Why saying "use more AI" doesn't work.

The AI news is all doom and gloom at the moment, and people are scared. The people in your company are reading those headlines too. Then the message comes down from the top: "We all need to use more AI."

Very reassuring. Also very actionable!

Maybe you're the one saying it. Maybe it's being said to you. Either way, it doesn't work.

I know because I've done my own version. Every week I pushed people to build their own AI skills, coming at it from a different angle each time, and nothing stuck beyond the session itself. Turns out my powers of persuasion have limits.

When I think about what actually gets people using AI, it's a mix of things. The framing matters most, and the rest follows in this order:

• The way it's framed. AI takes the tedious bits of their job so they can get on with the bigger stuff.
• A tool doing a real job. The kind of thing they'd normally wait on another person for.
• A peer using it well. Someone like them, getting real work done with it.

Here's the main idea: help people see how AI can elevate their role, don't scare them with adopt or die.

Lovely on paper. So how do you actually do it with a team that's already scared?

Let's go!

"Use it more" sounds like a threat right now

Put "use more AI" next to those headlines and your team hears something else: adopt or die. It doesn't matter that you meant it kindly, or that you're genuinely excited about it.

I think adopt or die is the wrong message, full stop. Tell people to use this or else and they get scared. And if you ask me, scared isn't a great place to learn anything new.

It's expensive too. Every time adopt or die comes down from the top, I end up spending about six months unpicking it with everyone else. Six months! For three words.

Even with the fear taken out, the instruction aren't clear.

Which job? With what? Is it OK to get it wrong?

Put yourself in their shoes: picture whoever does your month-end report hearing "use more AI" on Monday. On Tuesday they've still got the same report in front of them, no idea where AI fits in, and a quiet worry about what happens if a wrong number gets through. A bare "use it more" has none of the three things that actually worked: no framing, no real job and no peer to watch.

So start where it goes wrong, with the words you use.

Start with the tedious bits

The framing goes first on that list for a reason. Get the words wrong and nobody gets as far as the tool.

So this is what I'd say to your team, out loud and more than once:

AI is here to take the tedious parts of your job, so you can concentrate on the bigger, more exciting problems.

You're still the one doing the job. You just get to skip the worst bits of it.

And don't pretend the headlines aren't scary. They are. I think AI is scary and exciting at the same time, and I lean towards excited. Say both. "This is a big change, some of it feels uncomfortable, and I think it'll make your job better" is far easier to believe than a cheery "nothing to worry about, everyone!" from the person who sent the "use more AI" email.

Then stop talking about AI in general and talk about their job. Nobody gets excited about "AI". They get excited about never doing that one awful bit of their week again. So ask each person one question:

Which bits of your week would you happily never do again?

If the answer comes back vague ("admin, I guess"), ask when it last happened and what they were actually doing. You want a real task from a real week. Write every answer down, because that list is where your first real job comes from.

For engineers it can be as simple as no more debating formatting rules. AI sorts that out, and they just build features. For everyone else, it's whatever is on that list. The time that comes back goes to the bigger problems, to getting better at the job and to the relationships with colleagues and clients that deserve more of their week.

The words get people listening. They start actually trying it when they see it do something real for them.

Do it with them, on a real job

What moves people is watching AI do a real job. Ideally one they'd normally sit and wait on another person for. Pulling the data they need, fixing a bug, adding a small feature that makes their own job a bit smoother.

In your company, that's probably the figures for the monthly report, the quote waiting on whoever knows the pricing, or the client reply that needs three bits of information from three different places.

That waiting matters. When someone usually waits on another person and AI gets it done while they're sat right there, they get it. A slick demo of AI writing a poem about your company values won't do that (please don't).

So leading by example means two things your team can actually see.

1. You use it yourself, where they can see it. Use it on your own real work, and say out loud that AI helped with it. If they never see you use it, "use more AI" is just homework you've handed out.
2. You sit next to one person and do one of their jobs together. Use their report, their numbers and their actual week (only stuff you're both allowed to share, obviously).

Pick that first job from the list they gave you when you asked about the tedious bits. Choose the one that comes round most often and annoys them the most. If it goes a bit wrong, that's fine. You're both there to see why and try again.

You've got the right result when they have the real thing done and know how to do it again without you. If they'd still need you next time, sit with them once more.

I'm firmly on the learn-by-doing side. A course can tell people what AI can do. Doing their own job with it, once, with you next to them, shows them what it can do for them.

The catch is that one 1:1 helps one person. And you can't sit next to everyone. Well, you can, but then nothing else gets done.

Then they do it with someone else

So don't. Do the first few 1:1s yourself, then hand it on.

You sit with one person and do one of their real jobs together. Then that person sits with someone else and does the same with one of theirs. The change spreads because you started it and then handed it on.

That's the third thing on the list from the start: a peer using it well. I think it does something you can't do as the boss. When you use AI, it's easy for people to think "well, of course they do". When someone with the same sort of job uses it properly on real work, they can see that people like them can do this too.

That's how my skills pushing finally got somewhere (the full story's in Issue 08). I built one with a colleague, we used it internally and I gave them the credit publicly. After that, a fair few more skills turned up. My read is that people didn't see me as a peer. I was the one running the sessions, which turns out to make me the least convincing advert for them.

This is what I call the skills flywheel. Whoever built the last skill is the best person to show the next person how, because they've only just done it themselves. A 1:1 works the same way: the person you sat with last week is now the best person to sit with someone else. And the first colleague isn't special. Anyone on the team can pass it on.

So when someone has done their first one with you, ask them to do the same with one colleague. Then give them the credit out loud, somewhere the whole team can hear it.

Which leaves an awkward question. Are you actually doing any of this, or just saying "use it more" a bit more nicely?

Something you can implement by Friday

Now for the main event: let's find out whether your team actually sees you leading by example.

Send your team a short survey this week and ask them whether they see you doing it.

A bit scary? A little. That's sort of the point.

You'll need your team or department, whatever form tool you already use (Google Forms, Microsoft Forms, whatever's lying around) and about 10 minutes to set it up. Make it anonymous if you can (scared people don't answer honestly with their names on it).

Before you send it, write down what you think they'll say. Actually write it down, please. The gap between your guess and their answers is the interesting bit.

Then send it with a one-line note that uses the framing from earlier. Something like: "I want to find the tedious bits of your week that AI could take off your plate, and to check whether I'm actually helping. Be honest, it's anonymous."

The questions:

1. In the last month, have you seen me use AI on real work? (yes / no / not sure)
2. When I talk about AI, how does it land? (1 = it worries me about my job, 5 = it sounds like help with the boring bits)
3. Which tedious part of your job would you happily hand to AI?
4. Has anyone sat with you and used AI on your actual work? (yes / no)
5. Have you seen a colleague, not me, use AI well on a real job? What was it?
6. Have you shown anyone else how you use it? (yes / no)
7. What would make you more likely to try it this week?
8. Has anything you've heard about AI here worried you?

They run in the same order as this issue. The first two cover what people see you doing and how your words land. Questions 3 and 4 cover the real job done together, and 5 and 6 tell you whether anything's been handed on. The last two give people room to tell you what you didn't think to ask.

How to read the answers:

• Mostly "no" on question 1? Start using it where they can see. It costs nothing and you can start tomorrow.
• Low scores on question 2? Fix the words before anything else. Go back to the tedious bits.
• Question 3 is your list of first 1:1 jobs. Congratulations, you've got a backlog.
• "No" on question 4? Book the first 1:1. This week.
• "No" on 5 and 6? It's stuck with you. Nothing has been handed on yet.

Read every answer to question 8 properly. If someone's worried, start with that conversation.

Then do the check. Put your written guess next to their answers. The places where you were wrong are where you start. Tell the team what you heard and which 1:1 you've booked, so the survey doesn't turn into one more thing you asked them to do and then never mentioned again.

By Friday, aim for one survey sent, one list of tedious jobs and one 1:1 booked.

If it worked, tell me. If it didn't, tell me too. I'd love to hear what came back.

But I'm not AI-pilled, what can I do?

Maybe you read question 1 and winced. They haven't seen you use AI on real work because, honestly, you don't use it on real work yet.

That's normal. Nobody's born AI-pilled. There is no manual yet. (OMG so meta!)

But you can't lead by example with something you haven't done. So I'd start the same way as the rest of this issue: someone sits with you and your team and does the first real job together. That's what our 30-Day Hands-On Sprint at ViableCloud is.

• One larger job, done with your team. Your engineer comes to your office. We watch the work with your team, check past examples and run the job beside a person before it goes live.
• Time with you and your team. Four hours of founder coaching, four weekly 45-minute team sessions, the playbook and 60 days of support.
• Two founder hours a week. That's your bit. Probably less time than you spent on doom headlines this week.

The promise is a job running by day 30, or we keep working free until it is, as long as you give those two hours a week and the team shows up. The written scope sets the job, access, dates and support before you commit.

Why I think it's a great place to start: your team gets a real job done with them, on their own work, instead of another instruction. You get your first real example to point at when you talk about AI. And with the founder coaching, the playbook and 60 days of support, you've got something to lean on when it's your turn to do the 1:1s.

The full details are on how it works.

No manual yet

"Use more AI" doesn't work because it skips the three things that actually get people going.

The words you use, so it sounds like help with the tedious bits. A real job, done with them, on their own work. And a peer to watch, so they can see that people like them can do this too.

Then you start it and hand it on. You sit with one person, they sit with the next, and the change spreads.

I'm still ridiculously excited about what we can build now, especially when the person you sat with starts showing someone else.

I'd just like one headline about AI taking the most boring bit of someone's job.

You survived another week in this strange new world AND another one of my articles. Amazing!

Over to you now: what actually got you using AI, someone telling you to or someone showing you?

I've made Why saying "use more AI" doesn't work: the survey, the prompt and what to do next to use with this issue.