What is safe to automate in HR, and what is not?
Safe: policy questions, onboarding documentation, job description drafting, and the shared-services queue that consumes an HR team's week. Not safe: screening decisions, performance ratings, or anything that ranks people. Those carry discrimination exposure and, in a growing number of jurisdictions, audit requirements. We build the first and decline to automate the second.
- Improve operational efficiency
- Reduce cost
Human Resources
Start with the queue, not the decisions
An HR team’s week goes to the same questions repeated: how much leave do I have, what is the policy on this, where is that form, what happens to my benefits if. Every one of those has an answer sitting in a document nobody can find.
A policy assistant grounded in your actual handbook, citing the clause it drew from, removes most of that volume and makes the answers consistent. Consistency matters more here than speed: different answers to the same policy question across two employees is how disputes start.
The line we hold
We do not build systems that screen candidates, rank employees, or generate performance ratings. Two reasons, and both are practical rather than philosophical.
The exposure is real and growing. New York City’s Local Law 144 already requires bias audits for automated employment decision tools, several states have followed, and the EU AI Act classes employment decisions as high risk. If you want that capability, you need employment counsel and a bias-audit vendor, and we will point you to both.
The second reason is that it does not work well enough to bet a discrimination claim on. A model trained on your historical hiring reproduces your historical hiring, including the parts you are trying to change.
Drafting a job description is different, and that we will build.
What we measure
Policy question turnaround, shared-services ticket volume and handle time, and time to productivity for new hires. Track first-year attrition alongside, since onboarding quality shows up there first.
Typical projects
- A policy assistant that answers employee questions from your actual handbook, citing the clause
- Onboarding documentation and checklists generated per role and location
- HR shared-services queue triage, with sensitive categories routed straight to a human
- Job description drafting from existing roles and levelling frameworks
- Knowledge capture from departing employees before the notice period ends
The same work, by industry
How it starts
A three-week assessment scoped to this area: where the hours actually go, what is worth building, and what to fix first. You keep the findings whether or not there is a next part.