ChatGPT Dots for Business Teams: Put Agents Where Your People Already Talk

Picture your team’s busiest Slack or Teams channel. Questions, updates, “can someone pull that report,” “did anyone follow up with them.” Half of it is work an agent could carry. The other half is people asking where that work is.

That’s the opening ChatGPT Dots create for business teams. As of October 2026, you can message Dots right inside Slack, Microsoft Teams and other workplace platforms. The agent shows up where the work is already being discussed.

Easy adoption is a gift. It’s also how teams end up with five agents nobody owns, each holding credentials nobody remembers granting. Here’s how to roll Dots out to a team the right way.

What Dots bring to a team

From what OpenAI announced at DevDay on September 29, 2026:

• Always-on agents powered by GPT-6 Astra that pursue goals continuously in the background.
• A primary Dot you can name, plus specialist Dots and teams of Dots working together.
• Messaging through Slack, Microsoft Teams and other workplace platforms; texting announced as coming.
• Individual identities, credentials and tools for each Dot — so an agent can be set up through your existing systems like any other team member.
• Security work with Microsoft: OpenAI is working with Microsoft on Agent 365 security controls.
• Availability in ChatGPT on certain paid plans in eligible markets. Confirm yours.

New to Dots? Start with the OpenAI Dots overview, then come back.

Why “its own identity” is the feature to build around

Most team AI problems are access problems. Someone connects an assistant using their own login, and now the assistant can see everything that person can see — forever, even after they change roles.

Giving each Dot its own identity, credentials and tools fixes that, if you use it deliberately:

• The Dot gets its own account in your systems, with only the permissions its job needs.
• When the job changes, you change the Dot’s access — not a person’s.
• When you retire the Dot, you shut off one identity and the access ends.
• Logs show the Dot’s actions separately from your people’s.

Treat each Dot like a new team member with a badge, not a shortcut through someone else’s.

The team rollout, in order

1. Write the policy before the first Dot

Decide what data agents may touch, which tools are approved, where human review is required, and when people should be told they’re dealing with AI. My AI policy outline covers the sections; have an attorney review the final version.

2. Assign an owner to every Dot

One person per Dot. They write its goal, approve its credentials, check its work and answer for it. I build this into an ownership chart where every agent seat has a human owner — the model is on human and agent teams.

3. Pick one channel and one job

Start with a single channel and a single internal goal: a weekly summary of a project channel, a running list of open questions, a draft status report. Nothing customer-facing yet.

4. Set up the Dot’s identity with least access

Create its own credentials. Give it read-only where possible. Add tools only when the job proves it needs them.

5. Pilot with a small group

Two or three people use it for a couple of weeks. They note what it got right, what it got wrong and what it needed to ask. The owner turns those notes into a sharper brief.

6. Add specialists only when the first Dot is reliable

OpenAI describes specialist Dots and teams of Dots. That’s the multi-agent model: a coordinator plus focused specialists. It works when handoffs and accountability are clear — see multi-agent systems.

7. Review on a rhythm

Put agent performance on your weekly Momentum Meeting agenda: what each Dot finished, what it got wrong, what access changed. My AI operating rhythm page shows how to run it.

Good first team jobs for Dots

Start internal and low-risk. Candidates to test, each with a human checkpoint:

• Channel digests. A daily or weekly summary of decisions and open questions in a busy channel.
• Follow-up tracking. A list of “I’ll get back to you” promises made in the channel, flagged to their owners.
• Meeting prep. Pull the open items, project updates and issues into a draft agenda for the owner to edit.
• Research lists. Compile background on a topic for a person to review.
• Status reports. Draft from existing project notes for the project lead to approve.

What stays human: anything sent to customers without review, money, deletions, hiring and HR calls, and any lending or housing decision. Using AI in hiring, lending or housing carries discrimination risk.

Compliance your team needs to know

• Customer outreach is regulated. Since the FCC’s February 2024 ruling, an AI voice on a call counts as an artificial voice under the TCPA, so robocall consent rules reach it. If a Dot drafts texts to customers, those texts still need consent and an A2P 10DLC-registered number, and Do Not Call lists still count. Recording a call? Consent rules differ from state to state. Commercial email falls under CAN-SPAM.
• Disclose AI where people would expect to know.
• Don’t paste client or sensitive data into any tool until your settings and policy approve it.
• Route legal questions to counsel.

Dots vs my picks for a team

For my business, I pick Grok Business with Grok Bot — mainly for its team workspace, seat and role controls, retention settings and permission-aware Google Drive connector. The full reasoning is on the best AI agents hub.

But if your company already lives in ChatGPT and runs its day in Slack or Microsoft Teams, Dots meet your people exactly where they work, and that matters. Run both through the same three real tasks and judge on results, not keynotes. The first-week setup steps are on how to use ChatGPT Dots.

Your next step

Before anyone on your team spins up a Dot, name the owner and write the one-page policy. If you want a team to map your operations and install agents across the company with you, that’s what my AI company, ClientPro, does. If you’d rather start with an AI chief of staff for yourself first, see how Chief works.