The tool you checked out on in June already left you behind.

Somewhere back in the first half of this year, a lot of business owners looked at the AI tool their business was using, decided it was fine for now, and quietly stopped paying attention to what else was happening in the category. That decision felt reasonable at the time. There was a business to run, and checking in on software updates every few weeks isn't how anyone wants to spend their attention. The tool worked. There were more urgent things to look at.

Here's what that decision actually cost, stated as plainly as I can make it. It isn't a vague sense that things are moving fast in AI, the kind of line that gets said so often it stops meaning anything. It's a specific, checkable, publicly reported sequence of events, and once you see the sequence, the cost of looking away becomes a lot harder to wave off.

What actually happened, and in what order?

Claude Cowork, the tool doing the work behind this very newsletter, launched earlier this year as a preview for one tier of subscribers. It then rolled out more broadly, moved onto desktop apps for Mac and Windows, and expanded into a cloud, web and mobile beta. That's not marketing language. That's the literal sequence of a product going from a narrow preview to something available across most of the ways someone might want to use it, over the space of a handful of months.

On the 9th of July this year, OpenAI answered directly. ChatGPT Work launched, built on a new GPT-5.6 model family, and multiple industry outlets covered it explicitly as catch-up rather than an independent move. One headline put it about as bluntly as a headline can: "OpenAI Plays Catch-Up: ChatGPT Work Borrows Anthropic's Cowork Blueprint." ChatGPT Work bundles in Codex, OpenAI's coding agent, adds scheduled and triggered tasks, and connects to Slack, Teams, Google Drive and SharePoint. It sits in the same category Cowork established first: hand the tool a real, multi-step piece of work and walk away from it, rather than sitting there prompting it one line at a time.

I want to be careful about what this newsletter is and isn't arguing, because it would be easy to read this as a pitch for one platform over the other, and that's not the point at all. Which specific product ends up ahead in this particular race barely matters to a business owner who isn't in the software industry. What matters, and what the sequence above actually proves, is that the underlying category moved in a way you could set your watch by: a leading platform established a new way of using AI, and a serious competitor closed the gap within two to three months, covered by the press the entire way. That's not a slow-moving field where checking in once or twice a year keeps you current. That's a field where the distance between "ahead" and "caught up" is measured in weeks.

What shifted underneath the product names?

Here's the part of this that actually matters to your business, and it has nothing to do with which logo is on the app. A year ago, using AI well meant learning to write a good prompt. One request, one response, and the value you got out of it depended almost entirely on how precisely you asked. That was the whole skill.

That is not what these tools do anymore, not for the businesses using them properly. The shift is from a tool you operate one exchange at a time to a tool you can hand a genuine piece of work, walk away from, and come back to find handled: a report pulled together from several sources, a first pass at a client response drafted and waiting for review, a multi-step process carried from start to finish without you sitting over it the whole way. That's the actual capability jump the Cowork and Work sequence proves happened, and it happened for real, documented, checkable reasons rather than because a vendor said so in a launch email.

A capability jump like that changes what the whole conversation about AI adoption should be about. When the tool could only do what you told it, one instruction at a time, the skill that mattered was knowing how to ask. Now that the tool can be handed something and left to run, the skill that matters is knowing what you're comfortable handing over in the first place, and what stays yours to do regardless of how capable the tool becomes. That's not a smaller question than the prompting question was. It's a considerably bigger one, and most businesses haven't started asking it yet, because they're still operating from an understanding of these tools that was accurate in January and isn't accurate now.

It's worth making this concrete, because the abstract version of the argument is easy to nod along with and easy to forget by Monday. In January, a business owner using AI well might have opened a chat window, typed a request for a first draft of a client email, read it, edited it, and sent it. One exchange, one output, the owner's hands on every part of the process from request to send. That was the actual ceiling of what the tools could do well, and getting good at that one exchange was a real skill worth having.

The same business, using the same category of tool today, can hand over something considerably larger: pull the last three months of a client's correspondence, draft a status update that reflects where the project actually stands, flag anything that looks like it needs the owner's attention before it goes out, and leave the final version sitting ready for a two-minute review rather than a from-scratch write. That's not a faster version of the January task. It's a different task altogether, and it only works if someone has already decided, on purpose, which parts of that sequence the business is comfortable letting run unsupervised and which parts still need a human looking at them before anything goes out the door.

Why is the governance question right behind this?

I'm not going to make the governance argument in full here, because it deserves its own newsletter and it's getting one next week. But it's worth naming plainly that the capability jump above is exactly why that conversation is about to matter more than it has at any point so far this year. A tool that only responds to a single prompt at a time has a natural ceiling on how much damage a bad instruction can do. A tool that can be handed a real task and left to run for an hour, touching client records or draft communications the whole time, does not have that same ceiling, and the businesses that haven't updated their sense of what these tools can do are, almost by definition, the businesses that haven't updated their sense of what needs a guardrail around it either. The capability moved. The conversation about what's allowed to run unsupervised needs to move with it, and for most businesses, it hasn't yet.

Why is "it's fine for now" the wrong test?

The instinct to stick with what's working is not a bad instinct in general. Most of the time, leaving a working system alone is exactly right, and chasing every new release is its own kind of waste. The problem with applying that instinct here is that "fine for now" was never actually the right question to be asking about a category moving this fast. The right question is whether the gap between what your business is using and what's now actually available has widened enough that it's costing you something you can't see from the inside.

That gap is invisible in the same way most gaps like this are invisible. No single week feels different. The business that checked out in June and the business that kept adjusting through the middle of the year are doing roughly the same volume of the same kind of work in any given seven-day stretch, right up until the moment something makes the difference visible: a competitor turning around client work faster, a team member who's used the newer tools somewhere else quietly wondering why the business hasn't, a task that used to eat half a day now taking twenty minutes for everyone except you. By the time the gap shows up as a number, it's already been widening for months.

This is the actual argument for paying attention to a sequence like Cowork and Work, and it isn't an argument about brand loyalty or which company deserves your subscription fee. It's an argument that the category itself crossed a real threshold this year, provably, in public, and that a business still operating on a January-era understanding of what these tools do is not simply a little behind. It's answering a different, smaller question than the one that's actually in front of it now.

What should you actually do with this, this week?

None of this requires an overhaul. It requires one honest look, taken deliberately rather than left to happen by accident, at what your business is currently using AI for and whether that use still matches what the tools you're paying for can actually do. Most businesses I talk to are using a tool that's capable of carrying a multi-step task end to end for exactly the same single-prompt jobs it was doing back in January, not because the tool can't do more, but because nobody circled back to check.

That's a smaller ask than it sounds. It doesn't mean adopting every new feature or chasing whichever platform wins the next round of press coverage. It means picking one thing your business currently does the old way, one prompt at a time, and asking plainly whether it's now the kind of thing you could hand over and walk away from instead. If the answer is yes, that's the whole exercise for this week. If the answer is truly no, that's useful information too, because it tells you exactly where the judgement question from the section above actually applies to your business specifically, rather than in the abstract.

There's a second, smaller version of this exercise worth doing alongside the first, and it takes about the same ten minutes. Look at whoever on your team has picked up one of these tools fastest, the person who's clearly ahead of everyone else in how they're using it day to day, and ask them directly what they've noticed changing about what it can do. Most owners never ask this question, not out of disinterest but because it doesn't occur to them that the answer might have moved since the last time they checked. The person closest to the daily use of the tool is usually the first to notice when it starts doing something it couldn't do a few months ago, well before that shift shows up in a headline or a press comparison. That conversation costs nothing and tells you, in your own business's language rather than an industry outlet's, exactly how wide the gap this newsletter has been describing actually is.

None of this is a call to panic, and it was never meant to read like one. A business that checked out in June is not ruined. It's simply behind by an amount that's easy to close once it's actually named, and considerably harder to close the longer it stays unnamed. The tool you tried in January and the tool sitting in front of you right now are not the same tool wearing the same name. Treating them as though they are is the one mistake this whole newsletter has been trying to talk you out of.

How do you find out where your business stands?

Knowing whether your business is still operating on a January-era understanding of what these tools can do, or has actually kept pace with what changed, is exactly the kind of gap the AI Readiness Assessment is built to surface. It takes about ten minutes and the result is specific to your business. Click here now to take it.

My book, The Human Edge is launching early Q4 2026. To be first to know when it's available, connect with me on LinkedIn and comment or DM me there.