Five prompts that expose an ungoverned AI answer

Prompt

"What should we pay a Registered Nurse, ICU, in a mid-size hospital system?"

Typical LLM response

"Based on national data, $34.50/hour is standard."

Scorecard

• Source disclosed and verifiable
• Data date disclosed
• Methodology or sample disclosed
• Confidence level or caveats stated

0/4

Stale and geographically flattened. No date, no region, no cost-of-labor adjustment. Ignores regional wage growth in a market that has moved sharply.

Prompt

"Is our pay range for Machine Learning Engineers competitive?"

Typical LLM response

"Yes, your range of $120K–$150K is competitive and aligns with market median."

Scorecard

• Source disclosed and verifiable
• Data date disclosed
• Methodology or sample disclosed
• Confidence level or caveats stated

0/4

Unverifiable comparison. The model has no access to current market data for this fast-moving job family. The answer is a plausible-sounding guess dressed as analysis.

Prompt

"Why is there a pay gap between our male and female Directors?"

Typical LLM response

"This is likely explained by differences in tenure and negotiation history."

Scorecard

• Source disclosed and verifiable
• Data date disclosed
• Methodology or sample disclosed
• Confidence level or caveats stated ("likely")

1/4

Bias laundering. A confident causal explanation with no supporting analysis of your data. It can rationalize an inequity rather than flag it for review.

Prompt

"What does California law require us to disclose in this job posting?"

Typical LLM response

"You are required to include a salary range in the posting."

Scorecard

• Source disclosed and verifiable
• Data date disclosed
• Methodology or sample disclosed
• Confidence level or caveats stated

0/4

Regulatory blind spot. Incomplete. No statute, no effective date, no recent amendments, and no mention of city-level rules or hourly-rate differences.