AI Definitions for HR, Part 2
Follow a policy review to understand tools, APIs, MCP, scheduled tasks, workflows, agents, and harnesses—and where human approval belongs.
Writing
Notes from my HR work and the AI tools I build and test.
About 9 min readUpdated
Explaining Tokens to HR Who Are Too Afraid to Ask
Tokens become less abstract when an allowance or spending limit appears. This guide explains what the meter is counting and how to stop paying for work you will not use.
About 2 min readUpdated
HR should help design the work AI changes
Software developers reporting into HR sounds strange until you look at the work ahead. HR needs to help design what people and AI will each do.
About 3 min readUpdated
What to measure in an HR AI pilot
A source error in my own work changed how I evaluate an AI pilot. Here are the checks I would include, along with the time spent correcting mistakes.
About 2 min readUpdated
A source-verification error in my AI workflow
A shortlink sent my assistant to a plausible but incorrect post. The answer looked fine, so I changed what the workflow counts as a verified source.
About 3 min readUpdated
Testing HRMC with a misleading source
I gave HRMC a made-up leave-law alert that tried to bypass review. It refused the obvious instruction and still exposed a quieter human-review gap.
About 3 min readUpdated
What is agentic AI in HR?
Agentic AI in HR can check approved sources, prepare a brief, route work, and keep a record. It should not make employment decisions.
About 2 min readUpdated
What real human review looks like in an HR AI workflow
A reviewer needs the evidence, time, authority, and a way to stop the next action. These checks show whether the person really has control.
About 2 min readUpdated
Preparing an HRBP intake brief with AI
An HRBP intake should organize the timeline, policy, known evidence, missing facts, urgency, and owner. Employment decisions stay with a person.
About 3 min readUpdated
Keeping the record of an AI-assisted HR task
What I keep with a completed task so someone can check the sources, understand a correction, and see whether the approved action happened.
About 2 min readUpdated
When AI does not belong in an HR workflow
Sometimes the useful answer is to leave AI out. Unsafe data, unclear ownership, weak sources, and irreversible mistakes are good reasons to stop.
What do you need help with?
A few sentences about the task and what you have tried are enough to start. Please do not send sensitive employee information.
Book a conversation ↗Or email mike@mikewinkleradvisory.com