With AI, You Need to Be a Manager

Business owner briefing and delegating to his team around a meeting table in teal and amber light

Here’s something I didn’t expect when I started working seriously with AI. The skills that matter most aren’t technical. They’re managerial. The more capable these systems become, the more they behave like staff, and the more you need to know how to manage people to get the best out of them.

Think about it. Copilot in Microsoft 365 drafts, summarises and analyses. Agents take on multi-step tasks. Assistants like the one I wrote about in Finally, AI That Just Gets Things Done go off and actually do the work. That’s not a tool you click. That’s a team member you direct.

Set clear goals

A new employee who’s told “sort out the client files” will guess what you mean. So will AI. Good managers define the outcome, the deadline and what done looks like. Do the same with your prompts and agent instructions. Tell Copilot who the audience is, what format you need and what success looks like. Vague direction produces vague work, whether it comes from a person or a model.

Delegate properly and give context

Delegation isn’t dumping. When I hand work to a staff member, I explain why it matters, what’s already been tried and where the traps are. AI needs exactly the same briefing. Point Copilot at the right SharePoint documents, the relevant Teams thread or the Outlook email chain. Context is the difference between a generic answer and one that fits your business.

Delegation also means deciding what not to hand over. Some decisions stay with you because they need judgement or carry real consequences.

Check the work

No sensible manager signs off on a new hire’s first report without reading it. Yet I see people paste AI output straight into client emails. Review it. Check the facts, the tone and the numbers. AI is fast and confident, which is precisely why it needs checking.

Give feedback

Good staff improve because someone tells them what worked and what didn’t. AI is the same. Refine the prompt, correct the draft, update the agent’s instructions and save what works. Over time your instructions become the equivalent of a well-written role description and training manual.

Trust, but verify

As people prove themselves, you give them more responsibility. That’s how I treat AI too. My monitoring bots across Microsoft 365 are read-only by default. When something needs changing, they list what they plan to do and wait for my explicit yes. I explained that pattern when I wrote about how I set up Grok Bot to close Sentinel and Defender incidents safely. Trust grows with evidence, and verification never fully goes away.

Why this matters for MSPs

For MSPs, this is a real opportunity. Most of your clients have never been taught to manage AI, and many have never been taught to manage people either. Picture helping a client’s office manager write clear Copilot instructions, set up review steps and decide which tasks an agent can own. That’s coaching, not configuration, and it’s worth paying for.

It also applies inside your own business. Your technicians are about to become managers of digital workers. Invest in their management skills now:

  • Writing clear briefs and expected outcomes
  • Choosing what to delegate and what to keep
  • Reviewing output with a critical eye
  • Giving feedback that improves the next result

The best AI users I know aren’t the most technical. They’re the best managers. If you want better results from AI, stop just learning the tools and start learning how to lead people well.

Leave a Reply