Free Prompts to Onboard Your AI Intern

The onboarding brief: give it context up front

The very first thing you'd do with a new hire, after you say hello, is tell them about your business. Do the same here.

You're joining my business as a new assistant. Before we start any task, here's what you need to know.

What we do: [one or two sentences about your business]
Who our customers are: [who they are, in plain words]
What they need from us: [the problem you solve for them]
How we help them: [your products or services, and what makes you different]
How we talk: [a few words about your tone, plus one example of something you've written that sounds like you]

Your role: [the job you want help with, for example "draft our social media posts"]

Keep all of this in mind for everything I ask you in this conversation.

Two things that make the brief work better, both from real research:

• Put the most important part first or last. AI models use what's at the start and the end of what you give them far better than what's buried in the middle (Stanford and UC Berkeley, "Lost in the Middle," 2024).
• Keep it relevant to the task. More context isn't automatically better; relevant context is (IBM Research and Rensselaer Polytechnic Institute, 2024).

Ask it what it needs

When someone asks me how to get AI to do something, my first question is: did you ask it what it needs? Usually they didn't. AI will often tell you exactly what's missing.

Before you start: what do you need from me to do this well? Ask me your questions one at a time, and don't start the task until you have what you need.

Its first assignment: one concrete task

Don't start with "help me with my business." Start the way you'd start a new intern: with one specific job. The best first task is one you do over and over and would rather not.

Here's your first assignment. Every [day / week], I want you to write [a social media post] for [where it gets posted].

Start from this: [paste a customer experience, a review, or a note about something that happened this week].

Here are two posts I've written that sound like me:
[paste them]

Keep it under [length]. End with [what you want readers to do].

Coach it, like you would a new hire

Its first try won't sound like you. It won't know your customers' language or your industry's jargon yet. That's normal. You'd gently correct a new intern, show them examples, and they'd get better every time. Same here.

Close, but not quite. Here's what to change: [what's off: the tone, a word we'd never use, a detail that's wrong].

Here's a version I'd actually post:
[your fix]

Remember this for next time, and tell me what you changed.

If your AI tool has a memory setting, turn it on. The coaching you do now carries over, and that investment compounds.

Trust, but verify: check the work before it goes out the door

You wouldn't let a brand-new hire's work go straight to a customer unchecked. Don't do it with AI either.

Before anything AI wrote reaches a customer:

• Facts and numbers. Check every stat, date, price, and name against the real source. AI can state things confidently that aren't true.
• Your business. Are the details about your products, prices, and policies right?
• Your voice. Does it sound like you, or like a brochure?
• The reader. Could anything confuse, overpromise to, or annoy the person reading it?
• The next step. Is the call to action (call, reply, book, buy) the right one?
• The gut check. Is there anything in it you wouldn't want to explain later? Cut it.

Why this matters: in a 2026 study reported in Harvard Business Review, managers caught 18% fewer errors in the same AI work when it was labeled an "AI employee" instead of an "AI tool." When AI feels like a colleague, we trust it too much. Treat it like an intern, and still check the work.

Ask for its ideas

The best interns I've worked with brought ideas I never would have thought of. AI can do the same, but only if you ask.

Now step back. What would you do differently here? What am I missing? Give me two or three ideas I haven't asked for.