There’s a moment I see in almost every business that’s struggling with AI. They bought a tool, set it up, gave it to the team — and then waited for things to improve. A few weeks later, nothing has really changed. The tool is technically running. Nobody’s using it well. And somewhere in the back of their mind, the founder is wondering if AI just isn’t right for their business.
It is right for their business. They just adopted it the wrong way.
## Most businesses treat AI like a feature, not a hire
When a business buys new software, the instinct is to flip it on and expect it to work. Software mostly behaves like that. You install accounting software, and it does accounting. The rules are defined. The output is predictable. You don’t need to manage it.
AI doesn’t work like that.
A better analogy is hiring someone new. When you bring on a new employee, you don’t hand them a login and disappear. You explain what you need. You show them examples of what good looks like. You check their work in the early weeks, not because you don’t trust them, but because alignment takes time. You expect a few mistakes. You correct course. Gradually, they need less of your attention.
The businesses I’ve seen get real, lasting value from AI are doing exactly that. The ones that treat AI like a feature — something you activate and walk away from — almost always stall.
## What the onboarding mindset actually looks like in practice
The pattern shows up most clearly in how leaders respond to the first month after introducing an AI tool.
A client I worked with runs a small professional services firm. Before we started, she’d tried two different AI tools and written them both off as “not quite there yet.” When we introduced a third, she was skeptical but willing. What changed wasn’t the tool. It was how she approached it.
We started by defining the role clearly — not in technical terms, but in business terms. What tasks would this tool own? What did good output look like? What were the edge cases where a human needed to step in? She treated it like writing a job description, because in a meaningful sense, that’s what it was.
For the first four weeks, she or someone on her team reviewed every output before it went anywhere. Not to be a bottleneck, but to build the kind of understanding you can only get by watching something work up close. They caught patterns. They adjusted the instructions. They trained it — and trained themselves.
By week six, the review process was down to spot-checks. By month three, the tool was handling a full category of work that had previously taken her team several hours a week.
None of that would have happened if she’d installed it and moved on.
## The three questions that make the difference
When I work with founders on an AI rollout, I ask three questions before we touch any tool. They’re the same questions you’d ask before bringing on a new hire.
**What is this person here to do?** Not in vague terms — specifically. Not “help with content” but “draft first versions of client proposals based on a brief, in a tone that matches our voice guide.” The more precisely you define the role, the better the output. Vague roles produce vague results, whether the person is human or not.
**How will you know if they’re doing it well?** You need a standard before you can evaluate. What does a good output look like? What would make you send it back? If you can’t answer this, you’re not ready to delegate the task to anyone — AI or otherwise.
**Who is responsible for reviewing the work in the first 30 days?** This is the question most businesses skip. They introduce a tool with no one accountable for its performance. If no one is watching, no one will catch the drift when quality slips. Assign it. Put it in someone’s role.
These questions don’t require any technical knowledge. They require you to think clearly about your business and your standards, which you already know how to do.
## Start with one role, not a transformation
The businesses that struggle most with AI are usually trying to do too much at once. They want to overhaul their marketing, automate their customer service, and streamline their operations simultaneously. That’s not a rollout, that’s a renovation with every wall open at the same time.
Start with one role. One task category. One clear owner for the review process. Give it 30 days of real attention. Adjust as you learn. Then, and only then, expand.
The same discipline that makes a good hire successful makes an AI adoption successful. Clear expectations. Early oversight. A willingness to correct course without abandoning ship.
If your previous attempt at AI didn’t work, it’s worth asking whether the problem was the tool — or whether the tool was never really onboarded at all.
