---
type: "synthesis"
tags: ["synthesis", "scaling", "pilots", "discipline"]
sources: ["execution"]
id: "cross-scaling-discipline-sunsetting"
sourceVaultSlug: "hbr-seg-execution"
originDay: 8
articleStem: "hbr-seg-execution"
sourceUrl: "(unified vault: 7 sources)"
sourceTitle: "HBR — Firm Ⅱ-C · Execution quality — correct execution of AI"
---
## Pilots stall; the discipline to scale is the skill

The corpus converges on a shared choreography for getting from experiment to enterprise value:

- **A060 names the failure modes**: [[concept-experimentation-trap]] (pilots never leave the lab) and [[concept-pilot-theater]] (celebrating activity over outcomes). The fix is Performance Drive — [[action-sunset-redundant-efforts|sunset low-impact efforts]] (echoing J&J consolidating ~900 pilots).
- **A062 supplies the unit**: [[concept-narrow-deep-use-cases]] validated by [[action-controlled-experiments]] before you scale or cut.
- **A089 supplies the enablers**: the [[framework-four-pillars-of-ai-success|four pillars]] — sponsorship, partners, cross-department communication, data management — are what let leaders scale where laggards stall.
- **A093 supplies the exemplar**: Moody's went from copilot to commercial product in ~5 months by [[action-deploy-gen-ai-company-wide|deploying to everyone]], enabling grassroots ideas, then resourcing the winners ([[framework-moodys-guiding-principles|'deliver impact']]).

## The synthesis

The discipline is a funnel: **experiment widely, measure narrowly, kill ruthlessly, scale the few.** Notice the shared enemy of both extremes — endless lab experimentation (A060) *and* premature enterprise-wide commitment (A062). The winners run cheap, disciplined experiments (Applied Curiosity in SHAPE), then commit hard behind measured winners. Scaling is not a technology event; it's a portfolio-management and prioritization skill. See [[cross-genai-measurement-problem]] and [[cross-winners-losers-execution-gap]].