There is a very obvious appeal of an assistant that keeps working while you get on with something more meaningful. That has been the promise of AI since the early days. Instead of you constantly checking whether something happened, reminding someone what changed and trying to keep a collection of small tasks from becoming tomorrow’s problem, you get to focus on the real tasks of growing your business. And it seems like that is exactly what OpenAI is targeting with Dots, which it announced on 29 September. Powered by GPT-6 Astra, these assistants have their own cloud computers and are designed to keep working towards goals between conversations, using connected apps rather than waiting for another prompt each time.
It’s an interesting direction, although I think the question worth asking is how much work comes off your plate once you account for the time spent checking what the AI has done.
What Dots is supposed to do
OpenAI’s pitch is that you give a dot an ongoing responsibility, supply the relevant context and let it figure out the steps, with results and decisions coming back to you for review. Its product page describes work such as preparing content from an interview transcript and keeping project materials up to date as information changes. You get to check progress, provide feedback or pause it altogether.
For a small business, you can imagine the appeal of having an assistant prepare a weekly overview, pull together the information needed for a client meeting in a few seconds or spot that two documents contain conflicting promises and need to be fixed.
The value however, would be in how well it understands the context around the request. Finding an unfinished proposal is useful, but understanding that the client changed the brief yesterday and bringing back a corrected draft would be considerably more useful.
That is why these products are so hard to judge from a launch demo. Under the lights of the well choreographed and PR infused launches, these demos are a neat example of what the system can do under almost perfect conditions, but an ordinary working week will reveal how often it gets confused, needs rescuing or confidently carries on with “BS” AI slop.
I’d be careful about what I give it access to
I can see myself using something like this for research and preparing drafts, but I’d be a lot more careful once client information becomes part of the job. A proposal can contain pricing, margins, plans that haven’t been announced and details a client shared with me in confidence, so before connecting an inbox or folder, I’d want to understand what the assistant can read, what it remembers and how that information might be used later. Having access to a document myself doesn’t automatically mean a client would be comfortable with me giving an AI assistant access to it too.
OpenAI says Dots’ proactive background research is read-only, which is useful, but that addresses only part of the concern. Preventing it from sending an email or changing a document doesn’t answer the separate question of whether it should be reading the information in the first place. I’d also want to check the data settings and terms for the account I’m using, rather than assume that a personal subscription and a business workspace offer the same protections.
That doesn’t mean I’d rule it out for client work, because there are useful things it could do with an approved brief or a document stripped of sensitive details. I’d start there and keep access limited to what the job actually needs, with anything going back to a client still coming through me. If I’m putting my name on a proposal, I need to know that the figures, promises and information in it belong there, regardless of how much of the preparation the AI handled.
What this means for South African users
Based on OpenAI’s published rollout rules, South Africans on an eligible plan should be able to access Dots, although it may take a few days to appear on your account. For now, that means Pro or Business Premium, with Enterprise users getting access to a beta if their administrator enables it. If you’re on Free or Plus, you’re not included in this first rollout.
Your first dot comes with the eligible subscription, but the distinction between chatting to it and asking it to do work is worth understanding. Conversations don’t count towards your ChatGPT limits, while tasks it runs through Codex or ChatGPT Work still use the relevant allowances, so having an assistant available around the clock doesn’t mean it can do an unlimited amount of work.
OpenAI lists its Pro plans at $100, $200 and $500 a month, depending on the tier, and those are the subscription prices rather than an additional charge for Dots. I haven’t confirmed the rand price at checkout, but even the entry point is enough money that I’d want a clear idea of what I’m going to use it for before upgrading. If you’re already paying for an eligible plan, there’s a different calculation to make, because you can try it on work you already need done and see whether it earns a place in your day.
You’ll also need to set it up on desktop, and the texting beta is currently limited to eligible Pro users in the US. Being eligible in South Africa doesn’t mean we get every feature shown at launch.
What would make it worth paying for?
When you run a small business, you can hand over a task and still find yourself carrying most of it in your head, because you’re remembering the deadline, checking the details and making sure the latest changes haven’t been missed. That’s where I’d want something like Dots to help, particularly if it can take the corrections I’ve already made and apply them without needing another explanation.
I wouldn’t expect it to get everything right, and I’d still want to review anything important, but there’s a big difference between checking a piece of work that’s nearly ready and spending an afternoon fixing something that looked finished when it arrived. I’d want to use it through a few messy projects, where the brief changes and some of the information is missing, before deciding how much responsibility I’m comfortable giving it.
I’m interested enough to try it, because there’s plenty of work I’d happily hand over if I could trust it to be done properly. Whether I keep paying would depend on how much of that work I can stop chasing, and whether I actually have more time for my clients and businesses once it’s part of the routine.
