How companies structure AI teams has changed dramatically.

And understanding the new structures helps you target the right role.

Old structure (2020-2023): centralized AI team that serves the whole company. Data scientists build models. Engineers deploy them. Long backlogs. Slow delivery.

New structure (2026): embedded AI engineers within product teams. Each product team has ML capability. A central "AI Platform" team provides shared infrastructure. Faster delivery. Tighter product integration.

What this means for job seekers:

Embedded AI roles are more plentiful and closer to the product. You'll work directly with product managers and customers. The scope is narrower but the impact is more visible.

Platform AI roles are fewer but more technically deep. You're building tools and infrastructure that multiple teams use. Requires strong systems engineering.

Applied Research roles still exist but are rarer. Reserved for companies where novel ML is the core product differentiator.

The optimal career path in 2026: start embedded (learn how AI creates product value), move to platform (build systems thinking), then specialize or lead.

The worst career move: joining a centralized AI team at a company that doesn't understand how to integrate AI with products. You'll build impressive models that nobody uses.

Before accepting any AI role, ask: "How does the AI team interact with product teams?" The answer tells you more about your future impact than the job description ever will.