HOW I USE AI

Why I'm sharing this

AI is part of how I work, and I'd rather be clear about how I use it and not leave anyone guessing. I also spend my working life on the future of learning and work, so how these tools get used, and misused, is not an abstract question for me. This is a practical account of how I use AI now: what it does for me, where I stop, and what I continue to own no matter how good the tools get. And yes, I wrote it with AI's help, following the guidelines below.

Core principle

I use AI to extend what I can do, not to outsource my judgment or my accountability. It helps me explore, understand unfamiliar material, find gaps, and sharpen things I've already made. I decide what matters, I check what's worth checking, and I own the result.

Where I use it

Most often, I take what I think I know and throw it into AI, asking it to challenge my assumptions, lay out counterclaims, and push me to sharpen my arguments.

Another use case is helping me organize and summarize the incredible amount of material being produced about work and learning. I'm building a tool to map this field: tracking how the conversation around growth, learning, and the future of work is actually moving, so I'm reading the shape of a market instead of a few loud voices. AI lets me cover more of that landscape than I could by hand and pull the signal out.

The rest is less exciting and just as useful: arguing the other side of my own recommendation before I commit to it, compressing a pile of market and customer inputs down to what’s meaningful in potentially shifting a decision, cleaning up a dataset, tightening a deck once the thinking is done.

I keep the thinking part.

What I check, and what I won't fake

AI is a good way to get oriented and a bad way to “be sure”. I am very aware that it could hand me a confident, invented number, a market size, or a competitor "fact," or a partner detail that sounds exactly right and isn't, and then might make its way into the C-suite or Board. So for anything that carries executive weight, I go to real sources and use my own read. The higher the stakes, the less I trust an unchecked answer. I don't pass off invented facts, quotes, experiences, or sources as real.

When I tell you

Not every use needs a label. Asking for a cleaner sentence is not the same as generating most of a finished piece. I'll flag AI's involvement when it's substantial, when someone asks, when the context calls for it, or when knowing would change how you'd read the work. If a project sets its own rules, I follow them.

What I don't put into these tools

There's a short list of things I won't put into an LLM. In my work, that's customer data; anything a customer told me in confidence; partnership terms; anything under NDA; and unreleased GTM or pricing. Credentials, personnel matters, and legally protected material sit in the same bucket. Unless a tool is approved for that use and I'm cleared to use it that way, it doesn't go in.

When an employer or client has their own tool and policy, theirs wins over mine. Where there's no policy, I assume anything I type could be read by someone else one day.

What stays mine

Some things I won't hand to a model even when it could produce a passable output: final calls, ethical judgments, the delicate human conversations. AI can help me think through sensitive conversations and tricky situations, but it can't be accountable for what happens next. That part stays with me.

This will keep changing

The tools will change and so will what I think about them, so these notes will change too. If you ever want to know how AI figured into something I made, ask me. I'll give you a straight answer about that piece.