Microsoft announces 365 Copilot
Built on GPT-4 and Microsoft's own data graph, it launched to a small group of testers with no price or general release date set.
- Money & business
- Culture & impact
- Notable
Microsoft announced 365 Copilot, an assistant built into Word, Excel, PowerPoint, Outlook and Teams that combined large language models — including GPT-4 — with a user’s own documents, emails, calendar and contacts, accessed through Microsoft Graph. A companion feature, “Business Chat,” let users query across all of a person’s Microsoft 365 data and apps in natural language rather than working within one program at a time.
The pitch was concrete rather than conceptual: draft a first version of a document in Word from a prompt and a linked file, turn a Word document into a PowerPoint deck, summarise an email thread, or ask Business Chat to produce a status update by pulling from recent meetings, emails and chats on a topic. Chief executive Satya Nadella framed it as a shift in interface — “the most universal interface — natural language” — layered on top of applications used daily by hundreds of millions of people.
Microsoft gave no pricing or general-availability date at announcement, saying only that it was testing the product with a small group of corporate customers and would expand access as it improved the underlying models. That is a common pattern for major platform announcements — capability demonstrated well ahead of a shipped, priced product — and it meant 365 Copilot functioned initially as a statement of intent for where Microsoft’s roughly $10 billion investment in OpenAI would show up in its own products, rather than something most users could try.
The announcement came two days after OpenAI released GPT-4 and set the template Microsoft, Google and others followed through 2023: generative features layered onto existing productivity suites, aimed less at novelty use than at the everyday tasks of drafting, summarising and searching that make up the bulk of office work.
In the commentary
What people were saying around this time — external links, from the record's commentary rail.
- 17 March 2023 · Ethan Mollick · One Useful ThingUsing AI to make teaching easier & more impactful
- 15 March 2023 · Latent.Space · Latent SpaceThe Multi-modal, Multi-model, Multi-everything Future of AGI
- 20 March 2023 · Arvind Narayanan, Sayash Kapoor · AI as Normal TechnologyGPT-4 and professional benchmarks: the wrong answer to the wrong question