46% don't want AI in payroll: Trust moves slower than tech

Podcast

Oct 5

Written By Chuck Gose

People will shrug when AI summarizes a meeting. They won't shrug when it gets their pay wrong.

That's the takeaway from PayrollOrg's 2026 Getting Paid in America survey, and Jenni and I spent a good chunk of Episode 77 on it. Of the 9,451 people who answered the question, 46% said they're uncomfortable with AI being used in payroll. 22% are very uncomfortable. Only 25% are comfortable.

And it's getting worse. In 2024, 39% were uncomfortable with AI calculating their pay. Two years of AI everywhere hasn't moved people toward trust on this one.

The release pairs the finding with a quote from UKG: only 11% of payroll teams use AI features, and "caution cannot become stagnation." Read that again. A payroll software vendor is telling payroll teams to go faster, in a release about employees asking them to go slower. Jenni's read was blunt: it's close to a non-story, a vendor trying to get its customers using more tech.

The line that stuck with both of us came from Hattie Roach at Unite 26 in London: AI moves at the pace of trust, not the pace of tech. Jenni tied it to hygiene factors. Getting paid correctly is one of the most basic things an employer owes you. If people don't trust the organization to get that right today, they're not going to trust it with AI in the loop.

None of this means AI stays out of payroll. Jenni's point was that AI with a human is fine. If something goes wrong and I can pick up the phone and reach a person, I don't much care what ran the calculation. Robots running pay with nobody to call is a different story. My version is simpler: the worst thing you can do to an employee is mess with their paycheck.

The harder question is for whoever owns the AI roadmap. Payroll looks like an easy efficiency win on a slide. But the things employees care most about are the worst place to experiment, and the efficiency case doesn't account for what one bad pay cycle does to trust in everything else you roll out. If payroll is on your AI list, has anyone told the people whose pay is involved?

Also in this episode:

Microsoft is moving communications out of marketing and under Brad Smith's legal and policy group, while Satya Nadella asks "the people closest to the work" to help tell the company's story (GeekWire). Jenni thinks clearer rules will get more employees posting. I worry employees hear "no" a few times and quit, or get handed a script instead of a voice.
ResumeBuilder says 18% of workers have a job that's significantly or completely different from the one they were hired for. We questioned the "bait and switch" label, since most roles drift without anyone lying, and the report's 74% stat, which is padded. The real problem is the career conversations that get delayed or avoided. My fix: write job descriptions around outcomes, not outputs.
LinkedIn's AI Confidence Index found only 38% of Gen Z plan to build AI skills, compared with 63% of Gen X. Jenni pushed back on the idea that Gen Z is refusing. They haven't worked out which parts are worth learning, and honestly, nobody else has either.
A Cornell study looked at who gets scarce training budget and how fairness norms shape the fallout. I asked Jenni how she'd split $5,000 between two people. She talked about having the conversation first. I'd give each person $2,500 and let them build the business case. She says we agree. I say she didn't answer the question.

Watch or listen to the full episode on Frequency, with timestamps, links to every story and the full transcript.