Sample data while we build. Names, prices and register numbers are illustrative.
search
Sign in
Methodology

How Bob will find and review every term.

AI will read the policy wording; a person will approve what it finds. Every value will link back to the document, page and section it came from. Today's catalogue is sample data; this process starts once real wording is in.

From document to comparison

Five steps, every time, from the first real policy.

01
Collect
We will gather published wording, disclosure statements and benefit tables from insurer sites.
02
Extract
AI will find each term in our shared vocabulary and record the page, section and its confidence.
03
Review
A person will check new and changed terms. Changed exclusions will need two reviewers.
04
Publish
The new version will go live. Older versions will be kept, never overwritten.
05
Correct
Insurers and readers can report errors. We will respond within five working days.
What the labels mean

Honest about what we know.

check_circleReviewedReviewedA person has checked the value against the wording.
smart_toyAI 94%AI with confidenceExtracted by AI and waiting for review. The percentage is the model’s confidence.
helpUnknownUnknownWe couldn’t find the term in the published documents.
blockNot offeredExcluded or not offeredThe wording says the benefit isn’t covered.
Versions

Your comparison keeps the wording you saw.

When an insurer publishes new wording, Bob adds it as a new version. Saved comparisons stay at the version you saved, and My Bob tells you what changed.

Your own policy usually follows the wording in force when it started. Upload your schedule and Bob will match the right version.

Keep exploring
How we stay fairThe rules every listing follows.OpenDeveloper docsAPI, MCP server and CLI.OpenGuidesPlain-language guides.Open