AI Centre of Excellence Playbook Template
A playbook for establishing and operating an AI Centre of Excellence (CoE) that drives AI adoption across your organisation. Defines the mission, team structure, operating model, service catalogue, and maturity milestones for a successful AI CoE.
Overview
What's included
CoE Charter & Mission
AI Centre of Excellence Charter
Organisation: Established: Sponsor: CoE Lead:
Mission Statement
The AI Centre of Excellence exists to [select/customise]:
- Accelerate responsible AI adoption across [Organisation name]
- Build internal AI capabilities and reduce dependence on external consultants
- Establish standards and best practices for AI development and governance
- Enable business units to identify, prioritise, and deliver AI use cases
Strategic Objectives (Year 1)
Operating Principles
- Business-led, technology-enabled — We start with business problems, not technology
- Teach, not just do — We build capability in business units, not dependency on the CoE
- Responsible by default — Ethics, governance, and security are embedded in everything we do
- Measure what matters — We track business outcomes, not just technical outputs
- Iterate and learn — We favour quick pilots over long planning cycles
Reporting Line
The CoE reports to:
Governance
- Steering committee: Meets (monthly/quarterly) with representation from
- Budget authority: £ annual budget; individual projects up to £ approved by CoE lead; above £ requires steering committee
Team Structure & Roles
Team Structure & Roles
Core Team
| Role | Headcount | Responsibilities | Reports To |
|---|---|---|---|
| CoE Lead | 1 | Strategy, stakeholder management, budget | CTO / CDO |
| AI/ML Engineer | Model development, MLOps, technical delivery | CoE Lead | |
| Data Engineer | Data pipelines, data quality, infrastructure | CoE Lead | |
| AI Product Manager | Use case prioritisation, requirements, delivery | CoE Lead | |
| AI Governance Lead | 1 | Ethics, compliance, risk, policy | CoE Lead |
| AI Trainer / Enablement | 1 | Training programmes, champions network, adoption | CoE Lead |
Extended Team (Part-Time / Shared)
| Role | Source | Allocation |
|---|---|---|
| Business analysts | Seconded from business units | % per project |
| UX/Design | Shared services | As needed |
| Legal/Compliance | Legal team | Advisory |
| Security | InfoSec team | Review and advisory |
Scaling Model
| CoE Maturity | Team Size | Focus |
|---|---|---|
| Year 1 (Establish) | 3-5 | First pilots, governance, quick wins |
| Year 2 (Grow) | 6-10 | Scale successful pilots, build platform |
| Year 3 (Optimise) | 8-15 | Federated model; CoE as enabler, not bottleneck |
AI Champions Network
Volunteer champions embedded in business units:
- 1 champion per business unit / department
- 2 hours per week dedicated to AI champion activities
- Monthly meetup facilitated by CoE
- Responsibilities: local AI advocacy, feedback loop, use case identification
Service Catalogue & Engagement
Service Catalogue
Services Offered
| Service | Description | Typical Duration | Who It's For |
|---|---|---|---|
| AI Discovery Workshop | Facilitated session to identify AI use cases | Half day | Business unit leaders |
| Use Case Assessment | Evaluate feasibility, value, and effort for a specific use case | 1-2 weeks | Project owners |
| Pilot Delivery | Design, build, and evaluate an AI pilot | 4-8 weeks | Approved projects |
| Production Deployment | Take a validated pilot to production | 4-12 weeks | Post-pilot projects |
| Governance Review | AI risk assessment and compliance review | 1-2 weeks | All AI projects |
| AI Training | Deliver training programmes (awareness, practitioner, builder) | Ongoing | All employees |
| Advisory / Office Hours | Ad-hoc guidance on AI questions | 1 hour slots | Anyone |
Engagement Process
- Submit request: Business unit submits an AI opportunity via [form/channel]
- Triage (1 week): CoE reviews and assigns priority based on value, feasibility, and strategic fit
- Discovery (1-2 weeks): CoE runs a discovery workshop to refine the use case
- Assessment (1-2 weeks): Formal use case assessment with go/no-go recommendation
- Pilot (4-8 weeks): If approved, CoE delivers a structured pilot
- Scale (4-12 weeks): If pilot succeeds, transition to production
Prioritisation Criteria
| Criterion | Weight | Score (1-5) |
|---|---|---|
| Business impact (revenue or cost) | % | |
| Strategic alignment | % | |
| Data readiness | % | |
| Technical feasibility | % | |
| Change readiness of the team | % | |
| Weighted total | 100% | ___ |
Instructions
How to use this template
Define the mission and get executive sponsorship
A CoE without executive sponsorship and budget will not survive. Secure commitment before building the team.
Start small with a core team
Begin with 3-5 people and deliver 2-3 quick wins in the first quarter. Expand based on demand and demonstrated value.
Establish the engagement process
Make it easy for business units to submit AI opportunities. A clear intake process prevents the CoE from becoming a bottleneck.
Measure and communicate value
Track business outcomes (not just number of projects) and report to the steering committee regularly. The CoE's budget depends on demonstrable ROI.
Watch Out
Common mistakes to avoid
FAQ
Frequently asked questions
Start with 3-5 people in Year 1. Most mature CoEs have 8-15 core members, supplemented by an extended network of champions and embedded analysts. The right size depends on your organisation's AI ambition and the number of business units you serve.
Common options: under the CTO (technology-led), under the CDO (data-led), or as a cross-functional unit reporting to the COO (operations-led). The best fit depends on your organisation's structure and AI maturity.
Track the business value delivered by CoE projects: cost savings, revenue impact, efficiency gains. Compare this to the CoE's operating cost. A well-run CoE typically delivers 3-5x its operating cost in business value within 18 months.
The CoE should be pragmatic: buy/configure for common use cases and build custom solutions only where AI creates competitive advantage. The CoE adds value through use case identification, vendor evaluation, and responsible deployment — not just coding.
Typically after 18-24 months, when business units have built enough AI capability to run their own projects. The CoE then shifts to a centre of enablement: setting standards, providing training, managing governance, and supporting complex projects.
Need a custom AI template?
Our team can build tailored templates for your specific business needs. Book a free strategy call.