I will try to do this job so well that you won't need me again.
When the people I work with already own the work and the skill to maintain it, teaching them to use AI on their own keeps the improvement going after I leave. When I build a solution, I license it so your team can keep changing and improving it in house, and I teach them how.
From three marketing pitches to an owner who runs his own AI
The owner of a small manufacturer of pressure, temperature, and humidity sensors wanted a better company website. He built the site himself years ago and has maintained it ever since, but he had never used AI tools and was skeptical of them. Three marketing firms reviewed the site and each proposed an ongoing package, with quotes of about $1,800 a month.
I recommended a different path. He already owned the website and the skill to maintain it, so the missing pieces were speed and a second set of eyes. I told him plainly that marketing was new to me, and that we would learn how to do this together.
- A diagnosis he could check. We first asked the AI to review the website and list errors and improvements. He recognized the problems as real, which built trust before we asked anything more of the tool.
- Letting the AI act, with the owner in control. I set him up with Claude Code, which edits the site's files directly on his computer. He kept the step that matters most: he reviews every change and publishes it himself.
- Teaching the context, then stepping back. I gave short guidance on how to give the AI the right context, how to scope a request, and how to check the result. The whole handoff took a couple of hours.
He now improves the website on his own. He has taught two family members to do the same, produced an audit of another company's website, and is planning a redesign of a third. The remaining work involves systems he cannot reach from his desk, such as advertising platforms, so I am building that layer as a system that runs on its own and reports to him.
The Independence Ladder
I generalized that engagement into a six-step method for helping a person or team take ownership of AI tools in their own work. It is a working hypothesis, version 0.1, and I am testing it against each new engagement.
Decide what to own
If the team already owns the work and the skill to maintain it, build their capability. If not, a built system or a service may be the better choice.
Start with a checkable diagnosis
The first use of AI should produce something the person can judge from their own knowledge, which turns skepticism into trust.
Let the AI act
Move the AI from advising to doing, and leave the one risky step, such as publishing or approving, in the person's hands.
Teach, then step back
Teach how to give context, scope a request, and review a result. Independence only shows up once the teacher is absent.
Watch for independence
The clearest signs are new uses nobody prompted and people teaching others. Both can be counted.
Build the hard layer
Once the person is independent on everything within reach, I build the remaining infrastructure as a system that runs on its own.
Teaching AI-assisted qualitative research to healthcare storytellers
I led a one-hour guest session for a special course of four students studying digital storytelling in healthcare. Their work involves collecting many personal stories and finding the themes that run through them, a process researchers call qualitative coding. Done by hand, it limits how many stories a team can study.
The students did not yet have technical access to build software themselves, so I designed the session to leave them ready for the day they would. I covered three things:
- How to set up Claude Code and work with it on a real project.
- How to think like a designer, starting from the people who will use a solution and the decisions it should support.
- The principles of scaling qualitative coding across many stories with AI, while keeping the researcher's judgment in charge of what the themes mean.
The session was arranged informally, which is how a lot of useful work gets started. It shows the same approach I bring to clients: learn the field's own methods first, then teach people to extend them with AI.
Teaching has been part of my whole life.
I learned to teach one person at a time, long before I taught with AI. That habit of meeting each learner where they are shapes how I work with clients.
- Served as a full-time missionary, where nearly everything I did was teaching people one by one.
- Teach Sunday school in my congregation.
- Tutored math for struggling students while I was in high school.
- Co-taught Probability and Statistics for Engineers at the University of Louisville, tutored Fundamentals of Optimization, guest lectured in Advanced Predictive Analytics, and led weekly peer-learning sessions on statistical reasoning.
- At the CDC, built an AI solution for reviewing public health reports and taught nontechnical colleagues what it did and why it worked the way it did.
- Trained research team members at Houston Methodist on context engineering for AI tools.
Want your team to run its own AI tools?
Tell me what your team already owns and where it gets stuck. I will suggest a first task the team can check for itself, and a path to doing the rest without me.