Skip to main content

Behind the Build: Inside Paychex's bet that AI could feel as personal as a local office.

Paychex rethought payroll processing to deliver twenty-four-seven service for the small businesses that can't afford to wait.

  • Client satisfaction

    95%

[Behind the Build is a series where Sierra customers share how they designed, deployed, and scaled their AI agents.]

Paychex pays one in eleven private sector workers in the United States. For the small and mid-sized businesses and their employees that rely on them to pay their bills and provide for their families, payroll isn't administrative. It's existential.

The company was founded in 1971 on local relationships: an office in your town, a person who knew your name, a service model rooted in "I know you because I know you." But as the business scaled and the product suite expanded to include retirement plans, HR, benefits enrollment, and more, that intimacy became harder to maintain, this is where technology enables Paychex service professionals to shine through.

The problem was always time

The friction showed up predictably during the busiest weeks on the calendar. Christmas and New Year's weeks drove surging call volumes that often outpaced Paychex’s staffing capacity. The result: extended hold times, mandatory overtime for employees, and service reps who didn't always have the knowledge to handle every question.

"Imagine you want to go to your child's holiday concert at school, and you've just been told you have to work mandatory overtime," says Liz Roaldsen, Senior Vice President, Operations and Customer Experience at Paychex. "People want predictability in their work life."

On the customer side, business owners — often already wearing every hat in the building — were waiting on hold for something that needed to happen now. And expectations kept rising. "This Amazonification of everything we do," as Liz describes it. As consumers, we’ve grown accustomed to having very little tolerance when it comes to our expectations for an immediate answer.

Designing for trust as much as throughput

When Paychex set out to build a voice AI agent for payroll processing, the brief went beyond automating calls. It was: deliver immediate, twenty-four-seven payroll processing with a human degree of accuracy…and make it feel effortless.

That distinction shaped every design decision. The voice. The greeting. The speed at which information was captured. Each element was measured against a single standard: does this feel local, human, trusted, dependable?

The early results from those using the agent validated the approach: 95% positive client satisfaction scores, 93% positive customer effort scores, and customers who try the agent coming back to use it again.

"That's the proof," Liz says. "They're reusing it. They're giving it high scores. It's getting the job done with great accuracy, and our team has more time to spend with customers."

The real work is change management

The team has shortened the greeting, condensed the legalese, and iterated based on ongoing feedback. The technology keeps improving. The bigger challenge is convincing customers to try it in the first place, and that happens one call at a time. When someone realizes the agent isn't the painful IVR experience they were bracing for, something clicks. "It's like an aha moment for them," Liz says. "To know that they could trust this technology to deliver what was once left for humans to deliver."

Paychex is being patient by design: proving the concept, collecting the data, and letting the experience speak for itself.

What AI frees humans to do

For Liz, the most important part of the story is what happens around the agent. Paychex has around 700 HR business partners who provide deep, expert guidance on HR matters ranging from attracting talent, benefits, and compliance. The agent handles the transactional volume — the "pay Tracy ten hours this week" calls — so service reps can show up for the moments that require human expertise and judgment.

"The moments that matter of keying in your payroll, those are not where customers probably want us to spend our time, and vice versa," Liz says. "What we want to be known for is having humans that provide deep expertise leveraging AI enabled technology."

It also gives service employees something less tangible but equally valuable: predictability. The ability to come in at nine and leave at five. To make childcare arrangements that hold.

Advice for CX leaders on the fence

Liz's counsel: start with the business problem.

"For us, the business problem was answering the phone as quickly as possible for customers. As simple as that."

She's quick to address alternatives and why they don't hold up. "Those benefits will erode over time. You've got to hire and retrain staff. You've got high occupancy roles that have high churn. Knowledge dwindles. All of that doesn't happen with the agent."

The through line is the same one Paychex was founded on more than fifty years ago: know the customer.

"When it's implemented thoughtfully, it doesn't dilute the experience," Liz says. "It really helps you deliver an enriched experience to your customers.”

Discover what Sierra can do for you

Find out how Sierra can help your business build better, more human customer experiences with AI.