Julie Yack

The “faster horse” gets a bad rap.

We use it as shorthand for thinking too small. For taking a new technology and using it to make the old thing a little faster instead of imagining something entirely new.

But you know what?

A faster horse can absolutely be useful. And I think that’s worth remembering as we talk about AI adoption.

If AI can help you write the email faster, summarize the meeting, analyze the spreadsheet, create the first draft, or take 20 minutes out of a tedious task, that’s a win.

Take the win.

Not every interaction with AI needs to transform the way we work. Productivity gains are real gains. Making something faster, easier, or better is valuable even if the fundamental work hasn’t changed at all.

Two things can be true at the same time. We can get enormous value from using AI to improve the work we already do. And we can ask whether AI gives us an opportunity to rethink the work itself.

I’ve been thinking about that distinction as:

Use → Adopt → Adapt

Use is pretty straightforward.

Can AI help me do this?

That’s where a lot of us start. We try the tool. We learn what it’s good at. We figure out where it saves us time and where it absolutely does not. We make some faster horses.

Then we adopt.

AI starts becoming part of how we work. We don’t just occasionally ask it to do something. We change our behavior because it’s available. We learn where to bring it into a workflow, how to iterate with it, when to challenge it, when to trust it and when to check its work.

We’re still doing recognizable work. We’re just doing it differently.

And then there’s adapt.

That’s when the question changes.

Instead of asking, “How can AI help us do this?”

We start asking, “Knowing what AI can do, should we still be doing this this way?”

That’s a very different question.

Maybe AI can make a fourteen-step process significantly faster. Great, then make it faster. But at some point, someone should probably ask why it has fourteen steps.

And the answer might be that it needs all fourteen. Some friction exists for a reason. Some reviews matter. Some human interactions shouldn’t be optimized away. Sometimes the right answer is that we have a perfectly good horse and we’ve made it faster.

That’s still success.

But sometimes the answer is that the process exists because of limitations that aren’t limitations anymore.

That’s adaptation.

Think about the monthly report:

Use: AI helps me build the report faster.
Adopt: AI gathers, analyzes, and summarizes the data as part of the reporting workflow.
Adapt: We realize nobody actually needs a monthly document. The information can be continuously available, with AI surfacing exceptions, changes, and things requiring attention. The report disappears.

And this is also where I think our obsession with measuring AI gets interesting. We like things we can count. And when it comes to paying for AI, it certainly counts.

Monthly active users. Prompts. Licenses. Hours saved. Tasks completed. Percentage of employees who have tried the tool.

Those numbers are meaningful. They’re what we have available to help justify the spend. If you’re trying to understand usage, they’re exactly the kinds of things you should measure.

But usage isn’t adoption. And adoption isn’t adaptation. If the goal is use, measure whether people use it. If the goal is adoption, look for changes in behavior and workflows.

If the goal is adaptation, look at the work itself.

What disappeared? What changed? What became possible? What are we doing now that we couldn’t reasonably do before?

And maybe, eventually, we stop measuring AI at all.

We don’t measure email adoption anymore. We don’t celebrate the percentage of employees who opened a web browser this month. Those technologies became part of the environment in which work happens.

AI will get there too.

At some point, perhaps the interesting metric isn’t whether AI was involved in getting us from here to there. It’s whether we got where we needed to go.

So maybe the progression is:

Use → Adopt → Adapt → ?

I don’t know what comes after adapt. And I’m increasingly suspicious of anyone who says they do.

Whatever we name today will be based on what we understand about work today. And if this technology really does change what work looks like, some of what comes next should be difficult for us to imagine from where we’re standing now.

Otherwise, we’re probably just imagining another faster horse.

And again, there’s nothing wrong with a faster horse.

Just don’t assume it’s the only thing we can build.