Check an AI report against the source numbers
Use a fictional, aggregated table, define each metric, and ask AI to show its calculations. Verify the totals and denominator yourself before relying on the explanation.
How to try it
Define the measure
For a practice quote-to-bind rate, say exactly which quotes and binds belong to the same cohort and period. A label alone is not a definition.
Use a simple test table
Fictional cohort: 40 quotes and 10 binds. The practice rate is 10 divided by 40, or 25%. This is an arithmetic example, not an agency benchmark.
Ask for evidence
Require the numerator, denominator, formula, and source row for every calculated metric. Ask AI to flag missing or inconsistent data.
Separate observation from explanation
A change in a rate does not establish its cause. Ask for possible questions to investigate, rather than a confident diagnosis.
Copy the starter prompt
Analyze this fictional practice cohort: 40 quotes, of which 10 became binds. Define quote-to-bind rate as binds from this cohort divided by quotes in this cohort. Show numerator, denominator, formula, and result. Flag missing context. Separate arithmetic observations from hypotheses. Do not invent causes, benchmarks, revenue, or forecasts.
What to review
Check arithmetic against the source table and confirm the cohort definition. Do not upload real client or employee records. Use approved aggregated data only when your agency permits it.
Adjust for your experience
- Novice: verify the single 25% example manually.
- Casual user: compare two fictional cohorts using the same definition.
- Power user: document metric definitions and test a reusable report against known answers.