ADRs with the debate attached.

Write the architecture decision record as a living document: the context, the drivers, the options considered, the objections raised in review and who agreed, so the next engineer knows why the system looks the way it does.

How it fits

Context and problem

State the problem and the constraints, and link the tickets, RFCs and incidents behind it.

Decision drivers

Name what the design must deliver: performance, cost, operability, security, team skills, lock-in.

Considered options

Lay out each candidate, including doing nothing; reviewers and AI probe assumptions and failure modes.

Decision outcome

Record the chosen option and who agreed, then lock the decision.

Consequences

Write down what gets easier and what gets harder. A later ADR supersedes this one explicitly instead of quietly drifting from it.

Questions

Does this replace the ADRs in our repository?

Not as files. Keep the index in the repository and link each ADR to its prorso record, where the options, objections and approvals live.

Is it heavyweight?

A decision starts from one sentence. Add drivers and options as the discussion needs them; nothing is required up front.

Who accepts the decision?

The owner names the approvers. When every approver has approved, the decision closes, and locking it stops the record from changing.

What do the AI agents do?

They read the decision and point out unstated assumptions, missing options and claims without evidence. They suggest; people decide.

Explore example decisions · Read articles

prorso for other teams: Procurement · Product management · Finance & investment · People & HR · Operations & supply chain · Projects & PMO · IT, security & risk · Strategy & governance