AI Didn't Make My Expertise Less Important...
Updated: Aug 16
It Gave Me More to Do.
I was sitting in my home office working on what I thought was a straightforward assignment: adapt an existing development workshop for a new audience.
Then, during a planning meeting, someone mentioned that a significant portion of the audience had already experienced the workshop while supporting an earlier rollout.
My internal response was a silent scream.
I could not change a few examples, add a different audience lens, and present substantially the same experience. The assignment I thought I had was suddenly the wrong assignment.
I had roughly five business days to respond. I also had planned time away with my niece and nephew, and I did not want another emergency assignment consuming the nights and weekends around it.
As a consistent high performer, I knew where this usually ended. I had become very good at being the magician who pulls the white rabbit out of the hat.
The trouble with being good at that is that eventually everyone starts bringing you hats.
The assignment changed, so I changed the question
The original workshop on personal development had been designed for retail leaders. My first question was:
How do I change this workshop on personal development so it can be delivered to our HR business partners?
That was the assignment as I understood it. Take an existing workshop, change the audience, and adapt the experience.
Once I learned that many of the HR business partners had already experienced the workshop while supporting the retail sessions, that question was no longer useful. Changing examples and adjusting the language would still leave me delivering substantially the same experience.
There was also not enough time or needs-analysis evidence to invent an unrelated workshop responsibly.
So I asked a different question:
How do I leverage our HR business partners to help uncover what would make this workshop on personal development actually stick and become embedded in our culture? What gets in the way of people owning their development?
That question changed what I was building.
The HR business partners were no longer simply the new audience for the workshop. They became co-designers who could help diagnose the conditions that made owning development easier or harder inside the organization.
That became the seed of the co-design sprint.
Then I brought in AI.
I gave it the audience, the problem, the existing learning architecture, the time constraints, and my emerging ideas. It generated possibilities quickly, which gave me more material to react to than I could have produced alone in the same amount of time.
That did not make the work automatic.
My actual responses throughout the build sounded more like:
This is too complicated.
That will not work in the time.
Simplify it.
Give me another option.
You did not actually change anything.
This is false.
Reset.
AI helped me move faster because it gave me something to evaluate at every stage. My expertise became the filter.
AI could produce the work. It could not own the decisions.
At one point, AI generated internal statistics that sounded completely plausible.
They were not real.
It also invented a personal anecdote about me visiting two stores after becoming a director. The story sounded credible because it was written to sound like me.
It had also never happened.
AI can sound authoritative about your own life. That doesn't make it true.
Another activity was so complicated that there was no reasonable way a room of people could complete it in the allotted time.
AI did not know that the activity would collapse in the room. I did because I understood facilitation, the audience, the constraints, and what people can realistically accomplish together.
That is the part of AI-supported work we risk skipping when we focus only on output.
A first draft is not a finished decision. A polished sentence is not necessarily a true sentence. An interesting activity is not automatically a useful learning experience.
Someone still has to know the difference.
AI also introduced a better way of working
AI did not only create problems for me to catch.
Because I spent most of those five days building rather than rehearsing, I needed a lightweight way to prepare for the live facilitation. During the collaboration, ChatGPT introduced me to facilitator cue cards as an alternative to a heavy facilitator guide or presenter notes.
They were compact, practical, and easy to use while standing in the room.
I kept them.
Later, I carried the method into Retail Team Week, where Regional Vice Presidents partnered with L&D to facilitate content using printed, branded cue cards. The method stayed useful after the original AI interaction was over.
Sometimes AI does not simply accelerate the way you already work. It exposes you to a better way of working that you keep after the AI is gone.
Friday was pencils down
By Friday afternoon, the workshop had to be finished. I was flying out Monday to deliver it.
During the session, participants worked in small groups with sticky notes and flip charts, debating the real barriers that prevented people from owning their development. The room became loud with conversation.
When it was time to move on, I tried to bring everyone back.
They could not hear me.
A colleague asked, "Do you want me to whistle?"
She let out the kind of ear-piercing whistle I thought only my grandmother could make. Like a room full of dogs, everybody stopped and turned.
Five days earlier, I was worried I had nothing meaningful to give this audience. Now I could barely get them to stop talking about the problem.
The whistle is not formal evidence that the workshop solved everything. It is evidence that we had found a question worth discussing.
AI helped me get there faster.
My expertise decided where we were going.
AI gave me more to do
AI did not remove my work from the process. It changed where I spent my time.
I spent less time generating everything from a blank page. I spent more time:
Framing the right problem.
Evaluating options.
Simplifying the design.
Checking whether claims were true.
Protecting the audience experience.
Deciding what belonged in the room.
Recognizing a better method when one appeared.
AI gave me more to do, but not more of the same work. It gave me more decisions to make.
That is why I no longer begin a task by asking, "Can AI do this?"
I ask:
Does this actually need me, or does it just need to get done?
Does this actually need me?
There is a difference between work that requires my expertise and work that benefits from AI's speed.
If it needs... | I keep... | If it needs... | I brief AI... |
My judgment | The decision | A first draft | The desired outcome and constraints |
My experience | The interpretation | Organization | The material and required structure |
My relationship with someone | The human interaction | Research support | The question, source boundaries, and verification requirement |
My read on the room | The adaptation | Summarizing | The content, audience, and emphasis |
My accountability | The final approval | Documentation | The decisions and format to capture |
Briefing AI does not mean blindly accepting whatever it produces. The review should match the stakes.
A summary of low-risk meeting notes may need a quick accuracy check. A leadership recommendation, employee communication, assessment, or policy interpretation requires much more scrutiny.
Delegation is not abdication.
What this looks like in learning and development
This decision shows up in L&D every day.
AI can draft the workshop.
I decide what people actually need to learn.
AI can summarize the survey.
I decide what the feedback means and which behavior needs to change.
AI can draft coaching questions.
I know which question to ask because I know the leader, the history, and what has not been said.
AI can generate an activity.
I decide whether it will work for this audience, in this room, with the time and trust available.
AI can adapt content for another audience.
I decide what must become more strategic, more specific, less formal, or completely different.
Those decisions are not leftovers after AI completes the important work.
They are the important work.
A practical way to divide the work
Before handing a task to AI, I work through five questions:
1. What outcome am I responsible for?
Name the result, not the deliverable.
"Create a slide deck" is a deliverable. "Help a new manager handle a difficult feedback conversation" is an outcome.
The outcome determines which decisions require human ownership.
2. Which parts require context, judgment, or a relationship?
Identify the moments where someone must understand the audience, organizational history, risk, politics, emotion, or quality standard.
Keep those decisions close.
3. Which parts need momentum more than expertise?
Look for first drafts, organization, synthesis, transformations, documentation, and option generation.
These are often good places to brief AI.
4. What does AI need in order to be useful?
Give it the audience, context, task, constraints, source material, and required format.
A vague request creates generic work. A strong brief gives your expertise something useful to evaluate.
5. Where must a human review, intervene, or approve?
Decide this before the output arrives.
What must be verified? What would make the work unusable? Who is accountable for the final decision? When should AI not be used at all?
The point is not to create a perfect division of labor on the first attempt. It is to stop handing off tasks without deciding what expertise the work still requires.
What your expertise is for
Most of us are not only short on time.
We are spending our expertise on work that does not actually require it.
Your value was never how fast you could make slides, summarize notes, or rewrite another email.
Your value is knowing:
What matters.
What does not.
What to do next.
That is the opportunity AI gives us.
Not to think less.
To spend more time thinking where it matters.
AI can do the work.
I decide what work is worth doing.
Bring this conversation to your leadership team.
Jennifer Baker speaks about human expertise, AI-supported work, and how learning leaders can redesign work without handing away the judgment that makes it valuable.
Comments