Enterprise Knowledge Base Management System (Deployed — AI Solutions)

Window: 15 apps • 2,000+ docs • 8-week rollout
Baseline → After: TTA ~15 min → ~2 min (≈85% faster)
Method: Time-boxed lookups • sampled tickets • SME spot checks

Deployed enterprise Knowledge Base Management System that ingests and structures content for 15 applications, publishes curated packs to NotebookLM, and cut time to answer (TTA) from 15 minutes to 2 minutes—saving 150+ hours/month.

Role: Lead DeveloperTeam: 1Duration: 6 weeks
Python
parsers
schedulers
JSON metadata
NotebookLM
Docker

Problem

Docs lived across vendors, SOPs, and old tickets; agents spent 15 minutes hunting per question.

Approach

Python scrapers/API pulls and file watchers convert sources to Markdown/HTML, tag metadata (app, version, audience, last-reviewed), chunk for retrieval, and publish curated packs to NotebookLM. Prompt catalog standardized queries.

Results

Time-to-answer dropped 15 → 2 minutes across 15 apps. Estimated 150+ hours/month returned to IT. Relevance increased after prompt-catalog rollout; curated packs keep drift low.

Architecture

Ingestion → Normalize → Metadata/Chunk → Deterministic folder/schema + JSON sidecars → Scheduled publishing → Prompt catalog

  • Python scrapers and API pulls
  • File watchers and change detection
  • Metadata tagging and normalization
  • NotebookLM publishing pipeline
  • Prompt catalog and standardization

Key Challenges

  • Normalization across vendors
  • Metadata freshness
  • Change-detection and safe updates

Lessons Learned

  • Schema discipline compounds
  • Prompt standardization lifts answer quality

Explore the Project

Interested in similar results?

Let's discuss how we can implement AI automation for your team.