Roll Forward: How breakthrough products are redefining ADP as a tech innovator
Roberto Masiero, SVP Innovation Labs
From my long tenure at ADP, I’ve learned that when a company gives you the latitude to move around—either within the technology space or from technology to business or sales—you get plenty of chances to reinvent yourself. And reinvention on the individual level influences the reinvention of the company as a whole, which I think we really see now with Roll™.
Roll™ is a mobile chatbot platform that uses AI and natural language processing technologies to anticipate users’ payroll needs intelligently. It’s the first-ever DIY payroll technology, and it’s so intuitive that our clients just download it and go; a lot of them never even talk to a customer service rep. But while we designed Roll™ to seem effortless, it’s the product of years of creative work with a unified team. The idea for Roll™ was to simplify payroll and HR using a novel UX and platform. I run the Innovation Labs at ADP, where we develop new products as quickly as possible. We’re a relatively small team, around 30 people from diverse backgrounds and with no hierarchy, allowing us to pull together tightly as a group. It’s important to me to have a flat organization because the moment you create hierarchies, you create ways to point fingers. In the way we work, everyone shares responsibility.
We came up with the product idea for Roll™ about four years ago when we were finishing up ADP Marketplace and wondering what to do next. At the time, most of our lab projects were satellite projects, adjacent offerings to our existing core services. I thought, “What if we reinvented the core?” We saw an opportunity to improve multiple facets of our payroll platform—the architecture, the design, the user experience. We had a chance to envision a whole new system.
We fixated on this idea of events—that everything done as an action within the system should be recorded as an event. In fact, we initially named the product “E” for “events.” For example, if you hire someone, pay someone, or terminate someone, we record each action as an event. This way, we know who did what, where they did it, what time of day, and from what device. All that information taken together feeds a machine learning engine where the system gets better the more it gets used. Instead of a system with a bunch of menus, forms, and reports, we imagined a vector of events where events cause other events. We basically built the software as a workflow.
But we didn’t stop there. We also wanted to transform the UI into something much simpler and more direct. People tend to design user experiences with a sense of engagement in mind, but that’s not what we needed here. We didn’t want people engaged; we wanted them to get the job done and exit the software. So with Roll™, the user goes straight to chat and tells the system what they need, and the software understands. If it’s to hire someone, change someone’s W-4, change a payroll schedule, the user asks, and the software guides them through the process using conversational UI.
We also built Roll™ to function 100% on mobile. We decided the UX would use a simple chronological timeline, similar to Facebook or Twitter. Clients love having one place to go to see their activity: “Yes, I ran payroll yesterday evening,” or, “Great, that new W-4 went through.” In addition to optimizing for mobile, we also wanted a strong desktop presence. We noticed our desktop users liked to grab info from the system and transfer it to Excel spreadsheets, so we decided to give them an Excel-like UX.
We finished Roll™ in July 2019 and got a pilot client in August. That fall, we presented the software to ADP’s executive leadership team. We got the feedback that we were sitting on something big that works for small to large corporations. But they encouraged us to focus on the smaller markets, those with one to ten employees. So we spent a couple of months designing an additional layer of software to cater to small businesses. In March 2020, we piloted Roll™ with about 50 smaller companies who all liked what we were offering, and then the executive committee told us to put Roll™ on the market and sell it as soon as possible. So we went from pilot program to full rollout in under a year, and today we’re getting dozens of new clients a day signing up for Roll.
A big part of what makes Roll™ stand out is integrating natural language processing with machine learning. We designed Roll™ to understand the mental model of our user’s meaning. We wanted the chatbot AI to talk the way people talk.
We brought in ADP’s business anthropologist, Martha Bird, and copywriters to advocate for the user, helping us to shape the Roll™ voice. We didn’t simply want AI to predict what our clients needed for payroll purposes––though that ability was definitely important. We wanted the voice of Roll™ to demonstrate human understanding. For example, Roll™ learned to respond more positively when addressing a new hire or giving someone a raise in pay, whereas it is more subdued when discussing termination. It’s that empathetic understanding that gives Roll™ an edge in human interaction.
On the backend, we decided that we didn’t want to run servers, or even containers, like Docker or Kubernetes. Instead, we made every event a function. The beauty of functions is that they only exist while that function is running. So our cost of running Roll™ is extremely low. Using cloud services and this idea of functions is another way Roll™ sets itself apart.
Of course, Roll™ didn’t come without its challenges during the development process. Fraud is something we have to consider whenever we engineer or develop a new product. But this is what I love about the Lab: We think of our challenges as opportunities to make our products better. How can we improve? How can we automate? How can we reduce the amount of burden on the system from someone trying to commit fraud? And when we meet a challenge, everyone jumps in to help. We either fail as a team, or we succeed as a team.
I’d say we’re succeeding right now, and the beautiful thing about Roll™ is that it’s always running. We change our models to pick up on new ways clients ask for things, and every new question pulls into Roll’s knowledge and experience. So the more clients we have, the better the software becomes. It’s an unprecedented level of automation.
A program like Roll™ can help further ADP’s digital transformation from merely a payroll company into a competitive tech company. What makes Roll™ exciting is that it almost creates its own category; it’s a technological solution no one else has. We can dominate this market and apply some of the same breakthroughs—machine learning, using functions—with other ADP products. For ADP as a tech innovator, this is just the beginning of the journey.
Interested in a tech career at ADP?
Click here to search for your next move and make sure to subscribe to our blog!