Skip to content

SKILL.md

Latest commit

 

History

History
64 lines (49 loc) · 4.22 KB

SKILL.md

File metadata and controls

64 lines (49 loc) · 4.22 KB
Raw
name hr-ai-evaluation
description Help HR technology teams evaluate AI tools and vendors for HR use cases against accuracy, bias, and fit-for-purpose criteria before adoption. Use when asked to evaluate an AI vendor for HR, build an AI tool evaluation framework, assess this AI tool for bias risk, compare AI vendors for [HR use case], or pilot an AI tool before rolling it out.
metadata
author version
Tuan Duc Tran
1.0.1

AI tool evaluation for HR

Evaluate AI tools and vendors being considered for HR use cases — screening, sourcing, chatbots, analytics — against accuracy, bias, transparency, and fit-for-purpose criteria before adoption.

Supported tasks

  • Building an AI vendor evaluation framework tailored to HR use cases
  • Assessing AI tools for accuracy and reliability claims against real evidence
  • Evaluating AI tools for bias risk and disparate impact potential
  • Comparing multiple AI vendors for the same HR use case
  • Designing a pilot program to test an AI tool before full rollout
  • Assessing vendor transparency around model training data and methodology
  • Evaluating data privacy and security implications of an AI vendor
  • Reviewing AI vendor claims critically rather than taking marketing at face value
  • Building evaluation scorecards for AI tool procurement decisions
  • Assessing integration feasibility of an AI tool with existing HR systems
  • Documenting AI evaluation decisions for audit and compliance purposes
  • Re-evaluating existing AI tools periodically as they update or as regulations shift

Key prompts

Building the framework

  1. "Build an AI vendor evaluation framework for [use case, e.g. resume screening, interview scheduling, chatbot] covering accuracy, bias, transparency, and cost."
  2. "What questions should we ask an AI vendor about their model's training data and bias testing before considering adoption?"
  3. "Design an evaluation scorecard to compare multiple AI vendors for [HR use case] on a consistent basis."
  4. "What red flags in a vendor demo or sales pitch should make us slow down and dig deeper before proceeding?"

Assessing risk

  1. "What bias risks should we specifically evaluate for an AI tool used in [screening/sourcing/performance assessment]?"
  2. "Critically assess this vendor's accuracy and fairness claims — what evidence would we need to actually validate them?"
  3. "What data privacy and security questions should we ask before allowing this AI tool access to employee or candidate data?"
  4. "What legal or regulatory review should this AI tool go through before we allow it to influence [hiring/performance] decisions?"

Piloting and deciding

  1. "Design a pilot program to test [AI tool] on a limited scale before full rollout, including success criteria."
  2. "Compare [Vendor A] and [Vendor B] for [HR use case] against our evaluation framework and recommend an approach."
  3. "How feasible is integrating [AI tool] with our existing [ATS/HRIS], and what are the risks of a poor integration?"
  4. "What would trigger us to pause or roll back a pilot of [AI tool] before it reaches full rollout?"

Ongoing governance

  1. "Document our evaluation decision and rationale for [AI tool] for audit and compliance purposes."
  2. "How often should we re-evaluate [AI tool] as the vendor updates the model or as relevant regulation changes?"
  3. "Who owns ongoing accountability for [AI tool] performance once it moves from pilot into standard operations?"
  4. "Design an offboarding plan for retiring [AI tool] if a re-evaluation determines it no longer meets our standards."

Tips

  • Ask vendors for evidence, not just claims — request bias testing methodology and results rather than accepting marketing language about "fairness" at face value.
  • Pilot before scaling; a small controlled test surfaces real-world issues that a sales demo never will.
  • Evaluate the specific use case, not the tool in the abstract — an AI tool that's fine for scheduling may carry very different risk when used for screening or assessment.
  • Involve legal and DEI stakeholders in evaluation, not just IT and procurement — bias and compliance risk in HR AI tools is a cross-functional concern.
  • Re-evaluate periodically; vendors update models and regulations evolve, so an approval from a year ago may no longer hold.