At DevDay on 29 September 2026, OpenAI launched dots: always-on agents that live inside ChatGPT and keep working when you close the laptop. Each dot runs on GPT-6 Astra and has its own cloud computer and browser. It can reach thousands of apps through ChatGPT plugins, and you can talk to it in ChatGPT, Slack or Microsoft Teams.

For Amazon sellers, vendors and agencies this is the first mainstream agent built to watch a business around the clock instead of waiting for a question. This guide explains what dots are, five Amazon jobs they suit, how to give a dot your live Amazon data, and where a human should stay in the loop.

What ChatGPT dots are

• Always on. A dot works toward goals you give it, runs recurring jobs on a schedule and follows up on its own. It does not stop when your session ends.
• Its own computer. Each dot has a cloud computer and browser, so it can open sites and tools the way a person would. You can watch what it is doing and take over.
• Connected to your apps. Dots use the plugins and apps connected to ChatGPT, which is how they read and change data in other tools.
• Rules you set. For each kind of action you decide whether the dot may act alone, must ask first, or should hand the job back to you.
• Plans and regions. At launch dots come with ChatGPT Pro and Business Premium, with the first dot included. Enterprise workspaces can enable them through their admins. The Pro rollout excludes the EEA, the UK and Switzerland, so check availability where you are.

Five Amazon jobs for an always-on agent

An agent that never sleeps is most useful for things that go wrong when nobody is looking. These five jobs fit dots well:

1. Featured Offer watch

The dot checks competing offers and your Featured Offer share every few hours. When a reseller undercuts you, it works out the price that wins the offer back without breaking your margin floor, prepares the change as a dry run and asks you to approve it.

2. Stock-out warning

Every morning the dot compares days of cover with inbound shipments and lead times. If a best seller will run dry before the next shipment lands, it tells you which SKU, how many days early, and how much revenue is at risk.

3. ACoS guardrail

The dot reads search terms and campaign performance next to your real margin, not just sales. When a keyword spends past its break-even ACoS, it proposes a lower bid or a negative keyword and waits for your yes.

4. Suppressed-listing alert

Listings get suppressed for missing attributes, image issues or policy flags, and every hour suppressed is lost sales. The dot spots the issue, drafts the fix against Amazon's product-type rules, and asks before it submits.

5. Weekly P&L brief

On Monday morning the dot writes a short summary for each brand or account: profit after fees, ads and COGS, what moved and why, and what it recommends. It posts the summary in Slack or Teams, where your team already works.

How to give a dot your Amazon data

A dot can open Seller Central in its browser, but that is the slowest and least reliable way to work. Seller Central was built for people, not agents. Screens change, reports take minutes to generate, and history beyond what the interface shows is out of reach. We cover the trade-offs in computer use vs a data layer.

A data layer is the better route. DataDoe connects Seller Central, Vendor Central and Amazon Ads through Amazon's official APIs, joins them by SKU and ASIN with your own costs, and keeps history. The dot gets this through MCP, the same protocol ChatGPT apps use. You add DataDoe to ChatGPT once, as described on our ChatGPT setup page. DataDoe is added through Developer mode as a workspace app, so check that your workspace makes custom apps available to dots before you build a routine on it.

Keep a human on the writes

An always-on agent should read freely and write carefully. DataDoe is built for exactly this:

• Connecting gives read access only. Every Action type, such as listings, prices, orders, A+ Content or Amazon Ads, is off until you switch it on.
• Prices sit behind their own permission, so a dot can fix a listing without being able to touch prices.
• Every change is validated with Amazon as a dry run before anything is submitted.
• ChatGPT asks before any write runs. Combined with the dot's own rules, you decide exactly what it may do alone.
• Every Action is logged with its payload, and you can revoke access at any time.

A sensible starting point: let the dot read everything and act on nothing, then allow ad bid changes within limits, and add listing and price changes once you trust its proposals.

For agencies: one agent, many accounts

Agencies can give each client account its own routine, or run one dot across the whole portfolio. In DataDoe, clients connect through an invitation link, each account can be read-only or read-write, and every Seller Central, Vendor Central and Ads account sits in one organization. A Monday brief for twenty clients becomes one instruction. See DataDoe for agencies.

FAQ

Which ChatGPT plans include dots?

At launch, ChatGPT Pro and Business Premium, with the first dot included. Enterprise admins can enable them. Dots are not part of Free, Go or Plus, and the Pro rollout excludes the EEA, the UK and Switzerland.

Can a dot change my Amazon listings or prices?

Only if you allow it. Through DataDoe, writes go through Amazon's official APIs, each Action type must be switched on, prices need their own permission, and every change runs as a dry run and waits for approval.

Is it safe to let an AI agent click through Seller Central?

We would not make it the default. Browser automation is slow, breaks when screens change, and can look like unusual account activity. Structured access through Amazon's APIs is more reliable and keeps an audit trail.

Does this work for Vendor Central?

Yes. DataDoe gives agents Retail Analytics, Net PPM, Amazon's demand forecast, purchase orders and invoices, and can update vendor listings, A+ Content and Amazon Ads. See DataDoe for vendors.

What does DataDoe cost?

$97 a month after a 14-day free trial, with every Amazon account included. See pricing.