A WEBSITE DIDN’T MAKE THE MALL AN INTERNET COMPANY.

THE SHOPPING-MALL TEST

Andrew Ng's AI Transformation Playbook — written from the experience of building Google Brain and Baidu's AI Group — opens with an analogy that has aged perfectly: in the internet era, a shopping mall that built a website did not thereby become an internet company. An internet company is organized to do what the internet does well — hundreds of A/B tests running at once, products shipped weekly. The same test now applies to AI: a company with ChatGPT subscriptions is not an AI company. An AI company is organized to do what AI does well — and that is an org-design question, not a procurement one.

FIVE MOVES, PARAPHRASED

• First projects buy momentum, not maximum value. Ng's advice: choose early projects to succeed — meaningful enough to convince the organization, feasible enough to show traction fast, with a measurable business objective defined up front. The flywheel matters more than the first project's size.
• Capability moves in-house over time. External partners accelerate the start; a durable advantage needs internal capability — a centralized AI function that serves every division and sets standards, rather than scattered experiments per department.
• Training is tiered, not uniform. The playbook prescribes hours per role: on the order of 4+ hours for executives (what AI can and can't do, strategy, resourcing), 12+ for division leaders (setting direction and tracking AI projects), 100+ for engineers. A lunch-and-learn is not an adoption strategy.
• Data is a strategy, not a warehouse. Unify the silos, yes — but Ng's sharper warning is the reverse mistake: terabytes of data are not automatically valuable, and over-investing in low-value data collection (or acquiring companies for useless data) is a documented CEO failure mode. Bring AI judgment in before the data spend.
• Communicate deliberately. Internal fear of job automation and external over-hype both corrode adoption; both are managed with explicit communication, not silence.

THE TIMELINES AGED; THE ORG LOGIC DID NOT

The playbook dates from December 2018 and predates the generative-AI wave; one number shows its age: Ng budgeted 6–12 months for a first project to show traction. With today's models and integration tooling, a first agent goes to production on live data in weeks — the entry barrier has moved from ML engineering to process design. But strip the timelines and the organizational logic is untouched: momentum first, capability in-house, training per role, data as strategy. Our five-phase protocol is, in effect, that logic compressed for the agentic era — with the training tier built in as its own service line, because the skills gap, not the technology, is still what stalls adoption.