7 min read
An AI that keeps working after you close the laptop is the most useful thing OpenAI has shipped for solo owners in a year, and also the easiest one to misuse. That is the honest verdict on Dots, the always-on assistant OpenAI unveiled at its DevDay conference on September 29. The pitch is irresistible: hand it a goal, walk away, and come back to finished work. The reality is more interesting, and more conditional, than the keynote made it sound.
This is not a review of every feature. It is a stance: Dots is worth paying attention to, but most one-person businesses should not hand it the keys yet. Here is why, and what to do instead while the product matures.
What OpenAI actually launched
According to CNBC’s DevDay coverage, Dots was the headline among more than 20 announcements. A dot is a persistent agent (an AI that carries out multi-step tasks on its own, not just answers questions) that lives inside ChatGPT. DataCamp’s breakdown of the launch lays out the specifics:
- It runs on GPT-6 Astra, OpenAI’s flagship model, with its own cloud computer and web browser.
- It keeps working between conversations. You set a goal, and it makes progress in the background, reporting back in ChatGPT, Slack or Teams.
- It can connect to thousands of apps, the same way ChatGPT’s connectors already reach email, calendars and file storage.
- It asks before acting by default. Custom rules let you set each type of action to “do it”, “do it if I pre-approved it”, “ask me first”, or “hand it back to me”.
- It is not cheap. The first dot is included only in the Pro plan (the cheapest tier with Dots runs $100 a month) and in Business Premium seats. Free, Go and Plus users do not get one.
One example OpenAI highlighted is telling: a user’s dot spotted a forgotten invoice, prepared it, and waited for approval before sending. That is exactly the kind of chore that leaks money from a one-person business. It is also exactly the kind of task where a wrong move costs you a client.
The case for getting excited
Let us give Dots its due. Most AI tools a solo owner uses today are reactive. You open ChatGPT, you ask, it answers, you do the work. The bottleneck is still you, and you are already the bottleneck for everything else.
An always-on agent flips that. The idea of an assistant that notices things (a client who has gone quiet, a review that needs a reply, a supplier price that changed) is closer to having a part-time operations hire than to having a smarter search box. For a business of one, attention is the scarcest resource, and Dots is the first mainstream product built to spend attention on your behalf.
The approval system is also genuinely well designed. Being able to say “draft anything, send nothing” or “you may update my spreadsheet but never touch payments” is the right mental model. It mirrors how you would onboard a human assistant: small permissions first, more trust as it is earned.
Why most solo owners should wait anyway
Here is where the skepticism comes in. None of these are deal breakers forever. Together, they make Dots a poor first purchase for most one-person businesses in October 2026.
1. The price is built for power users, not owners testing the water
A $100 monthly entry point is five times what most solo owners pay for ChatGPT Plus. That only makes sense if a dot reliably saves you more than a few hours a month, every month. Nobody, including OpenAI, has data on that yet. Worse, the terms for background work after the first month have not been published. We have already written about why flat AI subscriptions are giving way to metered pricing, and Dots looks like the clearest example yet: a capacity you rent, with the meter details still to come.
2. Availability is patchy
At launch, Pro subscribers in the European Economic Area, Switzerland and the UK cannot set up a dot. Setup also requires the desktop app or desktop browser. If you run your business from a phone between jobs, as many tradespeople and service providers do, you cannot even start.
3. Background work is where AI mistakes hide
When you watch an AI work, you catch its errors. When it works while you sleep, its errors arrive as finished facts: an email already drafted in the wrong tone, a spreadsheet already “cleaned” of rows that mattered. The approval rules help, but they only cover actions. They do not cover judgment. A dot that summarizes your week wrongly does not need permission to be wrong.
4. Its memory is all or nothing
At launch, deleting what a dot has remembered means deleting the dot entirely. For anyone who puts client details into AI tools, that is a real concern. If you have not already, read our piece on the mistakes solo owners make when putting business data into AI before connecting an agent to your inbox.
What to do instead, this month
You do not need Dots to start building the habits that will make an always-on agent valuable later. In fact, the owners who will get the most from Dots in 2027 are the ones who do this groundwork now.
- Write down your recurring chores. Spend one week noting every task you do more than twice. These are the jobs an agent will eventually take. Without this list, you will hand a $100 agent vague goals and get vague results.
- Automate the boring, rule-based ones with cheaper tools first. Invoice reminders, lead routing and appointment confirmations do not need a flagship model. A simple automation in Zapier or Make, or a narrower agent like the one in our Lindy review, handles them for a fraction of the price.
- Practice writing approval rules on paper. For each chore, decide: may an AI do this alone, draft it for me, or never touch it? That exercise is free, and it is the single most important skill for working with agents.
- Keep your setup portable. Store your prompts, instructions and process notes in your own documents, not only inside one vendor’s product. We made this argument when OpenAI announced it was retiring Custom GPTs, and it applies doubly to agents that hold weeks of accumulated context.
Who should try Dots right now
There is a narrow group for whom the math already works. If you are already paying for ChatGPT Pro, live outside the EEA, Switzerland and the UK, and work at a desk most of the day, you get a dot at no extra cost. Use it. Start with read-only goals (“tell me every Monday which invoices are overdue and which leads have not had a reply in five days”) and resist giving it send permissions for at least a month.
Also consider it if your business is heavy on research and monitoring: a consultant tracking a few industries, a freelance writer following beats, a reseller watching prices. Those are low-risk tasks where an always-on helper shines and a mistake is cheap.
Everyone else should watch the first few months of real-world reports, wait for the post-launch pricing, and spend the time getting their processes written down. The agent is coming either way. Being ready for it matters more than being first.
The bigger shift behind the launch
Set the product aside and look at the direction. Meta’s Muse for Small Business, OpenAI’s Dots, and the wave of inbox agents all point the same way: AI is moving from something you visit to something that runs alongside your business. That changes the core skill a solo owner needs from “writing good prompts” to “managing a worker you cannot see.”
That is a management skill, not a technical one, and it is one every solo owner can start building today, with or without a $100 subscription.
Your turn: if you could hand one recurring chore to an always-on assistant tomorrow, which would it be, and what would you never let it touch? Tell us in the comments.
