How I lead engineering teams

I want teams to make good decisions without waiting for me.

Earlier in my career, I sometimes mistook being involved in every important decision for being useful. That works for a small team. As the team grows, it becomes a bottleneck.

A large part of my job now is making sure people understand the problem, have enough context and know which decisions are theirs to make. I’m direct about expectations and I’d rather discuss a difficult trade-off early than discover it at the end of a project.

Agnieszka Besz smiling on stage with a microphone at the Decentraland Community Summit, in front of a neon Decentraland logo
Presenting the new Decentraland client at the Community Summit, Buenos Aires, 2024. More talks

How I work

01

Start with the result

Shipping is important, but finishing a roadmap is not enough. I want to know what improved for users, creators or the business.

02

Say the difficult thing early

Unclear expectations rarely become clearer on their own. I try to name the problem and discuss the trade-off while there is still time to act.

03

Add process for a reason

I add a process when a recurring failure shows that we need one. A new meeting does not fix unclear ownership.

04

Stay close without taking over

I stay involved in technical direction, but the people closest to the system should usually choose how to build it.

When the number becomes the goal

A proxy is an indirect measure we use when the outcome is slow or hard to see. It can be a useful KPI, but it is not the outcome. Fewer decisions escalated to a senior leader may mean stronger ownership. It may also mean people have stopped raising risks. The number tells me where to look; it does not tell me which explanation is true.

When escalations drop, ask why and check other evidence. It may mean stronger ownership, so keep the KPI, or risks being kept quiet, so add context or replace it.
I would read the escalation count alongside release quality, decision speed, team load and examples of decisions the team handled well or badly.

If people can improve the number without improving the result, the KPI needs more context or should be replaced.

Questions I come back to

Growing the engineering organisation

One team I led had grown faster than its way of working. Too many decisions returned to the same senior people, managers lacked a clear remit and QA was often involved late. I had contributed to that dependency by staying too close to too many decisions.

We clarified ownership, gave managers and QA leads more authority and made release quality easier to see. The change took time. It involved coaching, clearer expectations and occasionally letting someone make a decision differently from how I would have made it.

Shared objectives, explicit decision boundaries and visible release evidence move normal decisions from a senior approval queue to the teams closest to the work.
Senior leaders stayed available for unusual or high-risk calls; they stopped being the default route. The full case study

Changes I have led

Things I try to avoid