What is the 20-60-20 rule for Ai in HR?

What does Ai lead?

The first 20 percent is the work Ai does on its own. Drafting a job description from a short brief. Sorting applications against clear criteria. Summarizing a long policy into a one-page brief. Transcribing a meeting and pulling out the action items.

This work follows a clear pattern, and the output is easy to check. That is the test. If the output can be checked in seconds, Ai can lead. A draft is a draft. Nobody confuses it with a decision.

What does Ai support?

The middle 60 percent is where most HR work lives, and it is where Ai earns its keep. Screening a pile of resumes down to a shortlist, with a human reading the shortlist. Drafting a performance review that the manager edits before it goes anywhere. Pulling the numbers for a turnover report that a human interprets.

In the middle, Ai does the heavy lifting and a human does the thinking. The human reads the output, checks it, and decides what it means. Skipping that check is how teams get burned. The rule only works if the human step actually happens.

What do humans own?

The last 20 percent stays human. Judgment calls. Sensitive conversations. Final decisions on hiring, firing, pay, and discipline. Anything where accountability matters.

Accountability does not transfer to software. When a decision goes wrong, nobody accepts "the Ai did it" as an answer. The human who signed off owns it. That is why the last 20 percent is not a suggestion. It is the boundary that keeps the whole system defensible.

Why does the split matter?

Most teams get Ai wrong in one of two directions. Some hand it everything, including the decisions, and find out when something breaks. Others refuse to use it at all, and drown in work a draft could have cut in half.

The 20-60-20 rule prevents both mistakes. It gives the team permission to use Ai aggressively in the first 80 percent, and it draws a hard line at the last 20. The line is what makes the rest safe. Without it, every draft becomes a decision by accident.

How do I use the rule this week?

Pick one repeat task and sort it into the three buckets. What part can Ai do alone? What part should Ai draft while a human checks? What part must a human decide? Write the split down in one paragraph and share it with the team. That paragraph is the beginning of an acceptable-use policy, and it takes less time than one meeting.

Then do the same for the next task, and the next. After a month, the team has a working map of where Ai belongs. That map is worth more than any tool list.

What is the rule not?

The rule is not a promise that Ai will do 80 percent of HR. It is a way to think about the work. Some tasks are almost all human. Some are almost all draftable. The percentages are a guide, not a quota.

It is also not a legal standard. No law says Ai can do 20 percent and no more. The rule is a management tool, built from doing the work, and it sits inside a larger strategy that includes policy, training, and audits.

What to do next

The full framework lives on why Ai needs a strategy, including how the rule fits with acceptable use and compliance. If you want help mapping your own team's work into the three buckets, that is the first step of an enablement engagement. Start with the why-ai page.