Start with a specific question
Choose a question with a clear reporting period, business entity and reference answer. For example, compare an operating metric across two periods or find a particular rule in a set of public documents. Keep the first test focused on a manageable set of systems and topics.
Confirm the data and access boundaries in advance
Match the test data to the question. Specify the visible fields, user roles and update scope. The public demo uses synthetic data. For a real PoC, the responsible teams agree on the material, permissions and environment before testing begins.
Test how the system handles questions it cannot answer
Follow-up questions test whether the context stays clear. Questions involving missing data or access outside a user’s role test the boundaries. Check significant results against sources and reference answers, beyond recording a response that reads well.
- For an ordinary question, check the result, units and period.
- For a follow-up, retain the conditions and state each change.
- For a boundary case, explain the gap and identify a sensible next step.
Record the result for the next decision
Mark each test as passed, partially passed or failed, and record the reason and remaining work. The demo length can be adjusted to match preparation. Agree separately on the scope of data connection, cleaning, authorization and production delivery.
Continue with a local requirements brief
After choosing a scenario and requirements, copy or download a brief that contains no personal information. Use it for your own review and later discussions. Completing the page does not create a booking or service commitment.
