AI in the People Business: HR's New Toolkit

Human resources is being handed AI tools as fast as any function in the company — screening assistants, drafting tools, chatbots for policy questions, analytics over the whole workforce. Some of that is genuinely useful, and HR professionals drowning in administration deserve the relief. But HR is the people business: its raw material is livelihoods, dignity, and trust, and its mistakes land on individuals by name. That makes HR the function where the line between tool helps and human decides matters most — and where drawing it clearly is the real skill of the next few years.

Where the machine genuinely helps

Start with the work that is voluminous, structured, and checkable. First drafts: job postings, policy summaries, onboarding materials, the hundredth version of a routine letter — drafted by the tool, finished and owned by a person. Logistics: scheduling, reminders, routing questions to the right human. Retrieval: answering routine policy questions strictly from the company’s own approved documents, with the source shown. Synthesis: turning hundreds of survey comments or exit-interview notes into themes a person then reads carefully. In each case the pattern is the same — the tool handles volume, a professional handles judgment, and the output is something a named human has reviewed and stands behind.

Where a person must decide

Then draw the other line, and draw it before a vendor demo draws it for you. Who gets hired, promoted, disciplined, or let go. How an accommodation request is handled. What happens with a complaint about a manager. These decisions can be informed by tools — organized information, consistent criteria, a second read — but they cannot be delegated to one, for reasons that are practical as much as principled. A model trained on a company’s past decisions can quietly reproduce the biases inside them. An affected employee is owed an explanation a person can actually give. And when a decision is challenged, “the system recommended it” is not an answer anyone can defend — to the employee, to leadership, or beyond.

Say openly where the tool is in the room

One practice ties the two lines together: disclosure. People affected by HR processes — candidates, employees answering a survey, staff whose questions go to a chatbot — deserve to know when a tool is involved and how to reach a human instead. Secrecy is what turns a reasonable efficiency into a scandal when it surfaces, as it always surfaces. Disclosure also improves the work: a tool everyone knows about gets its errors reported, while a tool nobody mentions gets its errors quietly absorbed by the people it was used on. The same applies inside the HR team itself. Specialists should be able to say which parts of a document the tool drafted and which parts they wrote or changed — not as a confession, but as ordinary professional provenance, like citing a source. A function that can describe its own process that clearly is a function that can be trusted with it.

The working rule that holds up: the closer a task sits to a consequential decision about an identifiable person, the more human the process must be — with the tool, if it is used at all, in a supporting and reviewable role. Exactly where a given company’s lines must sit depends on its industry, its jurisdictions, and its obligations, and those are questions to settle deliberately with the company’s own legal and compliance advisors — not defaults to inherit from a software package. This article is a way of thinking about the toolkit, not advice about any company’s rules. The HR teams that get this right will not be the ones with the most automation. They will be the ones who can say, for every use, what the tool does, what the person does, and who answers for the result.

← All Insights