I was talking to a Director of Marketing at a small agency last week about his AI usage. This is how he described it:
"It’s one of those things where I get in there and I start playing around with it, it’s really cool and exciting, and then 3 hours later I’m like, what did I actually accomplish here?"
"What did I actually accomplish here?" That could be the tagline for a lot of AI usage these days.
If you spend enough time on LinkedIn, you might start to believe that every business challenge has the same answer: AI.
Need more leads? AI.
Faster ticket resolution? AI.
Need a stronger culture, happier employees, faster growth, better snacks in the communal kitchen? AI can help with all of that, and why haven’t you figured it out yet, dummy?
The result is that many people have found themselves in a strange position. They're surrounded by AI tools, bombarded with advice, and hearing daily stories about companies transforming their operations with AI—but they're not entirely sure what they should actually be doing.
So they do what most of us would do. They start experimenting. Using AI to draft blog posts, prepare for meetings, create email templates. Operations builds a workflow or two. Everyone gets excited.
A few months later, someone asks:
"So what are we actually getting from all this?"
Because while many businesses are using AI, they’re not seeing the kind of AI lift the headlines promised.
I think the gaping chasm between the promise of AI and the results people are actually seeing is because they’re starting with the tool first. They hear about the wonders of ChatGPT or Claude Cowork, and they start looking for places to use it.
That sounds reasonable. It’s also why many AI initiatives end up feeling random.
When your starting point is: "We have ChatGPT, what should we do with it?" You end up forcing AI into places where it doesn’t belong.
The better question is: "Where is work breaking down?"
Because "AI" is not a business strategy. It’s a tool. And like any tool, it works exceptionally well for some jobs, and spectacularly poorly for others.
AI tends to perform best when work is:
Think about the kinds of work that make you say: "I do this 40 times a week," or, "Why does this have to take so long?"
Those are often strong AI candidates. They’re the kinds of tasks where AI can create meaningful leverage because they consume time without necessarily requiring deep human judgment.
Considering this question is the part many businesses skip.
Some work should stay heavily human. AI tends to struggle when work requires:
Work like delivering feedback, navigating conflicts, negotiating partnerships, and defining company strategy—these situations contain nuance that doesn’t fit neatly into patterns. Humans are still much better at that.
And ironically, businesses often start their AI transformation in exactly these areas because they seem important. (Or maybe it’s the waterfall of articles about how an entire GTM team is run by Claude Cowork, so why isn’t yours?)
This way disappointment lies.
One of the biggest stressors about AI transformation is that it implies some kind of broad adoption. That feels huge, and it’s intimidating to start. What’s the right entrypoint in something so monumental?
In reality, many of the organizations seeing meaningful results start much smaller—with points of friction, not "AI opportunities."
They ask questions like:
Identify one of these areas of operational friction, verify it’s something AI could reasonably help with, and improve that process. Then move to the next one. And the next.
Over time, those improvements compound. The result is not an organization that uses AI everywhere. It’s an organization that uses AI intentionally.
That difference matters, because not every business problem is an AI problem. But the right business problem, solved with the right AI workflow, can create a surprising amount of lift.