I’ve noticed a quiet guilt creeping into conversations about AI. People tell me, almost apologetically, that they let Copilot write an email or summarise a meeting. As if they’d cheated. I don’t think the question is whether you should hand work to AI. You should, and lots of it. The better question is which skills you deliberately keep for yourself. Because if you don’t choose, the choice gets made for you, one convenient shortcut at a time. Some abilities fade quietly when you stop using them, and you only notice when you need them most.
Choose on purpose
My test is simple. I ask three questions about any piece of work.
- Does this define the value I bring? If a client pays me for my judgement, my judgement stays in my head.
- What would I need if the AI was wrong, or simply not there? Someone has to recognise a bad answer and fix it.
- Do I actually enjoy doing it? Some work feeds me. I’m not giving that away just because I can.
If a task scores on any of those, I practise it by hand. Not always, but often enough that the muscle stays strong. Everything else is fair game for delegation, and I hand it off without apology.
For me the keepers are clear. Thinking through a client’s real problem. Writing my own ideas in my own words. Core troubleshooting, the kind where you work from symptoms back to cause. And the relationships, the conversations where trust is built. I wrote earlier about why everyone needs basic technology troubleshooting skills, and this is the personal version of that argument.
What handing off looks like
Picture a typical Tuesday. My inbox is full of routine requests. I let Copilot in Outlook summarise the long threads and draft the straightforward replies, then I check and send. A recurring project call finishes in Teams and I lean on the meeting recap for actions instead of scribbling notes. Someone sends me a spreadsheet of licence data and I ask Copilot in Excel to spot the trends and build the summary. None of that defines me. It just needs doing, and doing well.
Now picture the same day with a proposal due for an important client. That one I write myself first. A rough draft, in my words, with my reasoning. Only then do I invite Copilot in Word to tighten it, challenge it or check the structure. The thinking is mine. The polish can be shared.
Same with a nasty technical fault. I form my own theory before I ask for help. If Copilot agrees, great, I’ve confirmed my reasoning. If it disagrees, I’ve got something to learn from, rather than an answer I can’t judge.
The order matters
That’s the real trick. For the skills you want to keep, do the hard part first and bring AI in second. For everything else, let AI go first and you review. Same tools, different sequence, very different outcome over a year.
None of this is about rejecting AI. I use it constantly and I’m better for it. It’s about being intentional, the way you’d choose which muscles to train at the gym. Agents in Microsoft 365 will take on more and more of the routine work, and I welcome that. It frees time for the things only I can do. As I said in Give It Time to Become Something, value builds through steady practice.
So pick your keepers. Write them down if it helps. Practise them by hand, on purpose, every week. Then hand off the rest and stop feeling guilty about it. Your attention is the scarce resource now. Spend it where it counts.
