A personal agent is not a team agent. Your org chart has to tell them apart.

The six weeks of agent launches since Meta's Muse produced two different things, and the headlines lumped them together.

Personal agents work for one person. Claude Duet, Muse and a person's own dot use that person's connected apps and ask that person before they act. They have no identity of their own at work. When they do something, they do it as their person.

Team agents work for the company. Google's Gemini agent can create coworker agents, and each one gets its own email address on an @agents domain, its own identity and its own storage. OpenAI's specialist dots, in preview, are provisioned by an employer with their own identity and credentials for one defined job, such as invoice processing.

The difference that matters is who the agent answers to.

Two kinds, two shapes on the chart

	Personal agent	Team agent
Works for	One person	A seat or a team
Logs in as	Its person	Itself
Asks before acting	Its person	The owner of its seat
Measured on	Its person's numbers	Its own number
Ends when	The person disconnects it or leaves	The company retires the seat
Examples	Claude Duet, Muse, a personal dot	Gemini coworker agents, specialist dots

A team agent belongs on the org chart as a seat of its own, with a human it reports to and one number it is measured on. We laid out what that seat has to hold in Agents are coming to the org chart. Here is what the chart has to hold.

A personal agent does not get its own seat. It sits next to its person, like an assistant sits next to an executive. Its work counts toward its person's numbers, and its person answers for it.

What goes wrong when you mix them up

Treat a personal agent like a team agent, and you give it its own seat, its own number and its own authority. Now there is a second accountable party in the chart that really just works for one person. When something goes wrong, the person says the agent did it, and the agent has no one else to answer to.

Treat a team agent like a personal agent, and you file the invoice-processing agent under whoever set it up. When that person goes on vacation or leaves, the agent keeps running with nobody watching. We saw a version of this at Sneeze It. A seat with no owner drifted for months before anyone noticed.

The two will meet

In practice the two kinds will work together constantly. Maria's personal agent will ask the finance team's invoice agent whether a customer paid. A dot in Slack will hand a task to a Gemini coworker agent.

That handoff only works if each side knows what the other is. The team agent should be able to see "this request comes from Maria's personal agent, on Maria's authority" and apply Maria's permissions, not its own.

The test

For every agent in your company, ask: if the person who set this up left tomorrow, should this agent keep working?

If yes, it is a team agent. Give it a seat, an owner and a number. If no, it is a personal agent. Record it next to its person, and disconnect it when they go.

Sources

Give your team's agents the plan

Personal agents read whatever they are connected to. OTP is where your company's chart, priorities, scorecard and commitments live, and any MCP client can connect to it. In Claude or ChatGPT, add a custom connector with this address:

https://orgtp.com/api/mcp

Then ask: "Using OTP, what are my priorities this quarter, what did I commit to in our last meeting, and what is due this week?"

Start free at orgtp.com. Every seat is free, for people and agents. You pay only for the AI you use.

The series

All ten on one page: Personal agents are coming to work

Series: Personal agents are coming to work. Part 8 of 10. Read this post as Markdown: /blog/personal-agent-vs-team-agent.md.