Healthgrades replaced a 100MB Excel catalog—but the context model did the hard work
The useful lesson in Domo’s case study is not “spreadsheet becomes chatbot.” It is that business aliases, permissions and user acceptance had to move with the data.
Healthgrades replaced an internal, macro-enabled Excel catalog with a governed AI agent on Domo, according to a September 29 case study published by Domo. The workbook was about 100MB, contained hundreds of tabs and held the master list of awards and ratings licensed to hospitals. Domo says staff updated it manually, needed high-performance computers to open it and received only five basic reports from it.[^1]
The headline sounds like a spreadsheet-modernization story. The more useful lesson for analytics teams is that moving the data was not enough. Healthgrades also had to reconstruct the business context embedded in the old tool and in the people who maintained it.
The migration had three separate jobs
First, the data and access path had to move into a governed environment. Domo says role-based access applied from the protected healthcare data through the agent’s answers, while underlying information could remain in platforms such as Snowflake, Databricks or Google Cloud.[^1] This is a vendor description of its architecture, not an independent security audit.
Second, the team built a context model. Domo says its context engine drafted an initial model from a dashboard and App Studio project in about 10 minutes by profiling columns, inspecting aliases and mapping relationships. A Healthgrades product manager then reviewed and edited that model.[^1]
That human edit was not cosmetic. The case study describes an award whose business-facing name had changed while an older code remained in backend data. Healthgrades mapped both terms so the agent could understand the language used by sales and account teams.[^1] A database migration that omits those aliases can preserve every row and still make familiar questions appear to have no answer.
Third, users tested the replacement against their own reference material. Domo says Healthgrades’ head of sales checked agent responses against existing spreadsheets during a roughly 10-minute session. The product manager later added methodology documents, regenerated the context model in under five minutes and queried relationships between a hospital’s award history and star ratings.[^1]
What teams should preserve
A migration plan should therefore inventory more than tables and reports. It should capture business aliases, historical names, methodology documents, access roles and a set of questions whose answers users already trust. Those artifacts become both context for the new agent and an acceptance test for the replacement.
The case study does not publish a formal accuracy score, user count, elapsed migration time or before-and-after operating cost. It also comes from the platform vendor. Practitioners should treat the timings as a reported demonstration, not a general benchmark.[^1]
Even with those limits, the implementation boundary is concrete: replacing a legacy analytics tool requires moving its meaning and review process, not merely putting natural language in front of the same data.
[^1]: Domo, “How to Replace a Legacy Tool with a Governed AI Agent,” September 29, 2026.
sources
comments · 0