Enterprise AI Sales Shift to 'Centers of Excellence'
The enterprise sales process for AI has reportedly shifted, with centralized AI Centers of Excellence (CoEs) now controlling procurement. This change turns what was once a single-threaded sale into a multi-threaded political process involving complex build-versus-buy debates. This dynamic requires vendors to navigate more complex organizational structures to secure deals.
- An AI Center of Excellence (CoE) is a central team created to fix problems arising from uncoordinated AI adoption, such as fragmented efforts, inconsistent governance, and duplicated spending across different business units. - The "build vs. buy" decision is a key function of the CoE; building a custom solution can take 12-24 months for full production, while buying an external tool can be deployed in 3-9 months. For large-scale use, building may offer lower long-term inference costs despite higher initial expenses. - CoEs are typically led by a senior executive like a Chief AI, Data, or Technology Officer and often use a hybrid "hub-and-spoke" model where a central team provides infrastructure and governance, while execution is embedded in business units. - A primary role of the CoE is to standardize the enterprise AI technology stack, creating an approved set of MLOps tools, cloud platforms, and data infrastructure to prevent tool sprawl and ensure scalability and security. - Key roles within these centralized teams that vendors must engage include AI Platform Architects, who own the reference architecture; MLOps Engineers; and AI Safety and Governance Leads, who define policies and risk controls. - The creation of a CoE elevates procurement from a cost-focused function to a strategic one, where decisions are based on total value, including supplier innovation, risk mitigation, and compliance with emerging regulations like the EU AI Act. - A typical 12-month roadmap for a new CoE involves defining the mission and governance in the first quarter, launching high-impact pilots in months 4-6, and scaling successful pilots across the organization in months 7-12. - For 90% of enterprise use cases, buying an AI platform is the most practical choice as it reduces time-to-value from over 18 months to just a few weeks. Building is typically reserved for when the AI itself constitutes core intellectual property.
