This week · Agents

AI agents now get staff IDs. Here is what that means for a service business.

On September 30 Microsoft published a roadmap item that says a lot about where AI is going: Copilot Studio agents can get their own account, with a mailbox, a Teams presence and Office access. Preview started in September and general availability is set for November 2026. Agents are no longer features inside apps. They are becoming members of the team.

In 30 seconds
  • Agents get an identity. Their own account, inbox and Teams presence, governed like an employee.
  • That changes the question from "which AI tool?" to "which AI staff do we hire, and who manages them?"
  • Big companies get this from Microsoft. Service businesses need the same idea, ready-made for their work.
  • That is what an AI operating system is: a team of named agents with jobs, rules and an owner.

What Microsoft announced

Its own account.An agent becomes a first-class identity in Microsoft 365, not a shared login.
Its own mailbox and Teams presence.It can be emailed, mentioned and handed work where people already work.
Governed by admins.Access, policies and use are controlled like any other account.
Timeline.Roadmap item published September 30, 2026. Preview September, general availability November 2026.

Tool vs AI staff, side by side.

AI toolAI staff (agent with an identity)
Who starts the workYou, every timeIt works from a goal and a schedule
Where it worksInside one appEmail, chat, documents, CRM
Who it answers toWhoever is typingA named owner with rules
How you check itYou read the answerA log of every action
What you buySeatsAn outcome, e.g. meetings booked

Why this matters for a service business

Agencies, consultancies, IT providers, accountants and trades run on the same loop: get noticed, start conversations, book the meeting, deliver, get paid. Most of that loop is repetitive work that never gets done consistently because people are busy delivering. An agent with an identity can own a piece of that loop the way a junior hire would.

  1. Give each agent one job.Research, writing, publishing, prospecting, follow-up. One owner per job.
  2. Write its rules down.What it may send, to whom, and what always needs your yes.
  3. Measure the outcome, not the activity.Meetings booked and replies, not messages sent.

How Mercury OS applies this today

Mercury OS is IAW’s AI operating system for service businesses. It runs a team of four named agents: Jarvis commands the team and plans the day, Studio creates the content, Social posts, replies and gathers reviews, and Maximus finds leads and books the appointments. Each agent has its own rules and a run log, and the owner approves what goes out. We run IAW on it ourselves: in its first two weeks our LinkedIn engagement went from about 150 to 7,500, and our first meetings were booked. Those are IAW’s own results, not a client case.

Frequently asked questions

What is an AI agent identity?

An account that belongs to the agent itself, so it can be given access, emailed and audited separately from any person.

Do small businesses need Microsoft Copilot Studio for this?

No. The idea, agents as accountable staff, works on any stack. An AI operating system like Mercury OS packages it for a specific industry so you do not have to build it.

Is it safe to let an agent send email?

With limits. Start with drafts, approve the first messages and templates, and keep a log you actually read.

Mercury OS · Live

Want a team of AI agents that books appointments and wins jobs for your service business? Mercury OS is live. Try it, or talk to us about setting it up.

Try Mercury OS Book a 20-minute call

Sources: Microsoft 365 Roadmap 570430 (via M365 Message Center archive).