Delegate to AI like it's a new hire

To get better results from AI, brief it like a new hire: give it a role, the outcome you want, the audience, examples of good work, constraints, and a review step. In a 2023 MIT study published in Science, professionals using ChatGPT finished writing tasks about 40% faster with about 18% higher quality, but results depend on clear instructions and careful review.

I've watched a lot of business owners try AI for the first time. It usually goes something like this.

They type: "Write a newsletter for my business."

They get back something generic, bland, and slightly off. It sounds like every other newsletter on the internet.

And they decide: "AI isn't that useful."

But imagine handing that same instruction to a brand-new employee on their first day. No context. No examples. No idea who your clients are or what you sound like. "Write a newsletter for my business."

You'd get something generic back from them, too.

Most bad AI output isn't an AI problem. It's a delegation problem.

What the research shows

In a 2023 experiment published in Science, MIT researchers Shakked Noy and Whitney Zhang gave 453 college-educated professionals (marketers, grant writers, consultants, HR professionals, and others) writing tasks from their actual line of work. Half were given access to ChatGPT.

The group with ChatGPT finished about 40% faster, and independent evaluators rated their work about 18% higher in quality.

That's a big boost. But there's a detail in the research that should make you pause. Many participants who used ChatGPT submitted its text with very little editing. And the researcher Shakked Noy noted that in the real world, the speed benefits would be smaller, because you need time to write good instructions and fact-check the results.

In other words, the gains are real. But they depend on how well you direct the work, and how carefully you review it. Which is exactly what makes a good manager.

Treat it like a talented new hire

Here's the mindset shift that changes everything: AI is like a new employee who's fast, capable, and knows a little about almost everything, but knows nothing about your business.

You wouldn't expect a new hire to read your mind. You'd train them. You'd show them examples. You'd explain your clients, your standards, and your voice. Then you'd review their first few projects closely.

AI needs the same thing.

If you read my post on delegation, you'll recognize this. The same brief that keeps work from boomeranging back to you from a person works on AI, too.

The AI delegation brief

Before you ask AI for anything important, give it:

1. The role. "You're helping me write my weekly newsletter for small business owners who want accountability coaching."

2. The outcome. Not just the task, but the goal. "The goal is to get past clients to reply and book a check-in call."

3. The audience. Who is this for? What do they care about? What are they struggling with?

4. Examples. This is the most important one. Paste in two or three of your best past newsletters and say, "Match this voice." Examples communicate more than any description can.

5. The constraints. Length, format, what to avoid. "Under 400 words. No exclamation points. Don't mention pricing."

6. A review step. Ask for a draft, not a final product. Then edit it like you would a new employee's first attempt: specific, direct feedback about what to change.

Build it once, reuse it forever

Here's the best part. With a new employee, you train them once and they remember.

With AI, you can do the same thing. Save your briefs. Most AI tools now let you create saved instructions or custom assistants. Build a brief for each task you do regularly (newsletters, client follow-ups, social posts, proposals) and reuse it every time.

That's the difference between using AI and building an AI system. One gives you random results. The other gives you a reliable assistant that already knows how you work.

Stay the manager

One last thing: a new hire, no matter how talented, still needs a manager. So does AI.

You're responsible for what goes out with your name on it. Review the work. Check the facts. Make the final call. AI can do a lot of the drafting. The judgment is still yours. So the next time AI gives you something generic, don't give up on it. Ask yourself: Did I delegate this well?

Usually, the answer is the fix.

Delegating well is a habit, and habits stick faster with accountability. If you want someone in your corner every week, here's how I coach.

Study to link: Noy & Zhang (2023), Science · MIT News summary

Frequently asked questions

Why does AI give generic answers?

Usually because the instructions were generic. Like a new employee on day one, AI knows nothing about your business, clients, or voice unless you tell it.

What should I include in an AI prompt for business writing?

Include the role, the outcome, the audience, two or three examples of your best past work, constraints like length and tone, and a request for a draft you'll review.

Does ChatGPT actually make professionals more productive?

In Noy and Zhang's 2023 experiment with 453 professionals, ChatGPT cut writing time by about 40% and raised quality ratings by about 18%. The researchers noted real-world gains may be smaller once you account for writing instructions and fact-checking.