There's no single right AI team, only the right one for your stage. Building a heavy platform team at seed burns runway; running Series B on contractors caps you. Here's the progression that works.
Seed: prove value, cheaply
Put a fractional lead in to set direction, plus one or two embedded builders (forward-deployed) to ship the first AI capability. Avoid permanent headcount until the value is proven.
Series A: own the core
Convert what works into an owned team: your first full-time senior AI engineer and, if AI is central, an AI PM. Keep embedding specialists for spikes rather than hiring for every skill.
Series B: scale and govern
- Specialize roles: ML, MLOps, data engineering, product.
- Stand up a platform team for shared tooling and reliability.
- Add governance: model review, compliance, cost management.
- Consider a hub-and-spoke structure as teams multiply.
