Cinnova Builds an AI Knowledge Base to Unify Tappi’s Decentralized Documentation

Client: Tappi
Industry: Music Technology / Creator Economy
Solution: An AI Knowledge Base built on Claude, delivering citation-backed answers with role-based access

The Challenge

Tappi’s internal staff, artists, and partners each needed answers about the platform, but no single trustworthy source existed. Information was spread across Jira, three separate application repositories, and a mix of assorted documents — with no reliable way to tell whether any given answer was current, still planned, or already superseded.

The result: people spent time manually cross-referencing multiple systems just to find out whether something was true today, rather than getting straight to the answer — or the work.

The Solution

Cinnova built an AI Knowledge Base powered by Claude that ingests Tappi’s real source material — the Jira epic export, all three application repositories, and curated reference documentation — and answers natural-language questions with:

• Citations back to the source material for every answer
• A status flag on each answer — live, planned, pilot, or superseded — so users know how much to trust it
• Audience-scoped delivery across three portals: Internal, Artist, and Partner
• An AI-guided, 10-stage onboarding flow for new artists
• A public, crawlable Help surface for external users

All three portals draw from one governed content base, but access boundaries between them are enforced at the database query layer — not just hidden in the UI — so each audience only ever sees what it’s meant to see.

The system is deployed and live in production.

Early Results

Support tickets and manual documentation lookups dropped from 800 to 150 per week — an 81% reduction since the AI Knowledge Base went live.

Average time-to-answer dropped from roughly 10 minutes of manual searching across Jira, repos, and docs to 10–15 seconds with the AI Knowledge Base — a ~97–98% reduction in time-to-answer.

The system answered 324 questions from staff, artists, and partners in its first week of production.
Answer accuracy: measured against a 16-question benchmark set on 2026-08-20, hand-graded against a fixed rubric, the system answered 14 of 16 correctly (87.5%), against a scope target of 80%.

What’s Next

Cinnova and Tappi are continuing to expand the platform’s content coverage across the Artist and Partner portals, building on its live production deployment.

“Cinnova’s AI Knowledge Base showed us how AI can make Tappi’s growing body of product and operating knowledge much easier to use without sacrificing source transparency or governance. Being able to see where an answer came from—and whether it reflects something live, planned, or superseded—is critical for us. This gives Tappi a strong foundation for improving support, onboarding, and partner enablement as we scale.”

— Matt Kovelan, CEO, Tappi