Krishna Dey Note
Three levels of execution risk
Delivery slips in the work, in the dependency graph, or in late decisions. Teams usually only look at the first.
When a date moves, I ask which layer actually failed:
- Task risk — the work is larger or fuzzier than the estimate. Specs were thin. This is the layer everyone can see.
- Coordination risk — the dependency graph is denser than the plan assumed. Integrations work lives here: Salesforce objects, vendor translations, another team’s API.
- Decision risk — the team cannot move because a choice is late or ownerless. Architecture, scope, or “are we building the generic version?”
Legal Intake was mostly (1) and (3): high complexity, real delay risk, and a temptation to start from a semi-spec. We made written specs mandatory before estimation, split the work into phases, and used velocity including leave to challenge underestimates. We found a two-week slip early and reset the plan instead of crunching.
Most teams only manage task risk because it is visible in tickets. The other two are where integrations and platform work actually fail.