OpenAI quietly released ChatGPT Agents to Pro users last week (it will arrive in other paid for accounts in the coming weeks).
This launch signals a significant shift in how generative AI (gen AI) will be used in everyday work.
Agents are tools that don’t just suggest. They do things.
In tests across our team, we’ve seen them:
Complete a full grocery shop, choosing substitutes and navigating stock constraints based on a list and health goals brief
Write and export an editable PowerPoint presentation, based on a live brief
Sort and file a messy desktop on a local computer
Operate like an executive assistant – organising email, calendar and notes to triage tasks and re-prioritise
Agents like ChatGPT’s don’t just act, they also record their steps as they go, showing every decision they made along the way. You can sit and watch as it completes tasks we’d usually do ourselves in its browser, pausing if it needs you to log in to a website or some other blocker. It’s early-stage and still a bit clunky, but the direction of travel is clear: generative AI is moving from text to tasks.
These are things humans can easily do – but they add up. There’s a cognitive cost to switching gears constantly, and decision fatigue creeps in where we don’t always notice it. AI agents can take on that invisible load. They can handle the small stuff, leaving more time for
It’s a reminder that progress in AI doesn’t always arrive in the form of big breakthroughs. Sometimes, it’s something that quietly changes how we work at a more mundane, practical level.
This is about distributed decision-making and a new kind of digital delegation.
So what will this mean for the average knowledge worker?
Much of the conversation around AI is shaped by ideas absorbed through headlines and hearsay rather than experience. It’s understandable: the technology is moving quickly, and the signals are noisy. “AI will never replace creative work” is a common refrain – but who, exactly, is saying it will? Often, we’re arguing with ghosts.
What’s happening is a reimagining of how we work. Tools are emerging that can complete workflows, not just brainstorm them. That can shift the division of labour between people and machines – not dramatically or all at once, but incrementally, and with far-reaching consequences.
You might have seen this perfectly articulated Tweet from last year:
It speaks to a very real fear about our future: who gets to decide how I spend my time?
One thing we noted over the weekend using ChatGPT Agents was that because it showed its ‘workings out’ it highlighted just how many of these decisions we make – and how easily they can be delegated. Not because they’re unimportant, but because they pull focus from the things that are: the higher-level choices and decisions of consequence. The imaginative leaps. The creative pursuits.
It’s easy to overlook the small decisions that quietly shape our working days. Choosing which version of a file to keep. Picking the right substitute for an out-of-stock ingredient. Tweaking the phrasing on a slide title. Each one is minor, but together they create a steady drain on our attention and energy. Neuroscientists call these microdecisions – they cost energy, and lots of small costs soon add up to less brain power for the things we want to do.
This isn’t necessarily about automation. It’s about the opportunity, however small or early-stage, to move our attention back to the work that requires our judgement, experience and taste. And perhaps it all begins with noticing the weight of the small things.
Not everyone will be trialling Agents this week. But everyone will feel their impact in time – through how tasks get triaged, and how we use AI to redesign our work.
The most interesting thing about ChatGPT Agents might not be the tasks they do today. It’s how quickly they make a new mode of working feel normal. Our hope is that they will allow for more creativity, if that’s what we enjoy, not less.