Most AI tools wait for you. You type, they respond. You forget to ask, nothing happens. Dots flips that entirely — it’s a ChatGPT agent that runs in the background, monitors what you’ve given it access to, and reaches out to you when it finds something worth your attention.
This isn’t a chatbot upgrade. It’s a different category of tool.
What Dots Actually Is
A Dot is a persistent AI agent that lives inside your ChatGPT account. Unlike a standard chat session, it doesn’t go dormant when you close the tab. It has its own compute environment, meaning it keeps working whether your laptop is open or not.
You give it a standing responsibility — not a one-off prompt — and it figures out the steps. Connect it to your email, calendar, Google Drive, or any of ChatGPT’s integrations, and it starts drawing on real context from your work life.
The key behavioral shift: it messages you. You don’t have to remember to check in.
What It Can Take Off Your Plate
The use cases that make Dots genuinely useful are the ones involving ongoing monitoring and follow-through — the tasks that fall through the cracks because nobody remembered to revisit them.
A few concrete examples:
- Contract review on autopilot. Give it access to your email and it can flag clause inconsistencies or missing terms in attached agreements before a deadline sneaks up.
- Content pipeline management. Tell it to always keep your next three newsletter drafts ready based on your notes folder. It pulls from your Drive, drafts, and notifies you when something needs review.
- Launch coordination. Hand it a folder of product docs and set rules — never send anything, flag pricing changes, escalate anything that touches the ship date. It cross-references the docs, catches mismatches, and surfaces them without being asked.
- Calendar and travel. With the right integrations, it can spot scheduling conflicts or find flight options for an upcoming trip you mentioned in an email thread.
None of these require you to prompt it repeatedly. You set the responsibility once.
How to Set One Up
Dots appears as its own section in the ChatGPT app. Setup is straightforward, but what determines how useful it is comes down to access.
Step 1: Name it and connect your tools. The more integrations you enable — email, calendar, Drive, Slack — the more context it has to act on. You control exactly what it can see. If you’re not comfortable connecting email, start with a single Google Drive folder.
Step 2: Give it a responsibility, not just a task. The framing matters. Instead of “summarize my emails,” try something like: “Keep me informed of anything in my inbox that could affect the Henderson project timeline, and draft a response for my approval when something needs a reply.” That gives it a standing goal with room to use judgment.
Step 3: Set rules before it starts. You can specify constraints upfront — draft but never send, never alter pricing, always ask before rescheduling anything. It’ll confirm those rules and check in before taking borderline actions.
Step 4: Let it run, then tune the signal. Early on, it may surface more updates than you want. That’s normal. You can tell it to batch non-urgent items into a single daily recap and only interrupt you for things that are time-sensitive. Getting that calibration right is worth the first few days of back-and-forth.
What It Can’t Do Yet
Dots won’t act autonomously — it drafts, flags, and proposes, but won’t send an email or post content without your approval. That’s a feature, not a limitation. The approval layer is what makes it safe to give it broad access.
Right now, each account has one primary Dot. The vision is multiple specialized agents working in parallel — one managing client communications, another tracking project deadlines — but that’s not the current state.
Availability is also limited. It’s launching on the $200/month Pro plan first, with broader rollout to follow.
The Shift Worth Understanding
Every previous AI workflow required you to remember to use it. You had to open the tool, write the prompt, wait for output. Dots removes that dependency. It runs on a responsibility you set once, learns from your feedback, and gets more useful as it accumulates context about how you work.
If you have access, start small: one integration, one standing goal, and a clear rule set. Get comfortable with how it surfaces information before expanding its reach. The real payoff isn’t any single thing it catches — it’s the compounding effect of having something that never forgets to follow through.