We already have access to an AI tool at work. Most of us have tried it. Most of us walked away unimpressed.
Not because it can’t write. Because it doesn’t know anything. It doesn’t know our reporting lines, our comp philosophy, which spreadsheet is the real one, or how our company sounds when it talks to employees. So we get generically reasonable HR advice from the internet, and we rewrite all of it, and we conclude the tool isn’t there yet.
The tool is fine. It’s missing context, and nobody hands us that part. Every guide about setting up an AI assistant is written by a consultant or a founder whose whole world is a calendar and a client list.
Ours is a company full of humans with salaries and managers and problems.
So I wrote the version for us. It’s free. It takes about twenty minutes and setup is a conversation, not a questionnaire.
What’s in it
Start Here
The five-step setup, a menu of first tasks worth trying, and what to do when the output isn’t right yet.
The People Partner Instructions
Tells your AI how to work with a People practitioner: when to push back, how to handle employee and manager comms, how to treat numbers headed for a leadership team, and when to hand the decision back.
The Context Builder
Interviews you one question at a time about your role, your org, and how People decisions actually get made where you work. Then turns it into context you keep and reuse.
The AI for HR Safety Guide
What never goes in, how to do the same task with less identifying information, what changes when your AI is connected to company systems, and what to say when someone asks you to cross the line.
How it goes
- Download the four files.
- Open the AI tool your organization already approves. ChatGPT, Claude, Gemini, Copilot, or something internal.
- Paste in the People Partner Instructions.
- Paste in the Context Builder and talk to it like a colleague. Out loud, if your tool has a microphone. Most of us say more than we’d type.
- Let it build your first version.
- Use it on one real, low-risk task.
Read the Safety Guide before that last step, not after.
Set it up once and it’s there every morning. Start Here walks through putting this where your tool keeps things permanently, so your People Partner is waiting in every new conversation instead of vanishing when you close the chat. That distinction trips up nearly everyone, so it gets its own section.
The part I care most about
We are the function that holds the most sensitive information in the company. Being early on a tool doesn’t change that. It raises the stakes. So the safety guide isn’t a disclaimer at the back, it’s a quarter of the product.
Where I don’t let it near
- Investigations. Not the intake, not the notes, not “help me think through this.”
- Health, accommodation, and leave details tied to a person. That was given to me in confidence by somebody who had no choice but to trust me.
- Performance decisions. Drafting a conversation guide, fine. Deciding whether someone goes on a plan, never.
- Anything that could become evidence. Terminations in progress. Complaints.
- Named individuals’ personal circumstances. The divorce, the sick parent, the reason they’ve been off.
This isn’t legal advice and I’m not a lawyer. It’s where I draw my own line and why. Yours might sit somewhere else. The point is to draw it deliberately, before you’re tired on a Tuesday and it’s convenient not to.
Nothing here asks you to upload company documents anywhere, connect your HRIS, or open a new AI account. Your information stays inside the environment you’re already authorized to use at work.
Who wrote this
I have a theatre degree. I got into tech sideways and spent thirteen years going from entry level to VP of People, mostly by being the person who’d figure it out. I led people through the pandemic while barely holding myself together, and burned out in the way that’s hard to see from outside, where you’re still shipping and still in every meeting and still the one everybody comes to.
Most of what gets published about AI and HR is written by people selling HR software or people who have never run a People function. I’ve done the job. This is what I built for myself, it works, and there’s no reason for it to be a secret.
“I know AI could save me time. I just don’t have time to figure out how to make it actually useful.”
That’s the thought this is for. Solo HR practitioners, People leaders, HR consultants, and anyone effectively doing five jobs under one HR title.
Take it
Free. Your first version will be incomplete, and that’s the point. It gets better as you use it.
Want to know when the next one lands? Follow me. Questions, or you tried it and something didn’t work? hello@kaitlingraves.ca — I read everything.