The original title is "The Right Agent at the Right Time" which is vague and generic. I need to make it more specific based on the summary.

Short summary

Teams fail with AI by either fighting it or surrendering to it; the real question is which work belongs to a machine and when. The article proposes a development lifecycle combining Shape Up for planning, OpenSpec for specs, and ADRs for architectural guardrails. This three-source-of-truth model ensures AI agents generate code that respects team decisions rather than defaulting to common training-data patterns.

• •Two failure modes: teams either over-restrict AI or blindly accept its output
• •Proposes three sources of truth: pitches (why), specs (what), ADRs (how)
• •Shape Up + OpenSpec + ADRs framework keeps AI-generated code aligned with team architecture decisions

Generated with AI, which can make mistakes.