What nobody tells you about leading AI experiments

There is a part of this work I had not named in the previous articles. It is not technical. It does not show up in a flow diagram or a token schema. But it is present every day of the project: the feeling of moving forward without knowing exactly where to.

I am living it right now. In the Harness Design System project and in the application of AI upstream, we are making progress. There are concrete results: the token governance I described in the previous article is real, it works, and the team can operate it. But real progress coexists with an unpredictability that does not decrease as the project moves forward. It only changes shape.

The problem with an experiment is that it is an experiment

When you are building something that has been built before, the scope is relatively stable. You may not know how long it will take, but you know what you will find along the way. Someone has already mapped the terrain.

When you are testing something that nobody in your context has done before, the scope is a hypothesis. Each week reveals something that changes what you thought you knew. In the article about tokens, I wrote that the estimate was to finish in one week, and it did not. That was not a planning failure. It was the project showing its real complexity as we got into it. Typography got stuck in a way nobody had predicted. The bind required a path that was not in the original plan. The question of where the JSON should live turned into an infrastructure discussion that really took time.

This is what an experiment does. It reveals what you did not know you did not know.

The problem is that teams, companies, and stakeholders are usually not calibrated for this kind of work. There is an implicit expectation, sometimes an explicit one, that you should know what will happen before you act. And when you do not know, it looks like a lack of preparation.

What this does to whoever leads

There is a human cost to this that is worth naming, because it is real and rarely talked about.

Leading an AI experiment in a company that depends on predictable results creates a permanent tension. On one side, the project demands attention to what is happening now, to course corrections, to the discovery that came in on Wednesday and changed what you were going to present on Thursday. On the other side, there is the legitimate pressure to report progress, defend priorities, and justify the time allocated.

This tension produces anxiety. And anxiety produces a specific temptation: promising more than you know, to look like you are in control. Anticipating results that do not yet exist. Presenting the plan as more solid than it is, so you do not look like you are improvising.

I understand this temptation. I have felt it. And I learned that it is the shortest path to a bigger problem.

The only way out I found

Transparency is not a communication strategy. It is the only posture that works when you are in genuinely new territory.

In practice, this means bringing leaders into the conversation before having answers, not after. Not arriving with the result, but with the map of what is being discovered. Saying what worked and what did not with the same weight. Protecting the team from the pressure to deliver certainties the project does not yet allow.

This takes a specific kind of courage: not the courage to move forward despite the risk, but the courage to be vulnerable in front of people who expect you to know what you are doing. To say "I do not know yet" to people who have a legitimate interest in knowing.

In our project, what has worked is keeping stakeholders in the loop continuously, not only at the milestones. When the scope changes, and it will change, they already have enough context to understand why. When a result is not what was expected, it arrives as learning within a conversation that was already happening, not as a surprise in a presentation.

This does not eliminate the tension. But it stops being about credibility and becomes about the real problem: how to move forward in terrain nobody has mapped yet.

What remains

I do not have a ready method for managing unpredictability in AI projects. I do not think anyone has one yet, at least not one that works in every context.

What I have is a posture: protecting the team from promising what cannot be promised, keeping leaders inside the reality of what is happening, and treating every discovery, including the ones that complicate the path, as valid information, not as a failure to hide.

Experiments are like this. The result is not only what works at the end. It is everything that was learned along the way.