Skip to content
PerslacePerslace Help Centre

Prompting and data-safety guidance

A useful Perslace Intelligence prompt identifies the question, scope, period, and comparison you need. Use the minimum information required, then verify the response against authoritative Perslace records.

Who can use this guide

Section titled “Who can use this guide”

This guide applies to everyone with verified access to Perslace Intelligence.

Before you prompt

Section titled “Before you prompt”
  1. Confirm that your role is authorised to review the subject.
  2. Define one question, one scope, and one period.
  3. Decide which scheme, report, or source record you will use to verify the answer.
  4. Remove names, email addresses, employee IDs, client names, transaction IDs, and exact financial values unless your organisation has explicitly approved their use.
  5. Never enter passwords, one-time codes, API keys, tokens, connection strings, raw transaction exports, bank or payroll details, or client-confidential information.

Write a focused prompt

Section titled “Write a focused prompt”

Use a structure like this:

For the selected group and period, identify the factors I should verify when comparing compensation share with revenue contribution. State the scope, period, currency, and any limitations. Do not recommend an operational change.

  1. Name the business question, not a desired conclusion.
  2. State the scope, such as organisation, branch, group, or individual, only when that scope is available and approved.
  3. State the period.
  4. Ask the response to identify its scope, currency, assumptions, and limitations.
  5. Ask for checks or evidence, not an automatic decision.
  6. Submit the prompt only after completing the data-safety review above.

Verify the response

Section titled “Verify the response”
  1. Confirm the returned scope, period, scheme context, and currency.
  2. Compare material statements with the authoritative scheme, transaction data, Dashboard, and report.
  3. Test a representative, sanitised sample where appropriate.
  4. Record the human reviewer, evidence, and decision.
  5. Do not apply a scheme, payout, data, or operational change from the response alone.

Expected result

Section titled “Expected result”

The prompt is focused and minimally sensitive, and the response is recorded as an AI-assisted finding with a separate human verification.

Important notes

Section titled “Important notes”
  • The interface advises users to verify critical information.
  • Product review found plan-state and currency inconsistencies in AI output.
  • Company Knowledge Memory and related guidance fields are sensitive configuration. Their governance, retention, and access rules require organisational approval before confidential content is added.
  • Perslace Intelligence, Copilot, and Strider AI wording was inconsistent in the reviewed interface. AI Impact Simulator (AIS) refers specifically to the historical scheme-impact simulator.

Troubleshooting

Section titled “Troubleshooting”

The answer is too broad

Section titled “The answer is too broad”

Restate one question with a specific scope and period. Ask the response to list limitations.

The answer includes the wrong plan or currency

Section titled “The answer includes the wrong plan or currency”

Do not act on it. Compare the result with authoritative records and follow the AI troubleshooting guide.

Sensitive information was entered

Section titled “Sensitive information was entered”

Stop using the prompt content, follow your organisation’s incident and data-governance process, and contact the authorised security owner. Do not repeat the sensitive content in a support request.

Section titled “Related articles”

Last updated: August 2026