AI Transformation Playbook

The core sequence

Brotman and Sack distill the methodology from dozens of interviews and case studies into five sequential steps:

AI education and proficiency — Build baseline understanding across the organization. This must come first because downstream steps (council decisions, use policies, pilot evaluation) require AI literacy to be effective. Challamel at Moderna: “You can’t effectively advise the company on an appropriate AI use policy if you don’t have a basic understanding of how the foundational AI systems work.”
AI council — A cross-functional group of AI-literate leaders that drives initiatives and ensures alignment with business objectives. Needs executive sponsorship, defined cadence, and clear purview. IT and legal should be involved from the start.
AI use policy — Ethical guidelines, data governance, compliance measures. Counterintuitively, a well-designed policy increases freedom to experiment — it provides the guardrails that let teams run pilots without fear of violating security, privacy, or ethics standards.
Road map and pilots — Auditing existing AI tools, assessing potential use cases across cost savings, experience improvement, and revenue growth. Prioritize ruthlessly. Paul Roetzer: “Have someone own the pilot project, benchmarking performance before and after, setting a ninety-day limit.”
AGI-horizon assessment — Regular evaluations of where AI capabilities are heading. Anticipating near-term capability jumps (the “middle era” between current AI and AGI) informs which pilots to fund and how aggressively to invest.

Three deployment approaches

The playbook is not one-size-fits-all. Case studies reveal three distinct approaches:

Approach	Champion	Mechanism	Speed	Risk
Bottom-up	Eric Vaughan, IgniteTech	Gamified scoring, mandatory participation, “AI Mondays” (20% of all hours), bottom performers fired	Fastest diffusion	Extreme disruption to operations
Middle-out	Alicia Parker, Tishman Speyer	CMO adopts for her department, shares learnings laterally to rest of org	Moderate, safe	May not reach full company
Top-down	Matt Britton (Suzy), Sal Khan (Khan Academy), Stéphane Bancel (Moderna)	CEO builds AI tools personally, demonstrates value, then cascades	High credibility	Depends on CEO’s AI literacy

Each approach works. The choice depends on organizational culture, executive sponsorship, and urgency.

Case study: Moderna

Moderna’s AI transformation under Brice Challamel is the most detailed example in the literature:

Foundation phase (2021–2022): Challamel conducted 270 interviews asking each person: “What would make you say that you have succeeded beyond your wildest expectations?” Discovered basic IT infrastructure had to be fixed first — “there were rooms with no plugs to charge laptops.”

Launch phase (mid-2023): After a second round of 150 interviews, the team built mChat (internal ChatGPT on the API), ran a companywide prompt contest modeled on Xprize, with CEO Bancel’s podcast as kickoff. Results: 3,000 active members (the entire company), 400+ prompt submissions, 180 identified game-changing use cases.

Champions phase: The contest’s voting mechanism naturally surfaced the most proficient and engaged employees. The top performers self-selected into the Gen AI Champions Team (GACT) — a 100+ person AI council that meets biweekly with 60–70% attendance. Challamel: “If you’re good at finding a great prompt and you’re popular so people want to vote for you, then you’re our champion.”

Scaling phase: CEO Bancel’s ambition — “We’re looking at every business process — from legal, to research, to manufacturing, to commercial — and thinking about how to redesign them with AI.” Target: same output as a 100,000-person company with a few thousand employees.

Challamel’s transformation framework: culture, business, technology (not the conventional “people, process, technology”). “I’m not an HR person. I’m a transformation person. I need to change culture, not people, to drive behaviors that lead to business value.”

Case study: Khan Academy — speed over sequence

Khan Academy had early access to GPT-4 (summer 2022, before ChatGPT launched). Rather than follow the full playbook sequentially, Khan moved from demo to working prototype of Khanmigo in two weeks. His approach:

Embraced AI’s limitations as design constraints, not reasons to stop
Got 40 employees under NDA within two months, entire team within three months
Used “show don’t tell” — the working prototype converted skeptics faster than arguments
Retroactively implemented formal playbook elements (training, governance) after initial momentum

Khan on the biggest mistake: “I wish I was even more aggressive in terms of the pace of our internal AI adoption in our everyday processes.”

Challamel’s historical pattern

Challamel observed that every major technology revolution follows the same democratization pattern:

Technology	Invention	Democratizing application	Time lag
Computers	Turing machine (1936)	Personal computer (1979)	43 years
Internet	ARPANET (1969)	World Wide Web/browser (1996)	27 years
AI	Neural networks (decades)	ChatGPT (2022)	Compressed

“AI itself is not a technology revolution. The technology has been there for a while. But transformer models and gen AI — now that’s a democratization revolution.” This aligns with Technological Revolution Cycles — the turning point arrives when a general-purpose technology acquires a consumer interface.

The services layer: executing the playbook at scale

The playbook provides the methodology, but execution in Fortune 1,000 companies requires an entire services ecosystem. Levie estimates “ten years of work for Accenture in every enterprise on the planet” — encompassing data remediation (fixing fragmented and legacy systems), workflow redesign (restructuring processes for agents rather than people), accountability architecture (defining who is liable when agent output fails), and continuous re-engineering (adapting workflows as models change).

This creates a new category of professional services that combines AI fluency with domain expertise and organizational change management. The agent operator role (see AI Agents and Job Redefinition) is the in-house version of this capability; the services firms provide the at-scale version. Both the in-house and outsourced models are necessary — the transformation is too large and too continuous for either alone.

The services boom reflects AI succeeding in the real world. As Levie observes: “We’re nowhere near eliminating the human from the workflow.” When the best models still require ~15% human correction on output, the professional services layer that manages the human-agent collaboration boundary becomes a permanent feature of the enterprise AI stack, not a transitional one.

Enterprise AI Adoption Lag — What happens without a playbook
Shadow AI and Organizational Enablement — The enablement approach the playbook formalizes
The Proficiency Gap — The gap the playbook is designed to close
AI as Organizational Force Multiplier — What the playbook unlocks when executed
Technological Revolution Cycles — Historical pattern Challamel’s framework parallels