Office management edition
OpenAI is letting ChatGPT agents rearrange appointments. Google is giving Gemini reusable skills. And a cheaper model makes automation more affordable.
ChatGPT agents can rearrange appointments and book flights. These are tasks you often handle yourself today. So first ask what the agent is allowed to do. And ask how you can check what it does. ChatGPT Space brings colleagues and agents together in a shared workspace. Agree who keeps track of changes and who can step in. The source gives no date for the Netherlands and names no subscription. So do not assume these features are already available.
OpenAI introduced 'dots'. These are agents: AI assistants that carry out tasks for you. They work in a shared workspace with your colleagues and their agents.
On 29 September, OpenAI presented more than 20 products at DevDay. The most important is called 'dots'. These are agents in ChatGPT that are always ready to help. An agent is AI that does more than answer questions. It also takes action. OpenAI says dots can rearrange appointments, book flights or edit code for you. The first two tasks sound a lot like office work. OpenAI also introduced ChatGPT Space. It is a shared workspace. Teams, ChatGPT and dots work together on projects there. Companies can use a Decisions API to let AI handle small, repeated choices. An API connects to your own software. Plugins, which add features to ChatGPT, work like complete apps inside ChatGPT. The source does not say when this will come to the Netherlands. It also does not say which subscription includes it. So do not expect to be able to use it straight away. A regular chatbot gives advice, but you take action. An agent takes the action itself. This shifts your work from doing to checking. Rearranging an appointment affects the other people involved. Booking a flight costs money. In ChatGPT Space, several people and agents work on the same project. You need to be able to see who changed what. So you need clear rules. First decide what the agent can and cannot do. Then make sure you can see afterwards what it has done. Also agree who can step in. The source says this news comes amid debate about oversight of AI that works on its own. That is the trade-off. You save time when you do not have to take every step yourself. But mistakes with calendars and travel can be costly and visible. An agent also does not know all the sensitivities you know. It may not know that a director is never free on Friday afternoons. So start with tasks that are easy to undo. Can you see dots in your workplace? Ask your administrator which calendars and bookings the agent can change.
Background for subscribersThese stories cover AI features that carry out tasks or automate set instructions. This can save time, but checks are still needed. Ask what an agent is allowed to change. Read confirmations before approving them. Check answers against a second source. Also agree who uses AI and who has had training. And ask how suppliers have their systems checked. Start with a small task. Choose one where you can easily fix a mistake.
OpenAI keeps its best model back and opts for a cheaper one
OpenAI is not releasing its latest top model, GPT-6.1 Astra. The reason is safety. Instead, it is launching GPT-6.1 Sol. OpenAI says Sol is almost as good as Astra at coding, using computers and professional work. Sol costs one fifth as much as Astra. You pay per token. A token is a piece of text that the model processes. Sol costs about 2 dollars per million tokens you put in. You pay 10 dollars for the tokens it produces. You can use Sol through the API, ChatGPT Work and Codex. The API connects to your own software. The other two are OpenAI products. This means companies using AI will pay much less for each task. You pay for what the model receives and what it returns. An agent often takes many steps to complete a task. Each step uses tokens, so the costs can rise quickly. A model that costs one fifth as much makes the same automation much cheaper. This fits with Ramp's AI Index. It shows that spending on AI is falling. OpenAI and Anthropic have cut their prices. Customers have also switched to cheaper models. Pay close attention to the words 'almost as good'. That is what OpenAI says. The company is deliberately keeping its best model, Astra, back. So you do not get the strongest model. That is probably not a problem for much office work. A summary or draft email usually does not need a top model. Ask your administrator if you can use a cheaper AI model for routine tasks.
Background for subscribersThis is a taster.
Subscribers read the whole Office management edition: every story, each with its background in plain language.
Subscribe →Not ready for that? Read the free letter first →
AI and tech, made simple.
This is the Office management edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026