Practical guide
Review an Agent Kay answer with its evidence
Use AI assistance to investigate, with a person responsible for the decision.
Ask a bounded question
Specify the business question, company scope, period and measure. Agent Kay capabilities depend on the enabled workflow and available records. A missing dataset or permission cannot be replaced by a confident answer.
Inspect the supporting records
For evidence-backed workflows, open the referenced records and compare dates, currency, values and statuses. Check whether the evidence is complete enough for the question. The finance workflow checks returned evidence references against the supplied set, but that does not guarantee that every interpretation is correct.
Separate observation from explanation
A movement in receivables is observable; the reason a customer has not paid may need a conversation or another source. Treat suggested explanations as items to verify. Ask what information is missing and look for contrary evidence.
Keep authority with the reviewer
Confirm permissions, source freshness and the approved process before acting. A suggested follow-up is not an executed payment, message or employee decision. Public demonstrations are scripted synthetic examples and do not call an AI model.