Detours · No. 4

Teaching yourself the next craft

I spent most of my career in sales and HR before I taught myself to build with AI, and this is what I'd tell anyone learning a new craft on their own.

Is it too late to learn a whole new craft once you’re well into a career? My answer is no, and you may be better placed than you think.

Most of my career was in sales and then on the people side of companies, from Sepracor to Roadr. The tools changed along the way, from Excel and SQL at HBO to Python and SQL at Roadr.

Then I taught myself to build with AI.

Plenty of people are learning alone

If you’re figuring out AI at work without much help, a lot of people are in the same spot. ZipRecruiter’s 2026 AI Worker Report found that 37.2% of workers learn AI tools entirely on their own because their employer provides no training.

So the self-taught route is common. For a lot of workers, it’s the only route on offer. Not everyone is sure it’s enough, though. The same report found only 34.3% of workers believe they can master AI skills without help.

Comfort with these tools isn’t widespread yet, either. In an Elon University survey of employed adults, 25% said they were already comfortable using AI at work, and 41% said they would be open to it with more training.

My read: good training helps, but you don’t have to wait for it to start.

Start with a problem you already know

Years of recruiting showed me why strong candidates never hear back. That kind of knowledge comes from doing the work, and in my view it’s the harder half of building something useful.

That’s the advantage of learning a craft mid-career. Someone starting out has to learn the tool and find a problem worth solving. You may only have to do the first part. Years in a field give you a list of things that should work better, and that list is a head start.

If you want to try this week, here’s how I’d begin:

  1. Pick one small, annoying problem from your own work that you understand better than most people.
  2. Use an AI tool to build the smallest version that helps, even if it’s a better spreadsheet or a short script.
  3. Show it to one person who has the same problem and ask what’s missing.

It will be rough at first, and at times discouraging. My view is that a nonlinear path is normal, and learning a new craft partway through a career is one more turn in it.

From Dabbling to Going All-In

I didn’t plan this pivot. I dipped a toe in, dabbled a little, and before long I was all in, excited to learn more and build better. On ReflectivAI, I learned as I went and kept reworking it as AI changed. CareerPaver and Lumira followed. One is for sales calls, one for job interviews, and one for any two people who matter to each other. Different people, same frustration.

One idea runs through everything I build: people get better at hard conversations by practicing them. Most of us know how to research. We read one more article, prep one more answer, and tell ourselves we need a little more time. What’s missing is the nudge that says you’ve researched enough, now go practice.

I built all three to give that nudge: a private place to rehearse the questions and the pushback, with coaching grounded in real science that fits your situation, and feedback without judgment. All of it happens before you’re in front of a client, an interviewer or a partner, so you walk in ready. It’s what I wish I’d had in a lot of the roles I’ve held.

The turns in my own path made me more confident walking into hard conversations. They rarely go exactly as rehearsed. Practice is what lets me adapt when they don’t.

Which conversation have you prepared for long enough to start practicing out loud this week?

Sources