Artificial Intelligence

Genie One hits 95% with governed context

Databricks says Genie One reached 93-95% on 125 insurance questions with agents plus governed Metric Views and Pages.

StartupHub.ai Staff

By StartupHub.ai Staff, reviewed by Daniel Singer, Editor

2 min read

Genie One accuracy test on insurance questions
Credit: Databricks

Databricks tested Genie One on 125 insurance questions and reached 93 to 95% with agents plus governed context, the company said. Curated agents alone scored 83% in the same test, according to the post.

Tables alone scored 64%, the post said.

Before Genie One, four team agents for Finance, Reserving, Underwriting and Claims each cleared 85% on basic questions on their own, the post said. Genie One then had to route across the same tables, 14 PDFs, three look-alike loss-ratio tables and two planted traps, including a vendor note pushing a 55% loss ratio and draft minutes citing $4.1 million versus $5.28 million approved. Microsoft Learn describes Genie as prioritizing curated Genie Agents as authoritative sources, a routing rule Databricks says is not enough when Genie One answers from its own context.

Genie One does not inherit agent instructions when it answers directly, the post said. It used unrelated Pages in 49% of answers with agents alone. Finance saw an 8.2 percentage-point loss-ratio deterioration while Underwriting saw 1.5 points of improvement, with both correct under different definitions.

Definition questions hit 100% with the governed layer, the post said. Document-plus-data questions hit 97% with agents plus the governed layer.

The pattern repeated in retail, where adding the governed layer lifted accuracy from 54% to 82%, the post said. Define each measure once in a certified Metric View and explain scope in a Page.

Mentioned in this story

Sources

Get the free daily brief every morning

Every morning the day's AI news, new AI startups and new funding rounds. Free and takes a few minutes to read, unsubscribe at any time.

Join 64,000 other subscribers.

TopicsDatabricksGenie OneGenie OntologyUnity CatalogMetric ViewsFinanceUnderwriting

StartupHub.ai StaffDaniel SingerWritten by StartupHub.ai Staff, Editorial teamReviewed by Daniel Singer, EditorAbout

StartupHub.ai Staff, Editorial team

The staff writers of StartupHub.ai, ranging from investment analysts to avid AI tool users, early adopters and critical enthusiasts. Backgrounds span engineering, business and the arts. We hold every piece to rigorous standards of research and review.

Daniel Singer, Editor

Daniel Singer is an AI expert who has been writing about, covering and building AI technology since 2018.

More from Daniel Singer

Test your Agent Readiness now

Paste your website. How well ChatGPT, Claude and the other agents can read and use a site, in about a minute.

Watch

The videos behind the news, one flip at a time

The feed mixes the videos from our articles with startup demos and you flip through them like TV channels for free with no signup.