Team Structure for AI Delivery

How you organise the people working on AI significantly affects what gets built and how well it works. The three common models - centralised AI team, embedded specialists in business units, and hybrid approaches - each have trade-offs. A centralised team builds deep technical expertise and avoids duplication but can become disconnected from business reality and create bottlenecks. Embedded specialists stay close to the problems but can feel isolated, lack career development paths, and reinvent wheels. Hybrid models try to capture the best of both but add coordination complexity. There's no universally right answer, and many organisations evolve their structure as they mature. What matters more than the specific structure is ensuring that AI teams include diverse skills - not just data scientists and engineers but also product thinkers, domain experts, and people who understand the operational context. Cross-functional collaboration isn't optional; AI projects that stay in a technical silo consistently underdeliver. Whatever structure you choose, make sure there are clear paths for AI initiatives to move from experimentation to production, with the right handoffs and support at each stage.