You bought the subscriptions. You sent the email announcing “we’re using AI now.” A month later, two people use it a lot, three use it once a week, and the rest quietly went back to the old way.
Sound familiar? That isn’t a tool problem. It’s a leadership problem, and that’s good news, because leadership problems are fixable.
Leading an AI-first team means your people start every repeatable task by asking “what part of this can AI draft?” and spend their own energy on judgment, relationships and decisions. Getting there takes more than access to tools. It takes a leader who goes first.
Here’s the order I’d follow.
Step 1: Model AI use yourself
Your team watches what you do, not what you say. If you announce AI and never use it in front of them, they’ll read that clearly.
Use it visibly. Share your screen in a meeting and draft the agenda with AI. Forward a summary an assistant wrote and say so. Talk about what it got wrong and how you fixed it.
I run my business AI-first, and I talk about it openly: my AI chief of staff, the agents I’m building, the things that broke last week. That openness gives people permission to experiment and to admit when something didn’t work.
Step 2: Start with pain, not features
Don’t kick off with a tour of everything AI can do. Kick off with the task everyone dreads.
Ask your team: what’s the most tedious thing you did last week? Meeting notes nobody reads. Weekly reports copied from three systems. The same customer question answered for the fortieth time. That’s your starting point.
When AI removes a task people already hate, you don’t have to sell it. They’ll ask for the next one.
- Ask each person for their two most tedious recurring tasks.
- Pick one task shared by several people.
- Build one simple AI workflow for it, with a human review step.
- Run it for two weeks, then ask what to improve.
My list of tasks to automate with AI is a good place to find candidates.
Step 3: Build a training rhythm, not a training day
One workshop changes nothing. A rhythm changes habits.
What works better is small and steady: a short live session to show one technique, a week to practice it on real work, and a few minutes in your weekly meeting to share what worked. Then the next technique.
The skills that matter most aren’t technical. They’re clear instructions, giving context, checking output, and knowing when not to use AI. My guide to how to learn AI lays out those skills for people with no technical background.
Step 4: Put AI into your Momentum Meeting
Your weekly leadership meeting is where AI adoption either becomes normal or quietly dies.
I run a weekly Momentum Meeting with a scoreboard, Rocks (our quarterly priorities), an issues list and to-dos with owners. AI helps prep it — pulling the scoreboard, summarizing Rock status, drafting the issues list. And AI adoption becomes part of the agenda:
- What did AI take off someone’s plate this week?
- Where did AI get something wrong, and what did we change?
- What’s the next task we hand to AI, and who owns it?
That last question is the engine. One handoff a week, every week, adds up. I break down the full meeting setup in AI for KPIs and meetings.
Step 5: Celebrate time back
When AI saves someone time, say so out loud. Then decide together what that time goes toward: more customer conversations, finishing the project that keeps slipping, training, getting home on time.
If the only reward for adopting AI is more work piled on, people learn to hide how much time they’re saving. Make the win visible and shared.
Step 6: Protect your people
Fear is real. People read the same headlines you do. Ignoring it doesn’t make it go away.
Talk about it directly. Be honest that the tasks are changing. Show them that the people who learn to direct AI become more valuable, not less. And back it up by investing in their skills. I wrote an honest piece on this worth sharing with your team: will AI take my job?
Protecting people also means rules. Make it clear what never goes into an AI tool, what always gets human review, and that nobody gets in trouble for flagging an AI mistake. The guardrails are in responsible AI for business.
Step 7: Redesign roles around judgment
Eventually, adoption changes the job itself. When AI handles the drafting, the data entry and the follow-up, roles shift toward review, decisions, relationships and improving the process.
Do this on purpose. For each seat, write down which tasks AI now drafts, which the person reviews, and which stay fully human. Every AI job gets a human owner. When you hire, hire for judgment and learning speed, not just for doing the repetitive work AI now handles. I go deeper on structure in AI and human team structure.
When to get help
Some leaders want to build this themselves. Others want AI running real work in their operation without becoming the IT department. That’s what Chief, my AI chief-of-staff install, is for: we set it up around how your company actually runs, and your team learns to work alongside it.
Your next step
At your next team meeting, ask one question: “What’s the most tedious thing you did last week?” Write down every answer. Pick one, and hand it to AI with a human owner by the end of the week.
For the big picture behind all of this, read the future of AI in business.
Questions people ask
How do I get my team to use AI?
Use it visibly yourself first, then start with a task your team already hates. Show them how in a short live session, give them a week of practice on real work, and share wins in your weekly meeting. Mandates without modeling rarely stick.
How do I handle employees who are afraid AI will replace them?
Talk about it directly. Be honest that tasks are changing, show them how AI removes their least favorite work, and invest in their skills. Make it clear that the people who learn to direct AI become more valuable to the business.
What is AI change management?
It’s the leadership work of moving a team from old habits to new ones with AI: explaining why, training, setting rules, measuring progress and redesigning roles. The technology is the easy part. The habits are where it succeeds or fails.