Where Does AI Fit? Use the Intern Test

Every week someone asks me where AI fits in their business. They usually expect a technology answer: which tool, which vendor, which feature. The most useful answer I have found is not about technology at all. It is a question about people.

Before you hand a task to AI, ask yourself: would I give this to a smart intern in their first week?

I use that test with clients as a fractional CTO and AI advisor, and I use it in my own companies, where I make these calls every week. It sorts work faster than any framework deck, and it is honest about what AI is actually good at today.

Why the intern test works

Think about what you would hand a sharp new hire in week one. You would give them work that repeats, so they get practice. Work with steps you can write down on a page. Work where a mistake is easy to spot and cheap to fix, because you know they will make some.

That description is also a near perfect description of where AI earns its keep. Sorting incoming requests. Summarizing a long email thread. Filling in a form from a document. Drafting the first version of a routine reply. Pulling the same numbers into the same report every Monday.

Now think about what you would never give a week-one intern. Pricing a large deal. The hard conversation with an unhappy customer. Deciding which supplier to drop. Approving a refund with real money attached. That work runs on judgment, context, and relationships nobody ever wrote down.

AI is not ready for that work for the same reason the intern is not. It does not lack intelligence. It lacks the years of context that make the judgment good.

Should, could, shouldn't

The intern test maps cleanly onto the three buckets I use to decide where AI fits:

• Should: you would hand it to a week-one intern with a checklist. AI probably fits now, and the hours you get back are real.
• Could: you would hand it to a junior person, but only with close supervision. AI can fit here later. Start it on drafts and keep a person on the final say.
• Shouldn't: you would only trust your best senior person with it. Leave it with them. That is not a failure of the technology. It is knowing what the job actually is.

Most businesses find that a surprising share of their week lands in the first bucket. That is good news. It means the first AI win is probably smaller, cheaper, and closer than they thought.

The mistake I see most often

Companies usually run the test backwards. They give AI the senior job first, because that is where the big number lives and the demos make it look easy. The AI produces confident answers that nobody can easily check. The senior person now spends their time supervising a tool instead of doing their job. And the intern-level work that was eating the team's week is still sitting there, untouched.

After 35+ years in software and security, I have learned that trust gets built from the bottom up. Nobody starts their career approving five-figure refunds. AI should not start there either.

Promote AI on evidence

The companies I see getting real returns from AI are not the boldest ones. They are the patient ones. They treat AI the way a good manager treats a new hire:

• First month: week-one work only. Sorting, summarizing, drafting, filling in. Someone looks at the output every day.
• Once you have watched its work: junior responsibilities. Drafting replies a person sends. Flagging the invoices that look wrong. Preparing the quote a person checks.
• Senior judgment comes last, if at all: and only for the narrow slices where you have months of evidence that it gets it right.

Each step up is a promotion, and promotions are earned. If the work at one level is not reliably good, it does not move up. That one rule prevents most of the expensive AI mistakes I get called in to clean up.

Where the hours actually are

Here is the quiet benefit of starting at the bottom of the org chart. Go look at your most experienced person's calendar. In most businesses, a real chunk of it is intern work: retyping details from one system into another, chasing missing paperwork, writing the same status update for the fifth time.

Hand that to AI and you are not replacing your senior team. You are giving them their week back, so the person with 20 years of judgment spends their hours on the work that needs 20 years of judgment. That is where AI fits, and it is where the return shows up first.

Try it this week

List ten tasks your team did last week. Next to each one, write the most junior person you would trust to do it alone. Anything marked "week-one intern" is your shortlist. Pick the one that repeats most often and start there.

If you want a fractional CTO and AI consultant to help you sort your own list and find where AI actually fits in your business, I take a few clients at a time. Book your free 15-min AI Fit Call →

Not ready for a call? Start with the free AI Fit Audit and see where your business stands in a few minutes.