In March, Suleyman handed his Copilot product responsibilities to Jacob Andreou, a former Snap executive, and turned to building new AI models.
The retreat has a face. Mico, the animated yellow blob Microsoft launched in Copilot's voice mode last October as a way to give the chatbot an "identity," is being removed from voice conversations. Microsoft's own support page says the character "helped us learn about warmth, expressiveness, and how people want to talk with AI," and that it now moves to Learn Live, the tutoring feature, "where the character has the most room to grow." The change started in August and rolls out in waves. Voice itself stays. The face goes. The Verge lists Mico alongside Clippy, Cortana and Rover in the company's line of retired helpers.
Where the money is
The numbers explain the turn. Microsoft has more than 450 million paid commercial seats for Microsoft 365. On 29 July, Satya Nadella announced that Microsoft 365 Copilot had "reached over 30 million paid seats," with net additions more than doubling quarter over quarter. That is a big business, and still well under a tenth of the installed base: most of the growth comes from very large customers, NHS England rolling it out to 505,000 clinicians and staff, HSBC committing to 200,000 seats, EY deploying it to 400,000 employees.
So the new Copilot is built for those buyers.
TechTarget describes three destinations. Home puts Teams chat and Copilot Cowork together with the full Word, Excel and PowerPoint apps. Code, powered by GitHub Copilot, lets users build custom apps and dashboards from natural language. Autopilot, a digital co-worker formerly known as Scout, sets up review processes, builds schedules and preps meetings. The whole thing, the site notes, turns Copilot from an assistant into a content workspace, and fewer than 20 million users paid for Copilot out of more than 450 million Microsoft 365 customers when the redesign was being planned.
PYMNTS summed up the logic: winning consumers meant building a new daily habit from scratch against products that already had one. Winning at work means adding AI to software that hundreds of millions of people already open every morning.
Why it matters
A company with Microsoft's reach tried, for two years and with real money, to become someone's AI companion, and concluded that people did not want that from it. That is worth sitting with. It suggests companionship is not a feature you bolt onto a productivity brand. People pick a companion the way they pick a friend, on trust and on tone, not on the logo of the software they use at the office.
It also narrows the field. Bloomberg's sentence is precise: the market for personal chatbots is being ceded to OpenAI, Google and Meta, three companies whose business is attention and whose assistants are designed to be talked to every day. Meta's Muse, an assistant that acts on users' behalf, has raced to the top of the download charts. Andreou and Lamanna told Bloomberg that Muse's success proved people were receptive to AI automation; Microsoft's opportunity, they said, is to bring similar capabilities to office workers.
The question this leaves open is the one Reflets keeps returning to. If the companion market belongs to the three largest attention businesses on earth, what kind of companion will they build? A system optimised for daily return is the companion that won't let you leave, and the research on what a year of that does to people is not reassuring: a year on Character.AI, fewer in-person ties. Microsoft's exit removes one large player that was, whatever its motives, at least trying to sell a companion to adults who pay for software rather than to users who are the product.
One caveat. Mico's retirement page and the Bloomberg piece describe a change of strategy, not a change of heart about anthropomorphism. Microsoft says the lessons from Mico "are shaping Copilot going forward." The blob is gone from voice; the warmth it was meant to carry is being folded into a work assistant. Whether an office tool should sound warm is a separate question, and the company has not answered it.