This standing room only crowd for an industry conference’s AI session, something seen with great regularity these days, is actually from last week’s American Payroll Association event in Orlando. You read that correctly.
While the payroll function and services market likely weren’t among the first AI or RPA candidates written on white boards in innovation labs, this obvious level of interest might suggest a “can’t see the forest through the trees” dynamic operating in some of those innovation labs. Back-office corporate functions such as payroll are in fact fertile ground for RPA and intelligent automation overall, given the preponderance of recurring manual tasks and transactions not dependent on person-to-person interaction.
Innovation labs are now on the case.
The speaker for this session called “Prepare Your Teams for the Future of Payroll: Robotics, Automation & Shared Services” was Brian Radin, President of global payroll services provider CloudPay and long-time entrepreneur in the HR Tech space as well. Brian immediately got everyone’s attention by factually reporting that the number of bank teller jobs did not decrease in the years following the introduction of ATM machines. Teller numbers actually went up due to shifting staff costs to support new, higher value services within retail branches, which ultimately allowed more local branches to open up, tellers in tow.
Using AI in the realm of HR operations, including cognitive computing and RPA (Robotic Process Automation) or bots, has been explored in my blog posts and also a recent POV. Radin’s session focused specifically on AI’s current and future use in payroll operations, including via services providers like CloudPay and over a dozen others to be profiled in my HfS Blueprint Report “Payroll-as-a-Service: 2017” (published this July).
Some Easy Questions, Some Hard Ones
Radin’s talk directly addressed some key questions about “AI in Payroll”; e.g., how can (or will) these capabilities help payroll clients spend less time on manually intensive, routine or recurring tasks, ones that machines can often handle with more alacrity? And are there other tasks where resourcing can be toggled between human and bot staff depending on availability? Here the presenter highlighted examples like data validations and checks pre and post-payroll run (payroll has quite a few of those), machines fixing errors or automating the consolidation of data, and of course, chatbots to answer recurring questions like “what is my accrued PTO?” or “when will I receive my first check?” (Questions which come up hundreds of times per year.) Allowing RPA tools to handle these will benefit clients of providers like CloudPay and any other vendor investing in these capabilities. And as far as highlighting a “resourcing agnostic” (bot or person) type of activity in payroll, the example given was using people or bot staff to train new staff.
One of the highlights of the session for me was listening to questions attendees were posing at the podium afterward, away from the large audience. One gentleman told Radin that training and re-skilling of staff were already going on in his company in areas where RPA would be heavily leveraged, but it sometimes provided only a year or so of “job runway” for employees until RPA would impact their next job. Then re-skilling would have to start again. Radin’s response was both admirable and accurate: “Re-skilling decisions in the RPA era is very much a work in progress.”
Machines that Do, Do and Think, and Learn
CloudPay’s VP Marketing, David Barak, elaborated for me after the session on Radin’s slide which highlighted these three different categories of RPA capabilities: “Do” describes the use of RPA to move and manipulate payroll data without human involvement, as one example. “Do and think” capabilities include the machine flagging and fixing hundreds of data issues pre-payroll run; and while “Learn” is an RPA capability in payroll processing that’s still being tested and improved upon (as with machine learning in most areas), it includes anticipating spikes in payroll processing costs based on time of year, business cycles, new regulations, etc. This information can then guide the customer in optimizing staffing levels.
Bottom Line: Payroll departments and services provider clients will increasingly benefit from emerging RPA and cognitive capabilities. It will probably be a few steps forward and a couple backward until something akin to a “human/bot hybrid resourcing homeostasis” is figured out – in general, and also reflecting specific customer contexts. Predicting how far / how fast with any precision, in any industry or discipline, is almost a total crapshoot. One thing we do know, machines are not nearly as susceptible to errors due to work overload or distractions.