Leadership in data and AI is often distorted by the FAANG fallacy: the assumption that every organization should reproduce the infrastructure, team shape, or spending pattern of a technology giant. Scale is not a virtue by itself. The right architecture is the one that matches the organization's decisions, data maturity, risk tolerance, and ability to operate what it builds.
An internal data platform should make good work easier without becoming another central bottleneck. It can provide trusted foundations such as identity, cataloging, quality controls, deployment paths, observability, and reusable patterns. Product teams still need room to solve domain problems, but they should not have to rebuild the same controls for every model.
Build the team around outcomes
Successful teams combine technical depth with ownership of business results. Data engineers make reliable inputs possible, data scientists develop useful signals, ML engineers operate models, platform engineers create leverage, and domain leaders define what “better” means. These roles may belong to separate people or overlap in a small team; the responsibility must still be visible.
The economics of AI includes more than API or compute spend. Count engineering time, data maintenance, review queues, incidents, retraining, opportunity cost, and the cost of a wrong decision. A strong business case connects those costs to a measurable outcome and includes a plan for stopping work that no longer earns its complexity.
A leader's checklist
- Which business outcome justifies this system?
- What capability should be shared and what should remain domain-specific?
- Who owns reliability, quality, cost, and retirement?
- What evidence would cause us to scale, change direction, or stop?