Finally, a yes — and now what?

I had tried many times before to gain interest around initiatives on AI and, although I generally got positive responses, I never really saw any interest in going forward.

When I finally got the go-ahead it did not come from the department where I normally work but from another one and with an unclear project scope.

I thought I would feel thrilled to get working with someone actually supporting the idea after so many attempts, and in some way I did.

However, there were also some aspects that made that feeling ambiguous.

The message there and then was that I was allowed to spend some money on it along with moral support. At this point I didn’t know the frame or the timeline for the project. I also had no dedicated work time set aside for it.

The project, curiously named Buildmate, is a service for using natural language to build reports. Normally, this would require a higher level of skill and experience in using the internal report editor.

My first instinct was to set my own direction and put frames on my own work. I knew that, if I couldn’t present any numbers or, at least, a precise working use case, my mandate to work on Buildmate would run dry.

Regrettably, I found my ability to frame my own work was mediocre. I kept opening new features, working on evals and metrics for different models, studying various subjects that were not always relevant to the phase I tried to prioritize.

As an example, I spent time optimizing the workflow process making it faster and less expensive to run. It was a genuine improvement to the targets I set but when I presented the results at the weekly briefing it became clear that this was not a priority.

There was nothing obviously wrong with this. However, being honest, I saw that I would need some structural guidance in the project framing or I would drift away on an unclear path.