governing the context AI agents retrieve<\/a>. IBM research puts a number on the pressure: 81% of CDOs are already prioritizing AI investments, but only 26% feel confident their data can actually support AI-driven revenue. The gap between ambition and readiness is where agents go wrong.<\/p>\nWhat this means for your business<\/h2>\n
The failure mode this article names is worth taking seriously. An AI agent that writes a supplier evaluation using both a 2023 and a 2026 policy document hasn’t hallucinated, it has executed correctly on unvalidated context. That distinction matters enormously for accountability. If your organization has already deployed agents for budgeting, compliance, or vendor management, the question isn’t whether your models are good enough; it’s whether anyone owns the rules about which data those agents are allowed to treat as authoritative.<\/p>\n
Farmer’s four-capability framework, inventory, access control, governed delivery, and observability, is directionally correct, but the hardest one is access control and it’s the one enterprises tend to underestimate. Application-layer permissions govern what a human user can open inside Salesforce or SharePoint. An agent pulling from both simultaneously can assemble a view that no individual employee would ever have been permitted to see. Most enterprises have not extended their permissioning logic to cover that combination scenario, and most security reviews of AI deployments aren’t asking for it yet. That’s the real exposure for CDOs who are letting departmental teams spin up agents independently, each with different definitions of “authoritative source” and different thresholds for data freshness.<\/p>\n
The semantic layer concept Farmer builds toward, a governed translation layer that tells an agent whether “revenue” means gross, net, or recognition-adjusted, is where the CDO role either becomes genuinely strategic or collapses into a support function. If the CDO owns and enforces that layer across marketing, finance, and operations agents, they control the boundary of what the enterprise is capable of doing with AI. If they don’t, every new agent deployment is a negotiation between siloed definitions, and the CDO finds out about the inconsistency after the compliance report has already been filed. The renewal to weigh here is any enterprise data platform contract that doesn’t yet include semantic layer capabilities, because that gap is about to become visible in the worst possible moment.<\/p>\n
Concept deep-dive: Semantic layer<\/h2>\n
A semantic layer sits between raw enterprise data and the systems that consume it, translating technical data structures into business-defined terms with consistent meaning. Think of it as a contract: when an AI agent asks for “revenue,” the semantic layer specifies exactly which accounting definition applies, rather than letting the agent infer from context. Without it, agents trained on general knowledge substitute their own interpretations of business vocabulary. With it, business logic is explicit, auditable, and consistently applied across every system that queries the data.<\/p>\n
Based on reporting from The CDO’s new role is curating context for data governance<\/a>, originally published 2026-07-29 03:00:00.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"Share with your CDO Agentic AI, systems that retrieve data and execute multi-step tasks without waiting for human sign-off, has exposed a gap that data catalogs and master data management programs were never built to close. Donald Farmer argues that the CDO’s job is now less about stewarding data assets and more about governing the […]<\/p>\n","protected":false},"author":1,"featured_media":8240,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[237],"tmauthors":[],"class_list":["post-8239","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ai-data","tag-cdo"],"_links":{"self":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts\/8239","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/comments?post=8239"}],"version-history":[{"count":0,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts\/8239\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/media\/8240"}],"wp:attachment":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/media?parent=8239"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/categories?post=8239"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/tags?post=8239"},{"taxonomy":"tmauthors","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/tmauthors?post=8239"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}