Get all your news in one place.
100’s of premium titles.
One app.
Start reading
Fortune
Fortune
Sheryl Estrada

EY's generative A.I. payroll chatbot is answering more than 500 employee questions a day

Asia woman ask chat bot function in app on laptop screen looking for customer support online shopping shop with apps,digital marketing concept,artificial intelligence (ai) system (Credit: Weedezign for Getty Images)

Good morning,

Ever receive your pay stub, and something doesn’t look quite right? A ChatGPT bot may come to the rescue.   

"We've created a chatbot that answers all the questions employees have around their pay stub,” EY global chairman and CEO Carmine Di Sibio told CNBC on Tuesday. “It's ingesting 500 questions a day, and it's learning really, really fast.” 

To find out more, I sat down with EY global payroll operate leader Sheri Sullivan, who is the lead on the EY Intelligent Payroll Chatbot project that uses ChatGPT via Microsoft’s Azure OpenAI Service. The chatbot is currently in the pilot stage being tested by EY clients, including airline, tech, logistics, and pharmaceutical companies, Sullivan says. 

Clients are given access to the chatbot via an app that’s available online, Sullivan says. “Our goal is, when we fully finish all the industrialization, we will launch it and put it into our mobile app, which is currently being used in 131 countries in 49 languages.” EY plans to launch the chatbot in the fall. 

“We have tested the bot in over 27 languages and haven't had any issues,” Sullivan says. “Our clients are playing with it, seeing if the app can actually answer the questions.”

What’s the ChatGPT bot’s job? “What we did with this chatbot is we put the regulatory compliance knowledge that we have, the anatomy of a payslip, because most employees don't understand their payslip structure anyway, and what the different acronyms are and how those relate, and we also put company policies,” Sullivan says. “And right now, the chatbot is able to go between the three.”

The chatbot currently specializes in “tier-one questions,” she says. “They’re not overly complex questions, but they are specific to regulatory compliance or specific policies and they're specific to the employees,” Sullivan says.

Sample questions? Sullivan offers some examples:

— "I make 5,000 euros per month, and I live in Spain. What would be the effect on taxation if my monthly pay was increased to 6,000 euros per month?”
— "I'm an employee in France, and I'll receive a 20,000 euro bonus next month, how will it be taxed? What will my take-home pay be?”
— "Why is there a change in the tax amount that I paid on this pay stub versus the last pay stub?”

“I really think this bot is going to be used across the enterprise,” she says. But very heavily, with payroll, which is “50% owned by finance and 50% by HR, it touches both,” she says. “So it absolutely will have an impact with both HR and finance.”

Without a chatbot, how do those questions normally get answered? “Normally, clients deal with the tier-one questions in three ways,” Sullivan explains. “The first way is they don’t. There are a lot of companies where, literally, there's no place that really employees go to ask these types of questions. The second is they have some local HR person that people are directed to. And the third is they have some type of shared service center. It could be in the company, or a third party BPO [business process outsourcing] provider.”

When it comes to generative A.I. security remains a major concern. I asked Sullivan about the team’s approach. “The chatbot only has access to EY-vetted data sets,” she says. “We're actually making sure that the environment is per client, and not shared across clients.” Sullivan’s team, active in 159 countries, has been crowdsourcing for prompt engineering to determine the right rules around prompts, including questions the chatbot shouldn’t answer, she says.

“I think we're kind of crawling, walking, and running before we release it to production to make sure we’ve addressed concerns from a data privacy and security perspective and to make sure that it's fit for purpose,” Sullivan says.

Does she think the financial services sector would adopt such a chatbot? "I'm sure it will," Sullivan says. "But financial services will probably be one of the last ones. Let me tell you why. Financial services, like banks, tend to have more strict requirements around what they allow their employees to have access to."


Sheryl Estrada
sheryl.estrada@fortune.com

Upcoming event: Join us for Fortune’s Emerging CFO’s virtual conversation, in collaboration with Workday (a CFO Daily sponsor), “Maintaining a Growth Mindset In Turbulent Economic Waters,” Wednesday, June 21, from 11 a.m. to noon ET. Geoff Colvin, senior editor-at-large at Fortune, and I will talk with McKinsey & Company Senior Partner Ishaan Seth, and finance chiefs Dan Durn, CFO at Adobe, Mandy Fields, CFO at e.l.f. Beauty, and Xihao Hu, CFO at TD Bank in the U.S. We’ll hear from these experts how success can be significantly influenced by how we perceive our talents and abilities—our growth mindset—and how this applies to the role of the CFO. Click here to register!

Sign up to read this article
Read news from 100’s of titles, curated specifically for you.
Already a member? Sign in here
Related Stories
Top stories on inkl right now
One subscription that gives you access to news from hundreds of sites
Already a member? Sign in here
Our Picks
Fourteen days free
Download the app
One app. One membership.
100+ trusted global sources.