Choosing an Operating Model
Module 3 · Operating Model and Platform

Choosing an Operating Model

Compare centralized, hub-and-spoke, and federated AI operating models, decide what the center owns versus the business units, and design the core team a CAIO needs.

Who Does the Work?

Strategy says where AI should create value. The operating model says who builds, buys, runs, and governs AI, and how they work together. Get it wrong and you’ll either have a central team that’s a bottleneck or business units reinventing the same things with no guardrails.


Three Models

Centralized

A central AI team, often called a center of excellence (CoE), does most AI work for the whole organization.

  • Strengths: consistent standards, efficient use of scarce talent, strong governance, easier reuse.
  • Weaknesses: becomes a queue; distant from business problems; business units feel AI is “done to them.”
  • Fits: organizations early in AI maturity, or with very scarce AI skills.

Federated

Each business unit runs its own AI teams with a light central function for policy.

  • Strengths: close to the business, fast, strong ownership.
  • Weaknesses: duplication, inconsistent risk practices, hard to share platforms or talent, difficult to see the whole portfolio.
  • Fits: very large, diverse organizations with mature business units.

Hub-and-spoke

A central hub owns platform, standards, governance, and scarce expertise. Spokes, embedded in business units, own use cases and adoption.

  • Strengths: balances consistency with business ownership; reuse through the hub, speed in the spokes.
  • Weaknesses: needs clear division of responsibilities and strong relationships; can drift toward either extreme.
  • Fits: most mid-to-large organizations once AI moves beyond early pilots.
CentralizedAI work done by the centerFederatedEach unit runs its own AIHub-and-spokeCenter enables, units deliverCoEBUBUBUBUPolicy onlyBUBUBUBUHubBUBUBUBU= AI team embedded in a business unit (BU)

Many organizations start centralized and move toward hub-and-spoke as skills spread and standards mature. Plan for that evolution rather than treating the first model as permanent.


What the Hub Owns vs. the Spokes

The hub (central team) ownsThe spokes (business units) own
AI strategy, portfolio view, and prioritization processUse-case identification and business cases
Shared platform, model access, and approved toolsProcess redesign and change in their area
Standards: security, evaluation, documentationDay-to-day product ownership of their AI solutions
Governance, risk tiering, and reviewMeasuring and reporting value
Scarce expertise (ML engineering, evaluation, AI risk)Domain experts and frontline adoption
Enterprise vendor contractsLocal champions and training

Write this down and agree it with business-unit leaders. Ambiguity about who owns a failing AI product is a common source of conflict.

⚠️ Common Pitfall: The Center That Builds Everything

Common mistake: a strong central team keeps building every use case itself because it’s faster and higher quality than the business units’ attempts.

It works for a year. Then demand exceeds capacity, business units feel ignored, and the CAIO’s team becomes the reason AI is slow. The hub’s long-term job is to make the spokes capable: platforms, patterns, training, and review, not delivery of everything.


The Core Team

Whatever the model, a CAIO typically needs these capabilities, in-house or through partners:

  • AI product management: turning business problems into well-scoped AI products with clear metrics.
  • AI and ML engineering: building, integrating, and operating AI systems, including retrieval, agents, and evaluation.
  • Platform and architecture: the shared infrastructure, model access, and integration patterns.
  • Data partnership: close links to data engineering and governance, often in the CDO’s team.
  • AI risk and governance: policy, risk assessment, model documentation, and regulatory tracking.
  • Change and enablement: training, communications, champions networks, and adoption measurement.

Small, senior, and cross-functional beats large and specialized at the start.

🧪 Try It: Choose Your Model
  1. Describe your organization’s AI maturity in one sentence: early pilots, scattered production, or broad adoption?
  2. Pick the operating model that fits today, and the one you’d aim for in two years.
  3. Fill in a two-column hub-versus-spoke ownership table for your organization.
  4. List the three core-team roles you’d hire first, and why.
🧠 Knowledge Check

1. What's the main weakness of a fully centralized AI model?

2. In a hub-and-spoke model, who typically owns process redesign in a business area?

3. What's the hub's long-term job?

Up Next

The next lesson covers the foundations the operating model runs on: data, platform, and cost.