> ## Content Index > Fetch the complete content index at: https://foxhole-dispatches-from-the-saaspocalypse-frontlines.ghost.io/llms.txt > Use this file to discover other available public pages before exploring further. # AI's favourite department is sitting on its least AI-ready dataset. - URL: https://foxhole-dispatches-from-the-saaspocalypse-frontlines.ghost.io/ais-favourite-department-is-sitting-on-its-least-ai-ready-dataset/ - Published: 2026-08-03T04:02:10.000Z - Updated: 2026-08-03T04:02:10.000Z - Author: Warwick Boulter *Why the fastest AI rollout in the enterprise is running directly on top of its most ungoverned data — and what that means for every workflow downstream of it.* Marketing didn't just get AI early. It got AI *first and fastest:* more concentrated than almost any other function in the enterprise. Recent industry surveys put marketing as the second most common function for AI agent deployment, behind only customer service, and generative AI adoption within marketing teams has moved from a minority practice to near-universal in under two years — a faster shift than any prior marketing technology category managed. Among marketers already using generative AI, creative development — producing images, video and campaign content — is the single most common use case, well ahead of anything else. By any normal measure, marketing is the AI success story inside the modern enterprise. Here's the problem nobody's naming: almost all of that investment points in one direction — generation. Drafting more. Producing more. Personalising more, faster than any team could manually. What it does not do, in most organisations, is make the content marketing *already owns* any more visible to a machine than it was two years ago. **The asset library was never built to be read by a machine** Rich media — the images, video, and creative files that make up the majority of unstructured data in most large organisations — doesn't declare what it contains. An image file doesn't tell a governance system who's allowed to see it, whether a licence has expired, or whether consent was ever captured. A video doesn't surface its own topics or speakers. That content sits in shared drives and folder structures, discoverable only by the person who remembers where they put it. Traditional DAM platforms were built to manage that problem for a single team — helping marketing store and retrieve its own creative files inside a UI. That's a real, narrow problem, and most DAMs solve it well enough. But they were architected as closed, self-contained systems for humans clicking through an interface, not as a data layer for machines to query at scale. The moment an asset leaves that system, its governance goes with it. That was an acceptable limitation when the only consumer was a marketing coordinator. It stops being acceptable the moment the consumer is a generative AI tool, an agent assembling a campaign, or a compliance process that needs to know whether an asset is still licensed for use. **The governance gap is compounding, not closing** This isn't a hypothetical risk sitting quietly in the background. It's compounding in real time, and the data on AI governance maturity backs that up: across recent industry research, a large majority of organisations using AI in marketing report having no clear, organisation-wide policy governing how it's used, and only a small fraction have matured their AI governance capability at all. Adoption has sprinted ahead. Governance hasn't caught up — and rich media, being the least structured data category most organisations hold, is exactly where that gap shows up first. **What "AI-ready" actually requires** AI readiness for rich media isn't a feature you switch on. It's a state that has to be built, and it requires four things to be true simultaneously: - **Understood** — every asset processed and described in structured, queryable metadata, not left to whatever a human typed in at upload. - **Contextualised** — provenance, purpose and relationships documented, so a file becomes intelligence rather than just a file. - **Governed** — access, rights, consent and retention enforced as policy, automatically, not tracked in a spreadsheet somewhere. - **Connected** — the asset accessible via API, with its governance metadata travelling with it wherever it's used next. Most organisations can't confirm any of these are true of their rich media estate. Which means the generative AI tools marketing has adopted fastest are operating with no visibility into the one dataset they're supposed to be building on. **Where this actually leaves marketing** This isn't an argument against AI in marketing — the opposite. It's an argument that generation without a governed foundation underneath it is building on sand. An AI agent assembling a brief can't use an image library it can't search by content. A personalisation engine can't reliably select assets it hasn't been told are current, licensed, or on-brand. And every asset generated on top of an ungoverned library just adds to the pile the next AI initiative will have to work around. The fix isn't slower AI adoption in marketing. It's closing the gap between how fast generation moved and how far governance has been left behind — ingesting rich media into a controlled environment, enriching it automatically so it's legible without manual tagging, governing it as policy rather than folder structure, and connecting it to whatever needs it next, human or agent. Marketing didn't do anything wrong by moving first. It just moved first without anyone making sure the ground underneath it could hold the weight.