United StatesEducation2025 – ongoingGuild
Catalog Atlas, an internal AI agent for program and provider questions
Guild's internal AI agent cut manual QA checks by more than 90% and saved an estimated 450 hours a year
Problem
Guild's information on education programs and provider outcomes sat in dashboards, spreadsheets, wikis and Slack threads. Lookups could take hours, and employees sometimes waited up to 24 hours for a Slack reply. Subject-matter experts kept answering the same questions. Generic chatbots missed the company's business rules and provider aliases.
Approach
Guild built Catalog Atlas, a domain-specific agent, on Databricks AI and Agent Bricks. Structured and unstructured program data was unified under Unity Catalog, and alias mappings and business rules were added, for example to tell Cornell from eCornell. Automated AI judges in MLflow scored answers to replace manual checks. Employees ask questions through a chat interface with single sign-on.
Results
Takeaway
Guild encoded its own naming rules and data into the agent and automated the answer checks, so experts spent less time on repeat questions. The 450-hour figure is an estimate.
Categories
- Industry
- Education
- Business function
- Internal work and knowledge management
- Innovation type
- Process innovation
- AI technology
- AI agentsGenerative AI
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Databricks describes how Guild built Catalog Atlas, an AI agent that answers staff questions about education programs and providers. It reports an estimated 450 hours saved a year, worth $33,750, and a drop of more than 90% in manual QA checks. A Guild executive is quoted saying requests that took hours or a day now resolve in thirty seconds or less.
Full details are in the original sources.