Stop rolling it out to everyone on Day 1. Start with the person who already uses it well.
Most companies announce AI, buy licenses, run a training, and wait. Six months later a few people are getting real value and everyone else went back to what they were doing. The training is not the problem. The direction is.
You already have someone who figured it out. They are answering everyone's questions right now, which feels like adoption and is actually a bottleneck. Every answer they give is one person helped, once.
Have them build instead.
One client learned AI fluency from me, and we built a decision tool for her own work. Five versions in, it left her desk. Tens of thousands of people across her organization use it now. One pilot, one workflow, and people across the company make decisions differently. And it got far more people interested in learning to use AI.
If you want to run this properly, the AI Fluency Sprint for Teams is six weeks of exactly this. If you want to try the thinking first, try one of the Workshops.
Then you find out whether they were a bottleneck or a catalyst.
If they answered questions, the knowledge walks out with them. If they built things, the things stay.
This is the strongest argument for doing it the second way, and it is the one owners feel most sharply. You do not want the organization resting on one person knowing everything.
This is why the mapping work matters. What breaks when someone leaves is a question you can answer on purpose, before it happens, rather than discovering it in their last week.
They are right, and that is the problem with training as a category.
Nobody has a spare afternoon. So do not ask for one. Pick work they already owe someone this week and do that work with AI, together, once. The learning is a side effect.
That is why the workshops ask you to bring one real piece of work. You leave with the thing done and the method learned. Not notes.
Three signs it's working:
Same work, less time
Team asks to use it more
You can measure the savings
Three signs you're overcomplicating:
Managing the tool takes longer than doing the work
Team avoids it
Can't point to specific time saved
Start with one repetitive task. Measure before and after.
If one tool saves 30 minutes daily per person, that is 25 hours/day across 50 people.
If you're three months in and can't name three faster tasks? You've added complexity.
Scale back. Pick one thing. Make it work.
To give people a sense of calm and agency during a time of great change.
The disorientation most people feel right now is not a mindset problem. It is an accurate reading of what is happening. The tools change weekly, the advice contradicts itself, and everyone seems to be further along than you are.
What helps is thinking more clearly and seeing a little earlier. Better decisions follow from that. So does the calm.
Calm is not being told it will be fine. It is knowing you will see the thing coming in time to do something about it.
Most of what I teach is one thing in different forms: getting what someone knows out of their head and into something other people can use.
That takes a few shapes.
Setup. A context file, so AI stops guessing about your business. Guardrails, including the rule you are most likely to break. A rhythm for what to re-run and when, because this is a practice rather than a tool you visit.
Thinking instruments. A premortem before you commit. A root cause analysis you can actually finish. An advisory council that challenges you rather than agrees with you. These were always good and always too slow to run. That is why most companies do them once a year, if ever.
Mapping. What each person does and where the time goes. Who depends on whose work. Which numbers matter and who owns each one. What breaks when you grow, and what breaks when someone leaves.
Reading change. Position, velocity, acceleration, jerk. All by subtraction. Choosing the few signals worth watching, and knowing when an instrument is lying to you.
Handing off. Building a tool someone else can run. Writing down what was only in one person's head. Pages that answer the questions your buyers are actually asking.
None of this is new. What changed is the cost. Externalizing what you know used to be the expensive part, and it is not anymore.
Knowing what to hand over to AI or not, how to brief it, and how to check what comes back.
Most training is prompts. Prompts are the smallest part. Fluency is understanding that you are working with something capable and literal, that it knows nothing about your business until you tell it, and that the quality of what you get back is mostly a function of what you gave it.
Being the answer when your customer asks a machine instead of searching.
Ask an AI to recommend a business like yours. Then ask again tomorrow. Different answer. It is a roll of the dice, not a ranking, and most businesses have never checked what comes back.
The work is figuring out the questions your customers are actually asking, then making sure your pages answer them. Not keywords. Questions.
During an AEO workshop, Nancy Case put it better than I can: "Most taffy people have a favorite flavor, so instead of searching for taffy they search for banana taffy or black licorice taffy. Once you see that, you know which pages you need to write."
A way of reading change early.
Position is where you are. Velocity is how fast it is moving. Acceleration is whether that speed is changing. Jerk is the change in acceleration, and it is the earliest warning your numbers can give you.
All of it is subtraction. This period minus last period. You do not need calculus and you do not need private data.
I write The JERK Report every Monday applying it to whatever is moving.
Depends on how much time you have.
Two hours, want to see if this is for you: a single workshop. AI fluency, AEO, or JERK.
A week, want the whole picture: Workshop Week. Everything, in the order it makes sense.
Already fluent, want to keep getting better: the monthly catch-up.
It is your team that needs this, not just you: the AI Fluency Sprint for Teams. Six weeks.
Most people start with a workshop or workshop week as a sampler.
Yes.
You already know your business. What takes too long. What is repetitive. What your team struggles with. That knowledge matters more than technical skill, and it is not transferable, which is why you cannot hire it out.
The skill is recognizing the pattern. We write twenty similar proposals a month, could this draft the first version. That is a business question, not a technical one.
The tools will change. What you learn will not.
Permanent: spotting repetitive work, briefing well, evaluating output, knowing what to hand over.
Temporary: the specific features of today's tools.
Everything I teach is tool agnostic and works from first principles, which is partly so you chase fewer tools you do not need.
Letting AI stay with one person. The person who figured it out spends their day answering everyone else's questions. That feels like adoption and it is a bottleneck. Have them build something instead.
Going all in. Everyone, everything, at once. The result is an overwhelmed team and no clear result. Pick one process. Make it work.
Buying tools before you have a process. Most companies buy five hoping one sticks. Start with one and find out what is still slow.
Assuming AI knows your business. It does not know your numbers, your customers, or your constraints, so it invents them, and it sounds right doing it. Most people try to fix this with better prompts. Better context is what fixes it.
Training everyone and calling it adoption. The training happens and the behavior does not change. One team, one task, one thing they built and use.
Starting with your most important work. Mediocre output on high-stakes work sets you back further than doing nothing. Start where the cost of a bad draft is low.
Measuring output instead of decisions. Ask someone what they used AI for last week and you will hear about a faster draft. Ask what decision they made better and you may get a blank. The first kind of gain is real and everyone gets it. The second is where the impact and lasting change happen.
Watching more instead of watching earlier. One person left a workshop, built a matrix of twenty-two metrics in two days, and immediately said the true thing: I do not need twenty-two metrics. The discipline is watching less, earlier.
Waiting for AI to settle. AI will not settle. Waiting means closing a gap in two years that did not have to open.
The first month can feel slower, not faster. You are learning. Budget for that.
Output needs review. Someone has to know enough to catch what is wrong, which is an argument for fluency rather than against AI.
The tools keep changing, so this is ongoing rather than a one-time setup.
Sensitive client data should not go into public tools without a clear policy. Most small companies do not have one.
And some people will resist, and some of that resistance is fear about being replaced. That is a real conversation, not a change-management problem to be routed around.
None of this is a reason not to do it. It is what doing it honestly looks like.
Look for implementation experience, not theory. And someone who understands your industry.
Red flags:
Leads with technology, not your problems
Promises "transformation" without asking about your processes
Wants multi-month engagement before you see results
Green flags:
Asks about operational bottlenecks first
Names specific examples from similar businesses
Starts with focused pilot, not massive transformation
Teaches you to fish
For growing firms: avoid enterprise consultants. You need practical AI, not complex infrastructure.
Ask: "What's the smallest, fastest win you can help us achieve?"
If they can't answer specifically, keep looking.
Six Sigma consultant: Cut project analysis time from 5 hours to 1 hour using AI. Doubled her client load without adding team members.
Management team: Used AI for root cause analysis. Cut analysis time in half, freeing the team to focus on implementation instead of data gathering.
Pattern: They didn't automate everything. Picked one or two high-impact tasks. Measured results. Expanded from there.
No technical expertise required. No massive budgets.
Clear thinking about where AI eliminates repetitive work.
Time investment upfront. Learning curve. Risk of over-reliance on changing tools.
Drawback 1: First 30-60 days might feel slower, not faster. You're learning. Budget for this.
Drawback 2: AI output needs human review. Can't blindly trust it. Someone reviews and refines.
Drawback 3: Tools change constantly. Ongoing learning, not one-time setup. Skills transfer. Tools don't.
Drawback 4: Privacy concerns. Sensitive client data shouldn't go into public tools. Need clear policies.
Drawback 5: Team resistance. Some will resist. Some worry about being replaced. Change management is part of the work.
Firms that succeed acknowledge these and plan for them.
Not a magic bullet. A tool that works when applied thoughtfully.