How engineering leaders build team judgment through uncertainty.
An engineering team is discussing how AI should change its work. The conversation reaches a familiar question: what should we do next?
As the team lead, you are expected to provide direction. You do not have a reliable answer.
Throughout my work in software engineering, I have often turned to Stack Overflow and colleagues when I did not know something. Finding out has always been part of the work. Yet leadership can bring a different pressure: other people are looking to us, and admitting uncertainty can feel like failing them.
Leading through uncertainty means taking responsibility for how the team learns enough to decide, acts within its limits, and uses what happens to improve its judgment.
Brené Brown’s Dare to Lead discussion examines how the need to be seen as knowledgeable can make asking for help feel like weakness.
Saying “I don’t know” makes a limit visible. What follows determines whether the team has a way forward.
Understand what kind of answer is missing
A framework I have used for this is Cynefin. It helps people make sense of a situation and choose an appropriate response. One distinction is especially useful here: complicated and complex.
| Situation | What makes it uncertain | Useful response |
|---|---|---|
| Complicated | Understanding cause and effect requires expertise and analysis. | Bring in relevant expertise, investigate, and decide. |
| Complex | The effects of interacting people and conditions cannot be reliably established in advance. | Run small, contained experiments, observe emerging patterns, and adapt. |
These are two contexts within the broader framework. The distinction concerns how the situation works, beyond how difficult it feels. Dave Snowden’s guidance on responding to different contexts explains why analysis and experimentation serve different purposes.
An unfamiliar technical detail may require documentation or a specialist. Understanding how a change will affect the way people work together requires observing their responses and interactions.
That also means “AI” is too broad to put in one box. A tool’s documented capabilities can be checked. Whether using it improves a particular team’s work requires evidence from that setting.
The leadership question becomes more precise: what would help us understand this situation well enough to choose a next step?
Leading through uncertainty in practice
A team does not need a complete prediction of AI’s future to investigate a useful question about its current work.
Consider a team exploring AI to help write automated tests, the checks used to detect problems in software. A practical question would be whether the assistance reduces effort while preserving the usefulness and reliability of those checks.
The team could try an approved tool on a few limited tasks, keep human review, and compare different approaches with its existing practice. Are the checks useful and correct? Does any time saved survive review and correction? What unexpected difficulties appear?
For complex situations, Cynefin’s approach involves several small experiments in parallel, allowing different possibilities to be explored as patterns emerge. The experiments should be safe to fail: their downside is contained, their effects can be observed, and the team is ready to change or stop them. Snowden’s practical guidance develops this further.
The practical responsibility includes deciding what can be tried, what must be protected, and what would justify stopping.
A pause can be responsible too. Missing authorization, capacity, or essential information may prevent a useful next step. Name the constraint, involve whoever can address it, and establish when the decision will be revisited.
Ownership includes asking for support and making limits visible. It does not give us control over every condition.
Committing broadly before understanding the effects risks spending time on changes the team later has to undo. Waiting without a plan to learn leaves the question unresolved. A focused investigation or limited experiment helps the team gather evidence for its next decision.
Make the learning useful to the team
If the leader does all the investigating privately and returns with a conclusion, the team has had little opportunity to develop its own judgment.
Involve the people who will use the approach, review its output, and deal with its consequences. They can help identify what matters and notice effects the leader would miss.
Carol Dweck’s clarification of growth mindset emphasizes changing strategies and seeking input when learning stalls. In a team, that means being willing to revise the approach, including one the leader originally favored.
For one unresolved question, leave the discussion with four things clear:
- Decision: What do we need to decide, and why does it matter?
- Uncertainty: What do we already know, and which missing information could change our choice?
- Inquiry: Who could help, what could we verify, or what limited experiments would reveal more?
- Commitment: Who owns the next step, within what limits, and when will we review what happened?
At the review, examine unexpected effects as well as the result you hoped for. Decide what to continue, change, stop, or investigate further. The owner remains accountable for moving the decision forward; involving others should make that responsibility clearer.
This is where the connection to high performance becomes practical. Adaptability requires responding to new evidence. Future capability grows through opportunities to practice judgment, compare interpretations, and learn from consequences.
A useful sign of progress is hearing people explain what they observed and how it informs their recommendation.
What your response teaches
I also run out of answers after achieving a meaningful goal. There is joy, then gratitude, and eventually the question: what next? Uncertainty returns even after something has gone well.
For a team, a successful decision provides experience to draw on. The next situation still deserves attention. What worked before can inform our judgment without settling the next question.
When people look to you for certainty you cannot offer, you can still help them find a reliable source, shape a careful experiment, or name a constraint that needs a decision. They can participate in that work and learn how to approach the next uncertainty with greater independence.
The answer may still be missing. The team can leave knowing how it will move the decision forward.