7-40 Challenge | Planning Week
I work in data. A big part of my job involves tracing data pipelines, reading SQL, debugging problems, and figuring out where things break. I’ve been doing it for almost twenty years.
Recently I’ve been using AI — specifically Claude — as a working partner in that process. And it’s changed my productivity in ways I didn’t expect.
Here’s what AI does well. I can hand it a SQL statement and it will walk through it step by step, telling me exactly what it’s doing. For someone who already reads and writes SQL, this is incredible — I can cut through code at speed and get instant answers on things that used to take research and thinking time. Questions that would have meant twenty minutes of digging through documentation, I get answered in seconds. As I trace a pipeline from start to finish, I can see how each step is built and how real data moves through it.
It is, without question, the most powerful tool I’ve ever used at work.
Here’s what it can’t do.
It doesn’t know what the business meant.
I was debugging a problem recently. The AI looked at the data and told me everything was correct. It wasn’t. I had to go back and define what the data fields actually meant — what they represented in the real business process, not just what the code said they were called. Once I did, the AI admitted it had made a logic jump. It had assumed it understood the data because it could read the code. But reading the code and understanding the business are two different things.
That’s the gap. SQL is a language, just like English. AI can process the language. It cannot supply the intent. It doesn’t know what a data point is supposed to represent. It doesn’t know the standards. It doesn’t know the stage gates. It doesn’t know the rules that exist because a human made a decision five years ago that still matters today.
So I have to define everything — every data point, every standard, every rule — for the AI to have any framework for how it’s supposed to operate.
Without that, it’s just a machine sitting by itself, not knowing what it’s doing.
Data exists because people made decisions. Pipelines exist because someone had an intent. The code is just the execution of something a human needed to happen. Without the person who understands why the data is there, the tool has nothing to work with.
AI processes the language. It does not supply the intent.
Humans don’t like to follow rules. Computers do. We innovate outside the rules.
The tool isn’t replacing me. It’s making me faster at the work that still requires me to show up — the definitions, the context, the judgment calls. That’s not going away. If anything, it matters more now than it did before I had a tool this powerful sitting next to me.
David Willis 