Is an AI agent that never switches off a good idea?

Yesterday, OpenAI introduced dots, always-on agents in ChatGPT designed to take on ongoing work and keep making progress between conversations. You give a dot a goal, connect the tools it needs, and define what it can do independently.

The technology is impressive. But for us at ON-IQ Vision, the interesting question isn’t whether an agent can work 24/7.

It’s what we allow it to do when we’re not watching.

Our own agents sort git issues and summarise our projects at night. As we build and use agentic workflows ourselves, we keep coming back to three things:

→ Give the agent a clear job, not an open-ended goal.
→ Give it only the data and access that job requires.
→ Keep humans involved where decisions actually matter.

So yes, it can be good, when the task is clear and a person decides what is allowed to happen.

OpenAI’s dots make this conversation even more relevant: the future isn’t simply about agents doing more. It’s about designing the right boundaries around what they do.

That’s AI in Practice.

What would you leave to an AI agent overnight, and what would you never delegate?

#AIinPractice #AgenticAI #AIAgents #Startups #CoBuilding #TechPartner #AIForGood #ONIQVision