AI Agents for Repertoire Research, Claims and Ownership Conflict Resolution

In short: When a claim arrives, repertoire teams must research whether the claimed recording or work is the same asset already in the repertoire, and whether the claim is valid. Noctil runs AI agents through ProcessEngine to do that research. Agents gather evidence from internal repertoire data through retrieval over a vector-enabled source, and from external sources through tools. They compare the assets, enrich missing metadata, and present a recommendation with its evidence. Evaluations act as quality gates: claims that score above a configurable threshold can be resolved automatically, and the rest go to specialists for review. Continuous improvement loops raise accuracy over time. All of it runs on a scalable, resilient, secure and cost-optimized platform.

Let AI agents do the research, so your specialists can make the call. For repertoire managers and claims specialists at rights organizations, publishers, labels and administrators who audit sound recording and musical work data and validate incoming claims.

The Situation

Repertoire managers and claims specialists keep a catalog’s sound recording and musical work data accurate, and resolve the claims and ownership conflicts that arise around it. What they handle depends on where they sit. Rights organizations receive members’ claims and run dispute processes. Music publishers and administrators handle counterclaims and overclaims raised by collecting societies when the shares claimed on a work exceed 100%. Record labels handle ownership conflicts raised by neighbouring rights societies and video platforms when two parties claim the same recording.

For each claim or conflict, they have to establish whether the assets involved are the same, whether each party’s information is correct, and what evidence supports their position, often before a response deadline. The outcome decides who gets paid.

The Problem

Claim validation is skilled, slow and hard to scale:

• Research-heavy. Each claim means searching internal records, registrations, agreements and external sources, then comparing titles, writers, performers, identifiers and durations.
• Is it the same asset? Versions, covers, remixes, alternate titles and missing identifiers make it hard to tell whether two records describe the same recording or work.
• Incomplete data. Claims and existing records often lack the details needed to decide, so specialists have to find and add them first.
• Growing backlogs. Claim volumes rise faster than teams can grow, and unresolved claims hold up distributions. Counterclaims and dispute cycles also come with response deadlines, and a missed deadline can mean losing the claim.
• Consistency and audit. Different specialists can reach different conclusions, and decisions need a clear record of the evidence behind them.
• Trust in AI. Teams want automation, but they need to know AI outputs are accurate, explainable and secure before relying on them.

The Solution

Noctil runs agent-based research and validation workflows on the Noctil Platform, with the organization in control of what is automated and what is reviewed.

1. AI agents orchestrated by ProcessEngine. When a claim arrives, ProcessEngine starts an agent workflow that plans the research, gathers evidence and prepares a recommendation. It works with leading AI models and Noctil’s own models trained for music rights tasks.
2. Internal knowledge through retrieval (RAG). Agents search the organization’s own repertoire, registrations, agreements and past decisions through a vector-enabled source, so their research is grounded in the organization’s data.
3. External research through tools. Agents use tools to retrieve information from approved external sources, such as society databases, DSP catalogs, identifier registries, and public music data sets like MusicBrainz and Discogs, filling gaps the internal data can’t answer.
4. Same-asset comparison with MatchIQ. MatchIQ compares the claimed asset with existing records, across titles, writers, performers, identifiers and durations, to show whether they are the same recording or work.
5. Metadata enrichment. Missing details found during research are proposed as enrichments to the record.
6. Evidence-backed recommendations. Each recommendation shows the sources and reasoning behind it, so specialists can review it quickly and see why. For counterclaims and ownership conflicts, agents also assemble the evidence needed to respond, such as agreements, split sheets, letters of relinquishment and copyright registrations, and ProcessEngine tracks each response deadline.
7. Evaluations as quality gates. Every workflow is measured against evaluations, and each recommendation gets a confidence score. The organization sets the threshold: results above it can be resolved automatically, and results below it are routed to a specialist for review.
8. Continuous improvement loops. Specialist decisions and evaluation results feed back into prompts, models and rules, so accuracy improves with every claim processed.
9. Enterprise platform. Workflows run on infrastructure built for scale and resilience, with security controls on data access and model usage, client data never used to train Noctil’s models, the option of a private model per client, and cost optimization that selects the right model for each task.

The Result

• Faster claim resolution, with research done by agents and specialists focusing on decisions
• Consistent, explainable outcomes, with every recommendation backed by its evidence
• Richer repertoire data, as gaps found during research are filled
• Trusted automation, with configurable thresholds deciding what is resolved automatically and what goes to a specialist
• Accuracy that improves over time through continuous feedback loops
• Secure, scalable and cost-efficient operation as claim volumes grow