---
name: scale-evolve
description: S.C.A.L.E. step 5, Evolve. Scores the pilot against the charter (expand, adjust, or stop), builds the three-layer scorecard and operating rhythm, drafts the Center of Excellence with two co-chairs and named roles, and writes the AI policy sized to the company. Use when the user says "evolve the system", "step 5", "did the pilot work", "set the cadence", "who should own AI", "AI policy", "center of excellence", or after step 4.
---

# scale-evolve

Write `SCALE/5-evolve-system.md`. Don't scale chaos. Scale what works, with governance baked in.
## Preflight (every run)
1. Find the `SCALE/` folder (Desktop or Documents), or `GTM-AI-OS/artifacts/scale/` if the user runs the GTM AI OS. If neither exists, run `scale-start` first; it creates the folder and asks for access.
2. Read `SCALE/progress.md`: mode (quick or deep), AI experience level, which steps are done, the one-line context.
3. Talk short and direct. Say what each answer is for. Beginner: one question at a time, plain words. Advanced: batch questions, skip explanations.
4. Quick mode: ask only the questions marked (Q). Deep mode: ask all. Accept "skip" and "don't know"; record them as gaps.
5. If the previous step's artifact is missing, say so and offer: do that step now, or answer three questions and continue with a draft marked fast path.
6. When the user names a system (CRM, calls, email, spreadsheet), request access to that one or take an upload. Never assume a product.
7. On finish: write the artifact, update `progress.md`, say the next step in one sentence, and ask one open question if any remain.


## Part A: Judge the pilot
1. (Q) Each criterion: baseline, current, target. From the CRM if connected, else from the user.
2. Stage diagnosis: early-stage drop means ICP or signal (back to Chart); late-stage drop means motion or skill (Align); post-sale drop means product or onboarding.
3. The two headline metrics, restated with their numbers.
4. (Q) The call: Expand (to a defined next group, never the whole org), Adjust (what, and re-run the window), Stop (and what was learned).

## Part B: Scorecard and rhythm
1. Three layers: executive (impact, adoption, three numbers), management (driving KPIs, red / yellow / green), technical (system health, data quality, guardrail incidents, reversal-trigger status).
2. The weekly review, named: 20 minutes, one owner, reads the scorecard and the pilot's weekly row, decides what to change (a message, a criterion, a step), and the owner approves each change. In SCALE this is a cadence you run; the GTM AI OS adds the mechanism that proposes changes from a log.
3. (Q) "Which revenue meetings already exist?" Attach artifacts to them before adding any. Weekly: pipeline review by motion, signal refresh, one ride-along. Monthly: review and rewrite. Quarterly: ICP refresh, reallocation, re-run step 1. Owner, day, artifact read, artifact updated. Offer to place on the calendar if connected.

## Part C: Who owns AI
1. (Q) Company size and regulation. Small or unregulated: three to five people. Mid-size or regulated: the full model.
2. Executive sponsor as high as possible. If it's HR: HR on the team, not in the chair; say so plainly.
3. Technology co-chair (speed, build, safety) and business co-chair (ROI, adoption, prioritization). Core team by title: legal, security, enablement, GTM, product and engineering, RevOps as data steward, industry-specific roles.
4. Purpose, decision rights, intake, budget path, relationship to tiger teams. Biweekly to start, two standing items: the prioritized use-case list and performance of what shipped.
5. Deep: the 90-day and 12-month view, including whether a dedicated forward-deployed pod reports to the CRO, RevOps, or the CTO. Don't deploy it until the foundations exist.

## Part D: The rules
1. Inherit any existing IT or security policy; extend, don't duplicate.
2. Thirteen areas, one paragraph each for small orgs, full for regulated: roles, operating model, risk and compliance, rules of use, security, model management, bias, tools and monitoring, transparency, accountability, trustworthiness, responsibility, incident response.
3. Risk register seeded from the step 3 reversal trigger and guardrails: risk, likelihood, impact, control, owner, review date.
4. Quality gates before any AI output reaches a customer, a forecast, or a system of record.

## Part E: The 30-day commitment
Ask, one at a time: (Q) the one AI opportunity you'll act on in 30 days; (Q) the specific action; the resources needed; (Q) the metric that proves it. Write it as one sentence at the end of the file and read it back.

## Write
`5-evolve-system.md` from `references/evolve-template.md`. Quality gate: no "expand" to the whole org; a scorecard nobody is scheduled to read isn't done; a CoE with one chair isn't done.

## Hand off
Mark the step done in `progress.md`. Return to `scale-start` for the close: the map of files and the 30-day commitment read back.