Challenge:
Adobe’s Learn organization was expanding AI-powered customer experiences, but inconsistent terminology, fragmented content structures, and disconnected metadata limited content findability, reuse, governance, and AI accuracy. The gap became impossible to ignore when Adobe’s first public-facing AI chatbot began returning inaccurate answers and routing customers to the wrong products. Leadership traced the failure back to an AI experience that wasn’t drawing on Adobe’s governed taxonomy data at all — and asked why not. That moment reframed taxonomy from a back-office, content-tagging function into a strategic input for trustworthy AI, and created the executive mandate for the work below.
Solution:
Adobe’s Oxford Taxonomy as a Service (TaaS) already operated a suite of services — a taxonomy API, an auto-tagging service with human curation, a resource portal, and an emerging Taxonomy MCP — but these were still functioning as tagging infrastructure rather than an enterprise-wide semantic layer. Factor partnered with the Oxford TaaS team to establish a governed enterprise information architecture connecting business language, taxonomy, metadata, content, and AI services.
Outcomes
- Improved AI response quality by 78% using governed taxonomy data, validated through structured testing by Adobe’s engineering team
- Piloted AI-assisted taxonomy development to roughly double the taxonomy’s size in support of 25 new products, while reducing staff effort by 90% and building the rubric-driven guardrails needed to keep AI-generated suggestions governed
- Secured executive sponsorship and budget to migrate the taxonomy from spreadsheet-based tables into an enterprise knowledge graph, unlocking deterministic inference and web-scale content authority
- Created a scalable information foundation for enterprise search, personalization, automation, and future AI initiatives
