A feature is split across tickets and agent sessions. Someone has to keep reconnecting the pieces.
The project keeps its direction while connected tasks move toward a reviewable result.
Begin with a bounded product improvement
A customer export improvement is a useful example: the problem is specific, the current behavior can be inspected, and the result can be tested. Capture the customer need and acceptance criteria in the project.
Avoid using “improve the product” as the first assignment. A bounded task makes it easier to learn what the agent can deliver and where the brief or operating rules need work.
Make investigation useful to implementation
Ask the first task to identify the existing behavior, relevant code, and constraints. Pass that completed output into the implementation task through a dependency, then connect a documentation update.
Each role receives the project context and relevant configured skills. Keep credentials scoped to the agent’s responsibility and use your autonomy policy to define how independently work can proceed.
Measure delivered work, not agent activity
Review accepted changes, correction time, and the effort spent coordinating the project. Compare them with the workflow you used before. Run count is not a useful success metric on its own.
Norc makes the coordination and evidence easier to inspect. Your team still decides what customers need, whether the result meets that need, and when a change is ready to ship.
Practical questions.
Is Norc useful for a team with one agent?
It can be, if that agent works across a queue of tasks and you need shared context, run visibility, or project controls. Multiple agents are not a prerequisite.
What should our first workflow be?
Choose a contained improvement with a known reviewer and observable acceptance checks. Investigation, a small implementation, and documentation make a useful sequence.
By the Norc team · Product details checked against the Norc project documentation. View the project