At some point in almost every conversation I have with a founder, they say some version of the same thing: “We tried it and it didn’t really work for us.” They signed up for a tool. They gave their team access. Nothing much changed. They moved on.
That’s not an AI problem. That’s an onboarding problem. And once you see it, you can’t unsee it.
The mistake hiding inside “we tried AI”
Most businesses approach a new AI tool the way they’d approach new accounting software. You buy it, you set it up, you expect it to work. If it doesn’t produce results in the first week or two, it’s probably not the right fit. On to the next one.
That model works fine for software that does a fixed job. A invoicing tool either sends the invoice or it doesn’t. There’s no judgment involved, no interpretation, no ambiguity.
AI tools are different. They involve judgment. They produce outputs that need to be reviewed, corrected, and shaped over time. They have to learn your business — your tone, your priorities, your standards — and that learning doesn’t happen by accident. It happens because someone is paying attention and doing the work of teaching.
The businesses I’ve watched struggle with AI are almost always treating it like a feature they’ve turned on. The businesses that succeed are treating it like someone they’ve hired.
What the analogy actually means in practice
Think about the last time you hired someone new. If you’re a thoughtful manager, you didn’t hand them a laptop and say “figure it out.” You explained the role clearly. You described what good work looked like. You checked in during the first few weeks. You expected some mistakes, and you caught them before they became expensive ones. You gave feedback. Gradually, you trusted them with more.
That process — define, review, correct, trust — is exactly what successful AI adoption looks like.
A founder I worked with recently was using an AI tool to help draft client-facing content. The first two weeks were underwhelming. The output was generic. It didn’t sound like them. They were close to abandoning it. But instead of switching tools, we sat down and did something simple: we defined the role properly. We wrote down what the content needed to do, who it was for, what the business’s voice actually sounded like, what “good” meant in concrete terms. We looked at the early outputs side by side with examples they were proud of and talked through what was missing.
By week four, the tool was producing first drafts that needed about fifteen minutes of editing instead of a full rewrite. By week eight, they’d saved somewhere between eight and ten hours a month on content that used to be a genuine bottleneck.
Nothing changed about the tool. The onboarding changed.
The 30-day rule for any AI rollout
The pattern I see across the businesses that get this right comes down to a simple principle: the leader stays close for the first thirty days.
That doesn’t mean doing the work themselves. It means reviewing outputs regularly in the early weeks. It means being available when the team has questions about what “good” looks like. It means treating early mistakes as information rather than evidence that the whole thing was a bad idea.
Businesses that assign AI to someone junior and walk away tend to end up with one of two outcomes. Either the tool gets quietly abandoned because nobody had the authority to make real decisions about how it should be used. Or it gets used — but in a watered-down way that doesn’t touch the parts of the business where it would actually matter.
The founders who see real results are the ones who treat the first month as an investment, not a test. They’re not hovering. They’re onboarding.
This also means setting realistic expectations before you start. AI will make mistakes in the early weeks. It might misunderstand your tone. It might miss something obvious to anyone who knows your business. That’s not a sign that it’s broken. That’s what the first thirty days are for. You catch the mistakes, you correct them, and you build the context that makes the tool genuinely useful over time.
What to actually do this week
If you’re thinking about rolling out an AI tool — or if you’ve already rolled one out and it’s underperforming — start here.
Write down the role you want AI to play, as specifically as you’d write a job description. Not “help with content” but “draft the first version of our weekly client update email, in a tone that is warm but professional, based on bullet points I provide.” Not “assist with customer enquiries” but “answer questions about delivery timelines and our returns policy, escalating anything outside those topics to a human.”
Then commit to reviewing the outputs personally for the first few weeks. Not to do the work — to calibrate it. Give feedback as if you’re managing a new hire, because in most meaningful ways, you are.
The businesses that treat AI rollout as a management task, not a technology task, are the ones that end up wondering why they waited so long.
