01
Separate essential criteria from preferences
Write down the outcomes the role must deliver. Keep genuine requirements separate from useful extras such as a particular tool or industry background. Avoid adding criteria after seeing a candidate's résumé.
02
Score evidence consistently
Use a small rubric: evidence shown, evidence unclear or requirement not met. Record a short example for each decision. Read in batches using the same criteria, and flag missing information for clarification rather than guessing.
03
Use automation for administration, not final decisions
A workflow can organize applications, prepare interview notes or draft acknowledgements. Do not assume a general-purpose AI tool is a validated screening system. Check errors, restrict access to applicant data and have a human review consequential decisions.
04
Close the loop with candidates
Explain the next step and expected timeline. Send respectful updates, keep only the information you need and follow applicable retention requirements. Revisit the rubric if qualified applicants are consistently missed.