The difference between organisations where Claude adoption stalls after the pilot and those where it becomes enterprise-wide in 90 days is almost always the same thing: the Centre of Excellence model. Here's how to build one — the governance structure, the Champions programme, the prompt library, and the measurement framework that make Claude deployment self-sustaining.
In our experience across 200+ enterprise Claude deployments, the single most predictive factor for long-term adoption success is whether the organisation establishes a Centre of Excellence model during the initial deployment. Organisations that do see Claude adoption expand from one department to the entire company. Organisations that don't typically see adoption plateau — sometimes even regress — as initial enthusiasm fades and there's no structure to sustain momentum.
The failure mode without a CoE is predictable: the initial deployment succeeds, the department that piloted Claude sees real productivity gains, leadership is pleased. Then the three-month mark arrives: the external consultants have wrapped up, the initial training energy has dissipated, a few power users are still using Claude regularly but most employees have reverted to old habits, and there's no internal owner driving adoption forward. Claude becomes "that AI thing IT tried last year."
The CoE model prevents this by creating the organisational infrastructure that makes Claude adoption self-sustaining. It provides: a clear internal owner accountable for outcomes, a governance framework that enables confident use, a peer-to-peer adoption mechanism through Champions, a continuously updated prompt library that lowers barriers to use, and a measurement system that generates the ROI evidence that keeps leadership engaged.
Effective Claude CoEs are deliberately lightweight — the goal is to create enough structure to support scale without creating bureaucratic overhead that slows adoption. In companies of 500–5,000 employees, the CoE typically requires 0.5–1 full-time equivalent. In companies of 5,000+, 1–2 dedicated people plus a network of part-time Champions can govern adoption across the entire organisation.
The CoE owns five primary functions. First, usage policy governance: maintaining and communicating the acceptable use policies, data handling rules, and security requirements for Claude. This is not about restriction — it's about enabling confident use by clarifying what is and is not appropriate. Second, prompt library stewardship: maintaining and publishing the enterprise prompt library, collecting new prompts from Champions, and running quarterly prompt quality reviews. Third, Champions programme management: selecting, training, supporting, and connecting the Champion network. Fourth, vendor relationship management: owning the Claude Enterprise contract, communicating with Anthropic, tracking new capabilities, and managing licence provisioning. Fifth, ROI measurement: establishing and maintaining the measurement framework that tracks Claude's impact across the organisation.
The CoE operates with a lightweight governance structure. A monthly CoE steering group brings together the CoE lead, the Champions from each major department, a representative from IT/security, and a sponsor from the leadership team. The meeting covers: usage metrics, new use cases identified by Champions, prompt library additions, policy updates, and the ROI scorecard. This meeting rarely runs more than 45 minutes and produces a clear record of Claude's enterprise value over time.
We've built CoE structures for 50+ enterprise Claude deployments. Get a free assessment and we'll show you exactly what's needed for your organisation.
Request Free Assessment →The Champions programme is the most important element of the CoE model — and the most underestimated. In every deployment where we've implemented it, the Champions network has driven adoption more effectively than any other mechanism. The reason is simple: people learn new tools from trusted peers, not from IT training sessions or top-down mandates.
Champions are selected employees in each department who receive advanced Claude training, become the go-to resource for their colleagues, contribute to the enterprise prompt library, and represent their department in CoE governance. The selection criteria matter: Champions should be respected by their colleagues, enthusiastic about Claude, and credible in their domain — they don't need to be technically sophisticated. A highly respected senior associate in legal who's enthusiastic about Claude is a far better Legal Champion than a junior tech-savvy paralegal who isn't trusted by the partners.
Champion training covers three areas beyond the standard employee training. First, advanced prompting techniques specific to their department's highest-value use cases. Second, the CoE governance model — what policies are in place, how to answer colleagues' questions about appropriate use, and how to escalate edge cases. Third, how to identify and document new use cases, contribute prompts to the library, and report on adoption and outcomes in their department.
The activation moment for Champions is their first department training session — typically two to three weeks after their own advanced training. Champions deliver this training themselves, using the CoE-provided materials but drawing on their own experience and examples from their department. Peer-delivered training consistently outperforms externally-delivered training in our deployments: the credibility of "this is how I use it in our department's actual work" is irreplaceable.
Complete training materials for every level — including the Champions advanced training programme.
Download Free →The enterprise prompt library is one of the CoE's most tangible deliverables and one of the most significant productivity multipliers in the organisation. The library collects, tests, refines, and publishes the prompts that work best for the organisation's specific use cases — so every employee can immediately access effective prompts rather than starting from scratch every time.
A well-built prompt library at a 1,000-person company typically contains 50–150 tested prompts, organised by department and use case. Each prompt includes: the prompt text, the use case it addresses, example inputs and outputs, and notes on how to customise it. The library is published in a format that employees can access easily — typically a Notion or Confluence page, or a more sophisticated tool if the IT team supports it.
The library is a living document. Champions contribute new prompts from their departments. The CoE runs quarterly quality reviews where prompts are tested, updated to reflect new Claude capabilities, and refined based on user feedback. The best-performing prompts are promoted to "Featured" status and highlighted in communications. This continuous improvement process means the library gets better over time, which sustains engagement with it.
Not all prompts are created equal, and the library should reflect this. We recommend a three-tier classification: draft (contributed by a Champion, not yet reviewed), approved (reviewed by the CoE for accuracy and effectiveness), and featured (proven high-value prompts recommended for widespread use). This tiering helps users identify the most reliable prompts while still allowing the library to grow quickly through Champion contributions.
The CoE's measurement function serves two purposes: it generates the ROI evidence that keeps leadership engaged and investment flowing, and it identifies where Claude is delivering value versus where adoption is lagging and intervention is needed. Both purposes are critical to sustaining enterprise-wide adoption.
Effective measurement in our deployments combines quantitative usage metrics (active users by department, queries per user, prompt library utilisation) with qualitative outcome tracking (time savings reported by Champions, use cases generating positive feedback, tasks where Claude is now the standard workflow). We recommend a simple monthly metrics dashboard that the CoE presents to the leadership sponsor — showing adoption trends, top use cases, and estimated time savings across the organisation.
The time savings metric is the most compelling for leadership and the most straightforward to track. Champions collect a simple monthly survey from their department: for each Claude use case, how many times per week do you use it, and roughly how many minutes does it save per use? This data, aggregated across the organisation, produces a monthly hours-saved figure that can be directly converted to a productivity dollar value. In deployments where this measurement is done consistently, the CoE's annual value documentation typically runs to millions of dollars — which is extremely effective at sustaining leadership support.
For detailed guidance on the implementation methodology, see our Enterprise Implementation service and our Pilot to Enterprise white paper. For Champions programme training materials, see our Training Curriculum white paper. For governance framework templates, see our Governance service.
Get a free readiness assessment — we'll design a CoE model that fits your organisation and create a 90-day build plan.
Weekly insights on Claude enterprise deployment, adoption strategies, and implementation best practices.