{"id":34590,"date":"2026-10-05T18:00:00","date_gmt":"2026-10-05T16:00:00","guid":{"rendered":"https:\/\/askme.it\/insights\/the-ai-competence-center-is-it-really-needed-or-just-a-bureaucratic-layer\/"},"modified":"2026-03-26T12:23:40","modified_gmt":"2026-03-26T11:23:40","slug":"the-ai-competence-center-is-it-really-needed-or-just-a-bureaucratic-layer","status":"publish","type":"insights","link":"https:\/\/askme.it\/en\/insights\/the-ai-competence-center-is-it-really-needed-or-just-a-bureaucratic-layer\/","title":{"rendered":"The AI competence center: is it really needed or just a bureaucratic layer?"},"content":{"rendered":"
\n

Only one in five organizations is getting value from AI tools, and only one in fifty is seeing transformative value: this is the figure that Paul Delory, VP Analyst at Gartner, presented at the Gartner IT Infrastructure, Operations & Cloud Strategies 2025 conference in London. In the context of this evidence, Gartner’s advice was explicit: build an AI center of excellence within the I&O organization. Its main job is to be the air traffic controller of AI.<\/p>\n

The metaphor is precise. An air traffic controller doesn’t fly the planes: they coordinate movements, assign routes, manage priorities, and intervene when there’s a risk of conflict. An AI CoE doesn’t implement every AI project in the organization: it sets standards, validates use cases, manages governance, and enables business teams to work autonomously without creating risks.<\/p>\n<\/section>\n

\n

Four AI COE styles according to Gartner<\/h2>\n

Gartner identifies four distinct Center of Excellence styles for AI, each suited to a different stage of organizational maturity. The first is informational: the COE spreads knowledge, produces best practices, and documents use cases. It’s the natural starting point when the organization is in the early stages of AI adoption and needs to build internal awareness. The second is persuasive: the COE builds internal consensus, demonstrates value with targeted pilots, and helps business units develop their own use cases.<\/p>\n

The third style is prescriptive: the COE establishes standards, policies, and governance frameworks that apply to all AI projects across the organization. At this stage, AI has reached sufficient scale to require centralized control. The fourth is innovative: the COE drives experimentation with new AI capabilities, identifies opportunities not yet explored, and builds competencies the organization will need in the next transformation cycle. Gartner states that the COE’s objectives must evolve as AI usage matures: it’s not a fixed model, but a function that changes with the organization.<\/p>\n<\/section>\n

\n

The hybrid model: centralization and decentralization<\/h2>\n

Gartner recommends a distributed hybrid organizational model: a central COE team that collaborates and coordinates with decentralized communities distributed across the organization. The central team governs strategy, standards, and governance. The decentralized communities, led by hybrid profiles with both domain and technical expertise, implement projects in their respective business areas.<\/p>\n

This model resolves the tension between two real needs: the need for standardization and control that only centralization guarantees, and the need for speed and domain context that only decentralization allows. A purely centralized COE becomes a bottleneck. A purely decentralized one produces ungoverned proliferation, which is exactly the risk Gartner identifies as one of the most relevant in accelerated agentic AI adoption.<\/p>\n<\/section>\n

\n

AI skills: the real strategic issue<\/h2>\n

By 2026, the Chief Data and Analytics Officer’s ability to develop AI literacy, drive cultural change, and build a qualified workforce will be one of the three main determining factors in supporting business strategy, according to Gartner. Building a COE without building the skills within it is an exercise in structure without substance.<\/p>\n

The AI skills that Gartner identifies as fundamentally different from traditional ones are not about task execution: they’re about enhancing human capabilities. Where traditional skills serve to do things better, AI skills serve to become a better thinker, a better communicator, a more effective problem solver. A COE that limits itself to training people on tool usage without developing these deeper capabilities is building tool dependency, not organizational capability.<\/p>\n

Gartner also flags the opposite risk: skill atrophy. If people rely too heavily on AI and stop using their core competencies, they lose them over time. Effective AI teams maintain a balance between human capabilities and AI support, with periodic assessments to verify that critical skills are maintained in the roles that need them.<\/p>\n<\/section>\n

\n

The COE mandate: what it should decide and what it shouldn’t<\/h2>\n

The most common failure of an AI CoE is the wrong mandate. A COE that wants to approve every AI project becomes a bureaucratic funnel that slows down adoption. A COE that has no real decision-making power becomes an advisory body that gets ignored. Gartner indicates that the right mandate focuses on three areas: defining governance standards that apply to all AI projects regardless of who implements them; validating technical feasibility and risk of the most significant projects before go-ahead; and measuring the overall impact of AI investments across the organization.<\/p>\n

Everything else, vendor selection, technical implementation, managing agents in individual processes, should remain with business teams that have the right skills. A COE that overcentralizes loses speed. A COE that undercentralizes loses control. Calibrating this boundary is the real organizational challenge, and it needs to be recalibrated periodically as the organization’s capabilities evolve.<\/p>\n<\/section>\n

\n

When a formal COE isn’t needed<\/h2>\n

Not every organization needs a formal COE to govern AI. Gartner indicates that the starting point depends on scale and maturity. Organizations still exploring their first AI use cases don’t need a permanent structure: they need a clear owner for each project and a review process that involves security, legal, and privacy. A formal COE makes sense when the number of AI projects underway is high enough to require standardization, and when the risk of fragmentation and ungoverned proliferation outweighs the cost of maintaining a dedicated coordination function.<\/p>\n<\/section>\n","protected":false},"excerpt":{"rendered":"

Only 1 in 5 organizations is getting value from AI tools, and only 1 in 50 is seeing transformative value. Gartner recommends building an AI center of excellence that works like an air traffic controller. But building it wrong is worse than not building it at all.<\/p>\n","protected":false},"featured_media":34592,"menu_order":0,"template":"","insights_category":[577],"insights_tags":[645,729,733,803,853],"class_list":["post-34590","insights","type-insights","status-publish","has-post-thumbnail","hentry","insights_category-strategy-and-ai","insights_tags-center-of-excellence-en","insights_tags-generative-ai","insights_tags-governance-en","insights_tags-organization","insights_tags-strategy"],"acf":[],"_links":{"self":[{"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights\/34590","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights"}],"about":[{"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/types\/insights"}],"version-history":[{"count":1,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights\/34590\/revisions"}],"predecessor-version":[{"id":34591,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights\/34590\/revisions\/34591"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/media\/34592"}],"wp:attachment":[{"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/media?parent=34590"}],"wp:term":[{"taxonomy":"insights_category","embeddable":true,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights_category?post=34590"},{"taxonomy":"insights_tags","embeddable":true,"href":"https:\/\/askme.it\/en\/wp-json\/wp\/v2\/insights_tags?post=34590"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}