The Week AI Started Doing the Work

Welcome to Briefly AI. Saturday: The week AI started doing the work.

By Friday, something had shifted. Not in a single announcement, not in one headline — but across three days of stories that, taken together, told the same story. This wasn't a week about AI getting smarter. It was the week AI stopped being a tool people used and started being a worker people managed. Or in some cases, a worker that replaced the people who used to do the managing.

Let me take you through it.

Monday, we started with Google. Twenty-five years ago, Google gave the world a box. You typed words into it, it returned links, and that was the deal. Simple, fast, good enough. This week, they retired it. The new interface takes in text, yes — but also images, PDFs, videos, open Chrome tabs. You don't type a query anymore; you start a conversation. You show it things. It responds.

Now, on the surface that sounds like a quality-of-life upgrade. And it is. But the deeper thing is what it signals about how AI is being positioned. Google isn't helping you search anymore. Google is doing the searching for you. The human's job in that transaction — forming a query, scanning results, making judgments — is quietly being absorbed. You bring the context. The machine brings everything else.

That theme — humans providing context, machines doing execution — ran through the whole week.

Also on Monday, we covered Doozy Robotics. They're deploying humanoid robots onto actual production floors. Not labs, not prototypes, not press releases about prototypes. Real factories in the US, the Gulf, and Asia. And the reason humanoid robots specifically — the ones that look and move vaguely like people — is genuinely interesting. Most factory automation requires you to redesign the facility around the machine. Fixed arms, conveyor configurations, custom setups. Humanoids fit into the spaces humans already built. They can use the stairs. They can reach the shelf. You don't have to rebuild the factory; you just swap out the workforce.

The logic there is blunt when you say it out loud. Human-shaped environments plus human-shaped robots equals a very short distance between "this is a pilot programme" and "this is how we operate now."

And then on Tuesday, that distance got even shorter.

ClickUp — a project management software company, not a manufacturer, not some fringe tech outfit — announced that it had laid off a significant portion of its workforce and replaced them with AI agents. Not assisted them. Not complemented them. Replaced them. Customer support teams, data entry, workflow automation — all of it handed to autonomous agents.

What made this unusual wasn't that it happened. There are probably a hundred companies doing something similar right now, quietly, without the press release. What made it unusual is that ClickUp said so, plainly, in public. They named what they were doing. The quiet part was spoken out loud.

That matters because it changes the conversation. When a company frames AI adoption as "we gave our team superpowers," there's room for ambiguity. When a company says "we replaced the team," there isn't. And once one mid-sized SaaS firm does it openly and the sky doesn't fall — no major customer revolt, no regulatory intervention — it becomes a lot easier for the next company to follow. And the one after that.

So by the end of Tuesday, here's what we had. A search interface that does the cognitive work of finding things. Humanoid robots that do the physical work of building things. And a software company that decided the operational work of running a support function didn't require humans at all.

Three different industries. Three different kinds of work. Same direction of travel.

Now, here's the thing I want to sit with for a moment — because the most honest version of this week's story isn't triumphant, and it isn't dystopian. It's uncomfortable in a more ordinary way.

Earlier in the week we covered research on why white-collar professionals resist AI adoption. And the finding wasn't that people are technophobic or change-averse. It was that the tasks they resist handing over — judgment calls, prioritisation, relationship work — are the tasks that define their professional identity. The tasks that, if you took them away, would make the job feel like it wasn't really a job anymore.

The week's stories suggest that the distance between "AI handles execution" and "AI handles everything" is getting shorter faster than most people realise. The coders who've stopped coding without AI assistance — a story circulating at the end of the week — aren't lazier or less skilled. But there's a real question about what happens to the underlying capability when the tool does enough of the work for long enough.

Nobody has a clean answer to that. Anyone who tells you they do is selling something. But it's the question worth holding onto.

Right. Let me bring this together.

This week, AI moved from assistant to agent in three visible, documented, real-world cases. A search engine stopped asking you to search. A robotics firm started filling human-shaped roles with human-shaped machines. A software company replaced teams with autonomous agents and told you about it.

Individually, each of those stories has caveats, timelines to watch, questions to answer. But together, they sketch the outline of something that's been approaching for a while and is now, this week, close enough to touch.

That's the week.

Tomorrow on Sunday we go a bit deeper — into what this actually means for the people doing the work, and whether the psychological reckoning that's coming is one we're remotely prepared for. See you then.