Score a Score makes original music and sound: real composers and producers creating work for brands, film, and media. It is a creative business in the fullest sense — and like every creative business at scale, it runs on top of a decidedly uncreative machine: a catalog of works and versions, metadata that determines whether anything can be found and licensed, and production workflows that move briefs, drafts, and deliverables between people.
That machine is the quiet tax on creative companies. Every hour spent tagging tracks, tidying catalog records, or shepherding a production checklist is an hour not spent on the thing clients actually pay for. And the machine grows with the catalog — creative output scales linearly, but the mechanical work around it compounds.
The problem class
Music catalogs live and die by metadata. A track that is mislabeled, thinly described, or inconsistently categorised might as well not exist: nobody finds it, nobody licenses it, and its history is hard to trace. Keeping metadata rich and consistent across a growing catalog is exactly the kind of work humans do badly at volume and resent doing at all. Production workflows have the same shape — many small, well-defined steps, each easy, none optional, all together a drag on the people coordinating them.
What we built
We worked forward-deployed with the Score a Score team: audit the catalog and production flows, build against the real systems, roll out with the people who run them, hand over.
What shipped is a set of custom agents — not an off-the-shelf assistant — connected to the company's systems through the Model Context Protocol (MCP). The agents support catalog and metadata work (drafting, enriching, checking consistency) and production workflows (keeping the mechanical steps moving), while people review and decide. The creative work stays exactly where it belongs: with the humans.
Custom mattered here. A music company's workflows do not match the assumptions of generic agent products — the vocabulary, the systems, the way a brief becomes a track becomes a licensed asset are all specific. Building custom agents on an open protocol meant the automation fits the business instead of the business bending to a tool.
The pattern for creative businesses
There is a clean line to draw in companies like this: AI should never be the artist, and it should always be the librarian. Agents that keep the catalog rich, the metadata consistent, and the production wheels turning make the creative work more valuable — better organised, easier to find, faster to license — without touching how it is made.
Ownership
The agents and the integration layer belong to Score a Score. Because they run on MCP — an open standard — the company can connect new AI tools to the same layer as the landscape evolves, and extend the agents as the catalog and the business grow. No per-seat fees, no dependency on us.
For music companies, rights holders, and catalog owners facing the same mechanical drag, this is the pattern we bring through our AI for music and rights management practice.