AI Discipline: Having the Tools Isn't the Same as Using Them Well

A couple of weeks ago I wrote about AI eating the SaaS industry. The short version: if anyone can build exactly what they need, the per-seat subscription model is in trouble. A lot of people agreed. A few pushed back. That is fine. It is a real conversation worth having.

But here is the thing I have been sitting with this week. Having the tools is not the same as using them well.

I have been in a lot of conversations lately about AI usage, and the pattern I keep seeing is this: teams get access, teams start experimenting, token bills arrive, and then someone in finance asks a very reasonable question about what exactly they are getting for all of this. And the honest answer, more often than not, is "we are not totally sure yet." That is not a technology problem. That is a habits problem.

The questions that actually matter right now are not "should we use AI?" We are past that. The questions are: are you using the right model for the job, or just the most impressive one? Are you reviewing what the AI produces before it moves forward in your workflow, or just approving it because it looks plausible? Are your agentic pipelines doing one focused thing at a time, or are they sprawling tasks that are hard to debug when something goes wrong?

I do not know if this is true for every industry, but I know it is true for software teams: we are very good at adopting new tools and very slow at building discipline around them. We did it with version control, we did it with CI/CD, and we are doing it again now. The teams that figure out the discipline part first are going to have a real advantage. Not because they are using AI more, but because they are using it better.

Originally posted on LinkedIn: "AI Discipline: Having the Tools Isn't the Same as Using Them Well"