25/10/2026
At DAM New York this month, AI agents were the headline. The real story was metadata.
In his review for DAM News (13 October), Paul Melcher writes that "agents are becoming the main consumers of digital assets". Yet session after session came back to old problems: metadata, governance, rights. One panel cited audits finding that 20 to 30% of live web content on enterprise sites had bypassed DAM governance entirely.
Agents don't fix weak metadata, they amplify it. They "act on exactly what they're given", as Melcher puts it, at scale and at speed.
But "more tags" is not the answer. An agent needs more than a label on a file. It needs relationships: this clip shows this athlete, at this match, under this rights agreement, cleared for these territories until this date. In most media organisations, each of those facts sits in a different system (MAM, DAM, rights database, archive catalogue) with its own identifiers and vocabulary. Context scattered across silos is context an agent cannot use.
So the real work is making meaning explicit and connected: shared identifiers, an ontology the systems agree on, and a layer where an agent can ask "can I use this, here, now?" and get an answer it can trace.
Provenance raises the stakes. From 1 January 2027, California will require large platforms to detect and display provenance data such as C2PA, and to stop stripping it. That information must survive every hop between systems.
Before scoping an agent pilot, try one test: can you trace a single asset's rights, people and provenance across all your systems without opening a spreadsheet?
If agents become the main readers of your catalogue, who is designing the language they read?
Source: Paul Melcher's review of DAM New York 2026, DAM News, 13 October 2026 – https://digitalassetmanagementnews.org/features/dam-new-york-2026-agents-arrive-and-the-old-problems-come-with-them/