Enterprise AI Transformation Skills

Sixteen installable skills that help executives, operators, and consultants run enterprise AI transformation — from "where do we start?" to "is our agent governed correctly?" — distilled from 21 flagship research sources (Stanford, MIT, McKinsey, BCG, Deloitte, PwC, Accenture, NIST, EU AI Act, IMDA, WEF) published 2024–2026.

TL;DR. 95% of enterprise GenAI pilots return zero P&L impact. The 5% that win don't have better models — they operate differently across people, process, technology, and a layer of general strategy/diagnostics. This skill library encodes those operating patterns. Each skill is a structured prompt the AI runs for you; you ask the question, the skill produces a board-ready output.

The four buckets (30 seconds)

Skills are grouped by what they fix. Pick the bucket that matches your problem; pick the skill inside it that matches the question.

The default order is general → process → tech → people, but most real engagements jump around. The "Which skill when?" decision tree below does the picking for you.

Which skill when? — Decision tree

Start here. Read top to bottom; stop at the first row that matches.

If your question hits multiple rows, run the higher one first — it sets context for the rest.

Get started in 5 minutes

Install once

Claude Desktop (easiest) — Customization → Add plugin (+) → Create Plugin → Add Marketplace → Add from a repository → paste:

https://github.com/geledek/enterprise-ai-transformation-skills

Restart. All 16 skills activate from their trigger phrases.

Claude Code — claude plugin install https://github.com/geledek/enterprise-ai-transformation-skills

No paid plan? — Open a Project at claude.ai, paste any single skill's SKILL.md into Project Instructions, upload relevant references/ files as Project Knowledge. One skill per Project.

Full instructions for ChatGPT / Grok / others: see INSTALL.md.

Try your first skill (2 minutes)

Pick the question from the decision tree closest to your real situation, then ask the AI in plain English. Examples:

• "Should we pursue an AI agent that auto-approves expense reports under S$200?" → triggers general-idea-diagnostic
• "Where should we point AI first across our claims operation?" → triggers general-use-case-discovery
• "Design a 90-day pilot for an RM call-prep summarizer at our private bank." → triggers process-pilot-design
• "We deployed 11 agents but can't see what they're earning. Help." → triggers process-portfolio-observability
• "Our senior lawyers are openly hostile to the contract-review AI. What now?" → triggers people-frontline-engagement

Each skill walks the AI through 4–7 structured roles and produces a verdict (Fund / Reframe / Kill, Greenlight / Stage / Park, Approved / Conditional / Blocked, etc.). The output is meant to be sponsor-ready.

A typical executive flow

For a leader running an end-to-end AI transformation, this is the canonical sequence. You can branch off at any point — the buckets are designed to be invoked individually too.

┌─ general-maturity-assessment ─── where are we today?
│
├─ general-use-case-discovery ──── what should we pilot first?
│      │
│      └─ general-idea-diagnostic ─ is THIS specific idea worth doing?
│             │
│             └─ tech-data-deployment ─ where can the data legally run?
│                    │
│                    └─ process-pilot-design ─ design the 90-day pilot
│                           │
│                           └─ general-roi-gate ─ approve / fund / reject
│                                  │
│                                  └─ tech-buy-vs-build ─ buy or build?
│                                         │
│                                         └─ process-productionization ─ ship to prod
│                                                │
│                                                └─ tech-agent-guardrail ─ govern at runtime
│                                                       │
│                                                       └─ process-portfolio-observability ─ track net ROI
│
├─ people-readiness-conversation ── (run quarterly — surfaces gaps the tech can't fix)
├─ people-literacy-curriculum ───── (close the literacy gap, satisfy EU AI Act Art. 4)
├─ people-frontline-engagement ──── (when fearful experts are blocking rollout)
├─ people-tool-selection ────────── (choosing which AI tool to put in front of a group)
│
├─ general-peer-cases ──────────── (pull at any decision point: "what have others done?")
└─ tech-stack-diagnostic ───────── (run before any major scaling decision)

For full-transcript worked examples — four single-skill walkthroughs and five chained-flow operating decisions — see docs/usage-guide.md.

Cross-skill cautions

These skills were tested together. A few real interactions worth knowing about:

• Customer-facing pilots. general-use-case-discovery will surface customer-pain pools (NPS, conversion drop). process-pilot-design will downrate any first pilot with customer-facing blast radius. The discovery skill now defaults customer-facing candidates to Stage-and-watch unless the sponsor explicitly accepts the risk in the brief. Trust the pilot-design skill on first-pilot scope.
• Three different "70/20/10"s. Different sources use the same digits for different things. Each skill is now explicit about which one applies:
  • general-use-case-discovery / general-peer-cases → BCG 10/20/70 effort split (10% algorithm / 20% tech / 70% people-process)
  • process-portfolio-observability → 70/20/10 portfolio mix (70% core / 20% adjacent / 10% transformational)
  • people-literacy-curriculum → 70-20-10 training-budget split (70% sandbox / 20% delivery / 10% content)
• "BCG 88/25" can mean two things. people-readiness-conversation and people-literacy-curriculum use it for manager role-modeling. tech-data-deployment uses the same digits for broad-use-vs-value gap. Both anchors are real; the skills now flag which one they mean.
• Verdict vocabularies differ across gating skills by design — concept-stage (general-idea-diagnostic → Fund / Reframe / Kill), portfolio-selection (general-use-case-discovery → Greenlight / Stage / Park / Reject), funding-approval (general-roi-gate → Approve / Conditional / Return / Reject). They sit at different points in the lifecycle. Don't run all three on the same question.
• CEO public commitment vs. no-headcount-cut clause. people-readiness-conversation recommends a CEO-owned measurable productivity number. people-frontline-engagement requires a no-headcount-cut clause for pilot duration. These reconcile — public commitment is on outcome, the clause is on pilot-window — but the CEO must say so explicitly to the frontline.

Who is this for

• Investors evaluating AI-first companies and AI initiative funding proposals
• Operators running AI transformation programs, CoEs, or portfolio pilots
• Founders assessing build vs. buy, structuring AI product strategy
• Boards and executives needing structured frameworks for AI governance and investment decisions
• Consultants and advisors delivering enterprise AI strategy engagements

Skills produce board-ready outputs. No coding required to use them.

How to extend

Three forks:

1. Add references — drop an extract into references/, follow the file shape in references/_index.md, update _index.md. Skill descriptions reference files by filename, not path.
2. Add cases — drop a diagnosed case into the relevant skills/<bucket>-<skill>/cases/. The skill will pull the closest analog automatically.
3. Fork a skill — copy any SKILL.md, rename it, adjust the role instructions and reference pointers for your sector, internal processes, or role-specific context.

See CONTRIBUTING.md for the full contribution shape.

Sources

21 flagship sources, 2024–2026:

See references/ for curated extracts from each source.

License

MIT — see LICENSE.

Author

Ray Han · June 2026