Expanding AI's Presence from Isolated Trials to Everyday Applications
In the realm of artificial intelligence (AI), many organisations begin their journey with isolated pilots, known as Islands of Experimentation (IOEs). These agile, localised teams test AI tools on niche problems, generating initial evidence but limiting scale.
However, to scale and unlock the full potential of AI, organisations must progress beyond IOEs. This transition is facilitated through structured stages that build governance, expertise, and scalability into AI efforts. The key structural stages involved are typically a Centre of Excellence (CoE) and later a Federation of Expertise (FOE).
Initially, organisations establish a Centre of Excellence (CoE)—a centralised team that governs AI initiatives by setting standards, managing resources, and developing core capabilities and best practices. The CoE acts as the foundational hub to ensure strategic alignment, governance, and efficient resource use across AI projects.
One example of a successful CoE implementation is a major insurance company that established such a team and introduced a dedicated AI and data leadership role that influences board-level decisions. This strategic alignment has proven beneficial, allowing the company to prioritise AI initiatives that align with their strategic goals.
As AI maturity increases, enterprises evolve towards the FOE model, where AI capabilities are distributed but coordinated centrally. This federation integrates specialized AI teams embedded in different business units, supported by the CoE's governance and shared infrastructure. The FOE preserves domain expertise while enabling cross-unit collaboration, knowledge sharing, and consistency in AI adoption methods.
A global security firm, for instance, used its COE to tackle complex risk prediction and built a predictive model with 88% accuracy. The FOE model allows some projects to not yield immediate gains but lay foundations for future use cases, thereby enabling long-term value creation.
However, the FOE model presents its own challenges. Misaligned incentives can cripple AI scaling, and companies must adopt joint objective schemes to ensure appropriate resource allocation and motivate all stakeholders to pursue measurable AI outcomes. To prevent fragmentation, consistent communication, shared standards, and centralised oversight are essential.
Standardisation is crucial for COEs, involving establishing a consistent data infrastructure, formalising development and deployment processes, and creating clear roles and responsibilities across business units (BUs) and IT. Four key limitations of IOEs are: narrow training data restricts transferability, limited data architecture halts model evolution, lack of enterprise-wide visibility leads to redundancy, and resource constraints cause stagnation.
In conclusion, transitioning between the stages of Islands of Experimentation, CoE, and FOE requires clear roadmaps, governance backed by senior leadership, investment in technical infrastructure and data management, and cultural change including workforce training. This progression mitigates fragmentation, ensures ethical and regulatory compliance, and accelerates realizing AI’s full enterprise impact.
Machine learning, a key component of artificial intelligence (AI), is integrated into AI tools tested by agile, localised teams in the initial stages of AI implementation, known as Islands of Experimentation (IOEs). To scale AI and fully realise its potential, organisations must transition from IOEs to structural stages that incorporates governance, expertise, and scalability, such as a Centre of Excellence (CoE) and eventually a Federation of Expertise (FOE). In these subsequent stages, technology advances are further propelled through centralised management, collaboration, and cross-unit innovation.
