!
AI isn't structurally part of today's workflow. Individual contributors may reach for a chat tool now and then, but there's no shared discipline, no prompt library, no quality contract — and no role that depends on AI output to do its job. Every activity in this view is human-led by default.
AI is applied deliberately, not opportunistically. Three modes — agent-led, agent-assisted, human-led — are picked per activity. Spec quality is the primary AI quality lever. Human accountability sits at the gates that matter.
AI modes
Agent-led — AI drafts/executes, human directs & reviews
Agent-assisted — human leads, AI augments
Human-led — judgment, accountability, sign-off
Gates & markers
Hard gate — Trio or named owner co-signs
Soft checkpoint — ratify, not re-decide
Activity treatment
Parallel — runs alongside Build, not after
Continuous — runs throughout, not phase-bound
The takeaway when you toggle

Today's process is sequential, hand-off heavy, and AI-absent — every box is a person doing the work alone. Tomorrow compresses the phases, moves the Trio into co-authorship at G2, applies AI deliberately by activity, and runs supporting work in parallel rather than after.

What you'll notice

Tech writing and Product Marketing move into Build rather than queuing behind it. QA tests are scaffolded from the spec at G2, not authored mid-Test. Customer Success and DS Team work continuously — they don't start and stop with the cycle.