Kotter asks a powerful question in his work on change. What are the challenges in front of your team, and what’s possible if it innovates and thinks differently?
That question doesn’t change in the age of AI. The work to answer it does.
Something I keep noticing
Here’s something I keep noticing in my own day, and in the teams I talk to. People are doing more individual work in the age of AI, not less.
That’s the opposite of what the research says should be happening. AI’s real value shows up when people collaborate. We share knowledge. We pressure-test each other’s thinking. We make better decisions when more than one brain is in the room. And yet here we are, working alone.
And it isn’t a tools problem. We have more ways to work together than we’ve ever had. It isn’t a proximity problem either. Return-to-office has put us back in the same rooms for 18 months. The gap is somewhere else.
The 20-page PRD
I had a brief exchange with a senior leader recently. He told me his teams are now producing PRDs that run 20 pages or more. AI-generated. Heavy on detail. Light on judgment.
The people writing them love it. He’s frustrated. He can tell the human didn’t vet the work. It reads as fluff. It lands as unprofessional.
His phrase for it was sharp: “AI as a BA.” His read was that senior people were automating away the work they didn’t like (PRDs, stories, requirements) so they could focus on the work they did (talking to customers).
There’s a name for the output side of this. Work slop. Polished on the surface. Hollow underneath. Easy to spot when you know what you’re looking for, and corrosive to trust when it lands in a leader’s inbox.
The work doesn’t go away. It just changes shape.
Here’s where I’d push back a little, carefully.
The instinct to automate the work you don’t like makes sense. The risk is that you also automate the thinking that work was doing for you. The PRD wasn’t just a document. It was a forcing function. It made you sit with a problem long enough to know it deeply.
AI doesn’t make that analysis disappear. It relocates it. Pulling in domain knowledge, synthesizing what matters, and critically assessing the output all sit with the person delegating now. That’s a shift, not a subtraction.
That’s still business analysis. It just looks different.
We’re training the wrong skill
Most of the AI training I’m seeing right now is prompt training. How to phrase the ask. How to chain the steps. How to squeeze more out of the tool.
That’s the ‘how’ of AI. It matters, but it isn’t enough.
The piece that’s missing is the “who”. The human doing the delegating. What they choose to hand off. How they frame the ask. What they check before the work leaves their hands.
That’s a different skill. We aren’t teaching it.
Stop treating AI like a calculator
Most people are still using AI the way they use a calculator. Type the question. Trust the answer. Move on.
A calculator has earned that trust because math is the math. AI hasn’t earned it. The math isn’t the math.
A better mental model is to treat AI output like a draft. You expect to edit it. You expect to question it. You expect to put your name on the final version only after a human (you) has done real work on top of it.
I’ve been calling the skill, Delegation Intelligence. Knowing what to hand off. Knowing how to frame the ask. Knowing what to check before the work leaves your hands.
It’s a muscle many people haven’t built yet.
Three ways to start building it
The Red-Pen Test. Old school, on purpose. Print an AI-generated draft. Hand someone a red pen. Their job is to find what the AI got wrong, what it missed, and where it lost the voice. This builds the human-in-the-loop muscle in a way that staring at a screen does not.
Review the prompt, not just the result. In your next 1:1, don’t just look at the deliverable. Ask to see the prompt that produced it. If the prompt was vague, the output will be vague. Teach your team to delegate with constraints. (Documenting constraints, by the way, is business analysis.)
Move past FOBO with ADKAR. FOBO (Fear Of Becoming Obsolete) is real. Use ADKAR to build the Desire stage. Show people their value isn’t in the doing anymore. It’s in the directing.
A short note on citing AI
Anything that uses AI should say so. Quietly. Clearly. The same way research cites its sources.
This sits under transparency and explainability in any decent AI governance framework. It’s also how trust gets rebuilt after work slop. A team that cites its AI use is a team that’s paying attention to its AI use.
Stepping back
My research at Purdue looks at how deep work and concentration shift when AI enters the picture. Being an orchestrator is a different kind of cognitive fitness. More critical thinking. Less rote execution.
The manager’s job shifts, too. Coach the process. Teach people how to direct the work, not just produce it.
That’s the part I think we’re underestimating. Not the prompts. The people.
Final question
What’s one task your team is currently doing that they could be orchestrating instead?
Let’s figure it out together. đź’š
My Change Log.io 
Leave a comment