How do you handle unknown risks during project execution?

In today world where the generative AI exists is difficult to think about the existence of “unknown unknowns” risks. I am not writting this "by the book". I am writting this because my personal experience. Just to remember generative AI exists from long time ago. Just today we are using a new model, no more than that. On the other side, all you need to do is to put a real project risk management process which is defined from the organization risk managemenet process and it is a continuous activity along the initiative.

I handle unknown risks by encouraging early reporting, continuous monitoring, and quick impact assessment. When they arise, I respond fast, analyzing impact, adapting plans, and using contingency reserves wisely. I maintain stakeholder trust through transparency and by always presenting clear solutions, not just problems.

You don’t really “manage” unknown risks the same way you manage known ones.

You design the system to surface them early and absorb them when they appear.

Because by definition, you can’t plan for what you don’t know.

In my experience, this comes down to two things: detection and adaptability.

Detection is about creating enough visibility into the system that weak signals show up before they become real issues.

Unexpected dependency friction
Unusual delays in decision-making
Teams working around the process instead of through it

Those are often early indicators of risks that haven’t been formally identified yet.

The second piece is adaptability.

Unknown risks don’t follow your risk register or mitigation plans.

So the question becomes:

How quickly can you understand, decide, and respond when something new emerges?

That’s where things like clear decision ownership, strong operating cadence, and access to contingency (time, budget, capacity) actually matter.

I also think there’s an important distinction between contingency and management reserve.

Contingency helps you handle known uncertainty.

Management reserve exists for exactly this—unknown unknowns.

But even that’s not enough if decision-making is slow or unclear.

So I don’t try to predict unknown risks.

I focus on reducing the time between:

signal → understanding → decision → action

That’s what keeps them from becoming bigger problems.

Not the plan.

The system.