Companies are increasingly turning to AI to improve their employee experience and to save money. We examine some of the pitfalls of this approach and look at how companies can leverage the technology but have to be very careful and selective when doing so.
Payroll is often a source of frustration at companies, and it’s increasingly difficult to find and staff payroll positions. There have been some payroll case studies recently that highlight how AI can be used to answer employee payroll questions. Ernst and Young, the global accounting and auditing firm recently deployed a customized chat bot to answer employee payroll questions.
On the surface it seems like a great idea. Employees will be able to get the answers to payroll questions without having to interact with a human. But we have some reservations about the approach of increasing automation in payroll.
Employees Usually Only Contact Payroll When There is a Payroll Problem
For those of our readers that have worked in a payroll department, they will know that the most common causes for employee interactions are because their pay is either incorrect, or they believe it to be incorrect. Payroll teams should do everything that they can to avoid errors, we even wrote a guide on how to avoid payroll errors., but also know that they happen and are sometimes unavoidable.
When employees reach out, they are typically confused, angry or both. These are highly critical interactions that can either build employee loyalty or destroy employee morale. Because of the importance of these interactions, they should be handled by humans that can hear, understand and acknowledge an employee’s concerns.
By too aggressively pushing employees to interact with AI, it will only increase their dissatisfaction.
Behind the scenes, there are many ways that payroll teams can leverage ChatGPT to improve employee satisfaction. When it comes to front line problems, employees should have a quick way to access someone in payroll that can help guide them through and address the problem they are facing.
Most Companies Will Not Have the Knowledge to Train a Payroll Chat Bot
In the Ernst and Young example, they leveraged their tax and accounting experience to customize the training that their payroll chat bot received. Being one of the largest tax and accounting firms in the world, they already have this knowledge base available.
Most companies that do not specialize in accounting are not going to have the payroll knowledge base required to train a chat bot.
Fortunately, some of the largest payroll processing companies in the United States are working on AI products to help their clients and their employees. Paylocity launched an AI Assist feature in March of 2023 and while we haven’t heard of any AI features coming from ADP yet, we would be surprised if such features were not already under development.
Payroll Compliance is Difficult and Changes Rapidly
Another challenge with utilizing AI as a primary reference for employees is that payroll laws and tax change all of the time. Pop! Automation is based in Portland, Oregon and there seems to be a new set of payroll taxes every other year or so and the implementation of these rules change over-time as the government figures out how new tax structures should be implemented.
We normally lean on payroll processing companies like Paylocity to provide guidance on how to implement payroll changes, but there is a lag time between when the local laws change and when the information makes its way to a company’s payroll group.
It’s currently unknown how quickly an AI chat bot can adjust its advice to recent regulatory changes. We also don’t know if a bot will be able to decipher between competing sources of information on the internet or other training sources.
AI Chat bot provide a lot of promise for helping companies maintain a compliant work environment but we would be hesitant to fully trust one, especially when the information is more ambiguous and recent in nature.
Companies Will Need to Limit AI to Relevant Topics
OpenAI, the developers behind ChatGPT have done a decent job of setting up ChatGPT to only reply to answers within its scope of knowledge. They’ve done this in an effort to not mislead the public and to limit how useful ChatGPT will be for nefarious purposes.
Companies will similarly need to place controls around internal chat bots to prevent them from providing information that they are not qualified to do. This will protect companies from potential lawsuits in the future.
Our advice to companies who are hiring payroll staff is to coach their staff not to provide tax or investment advice to employees. Even though employees see the payroll department as experts in their field, very few people processing payroll are qualified or licensed tax or investment advisors.
In the same way, AI should be trained not to provide specific tax advice or at the very least surround advice with disclaimers that the information may not be correct and that employees should talk to a qualified advisor who will better understand their wholistic tax situation.
Conclusion
Unfortunately for companies that wonder if they should outsource payroll to artificial intelligence the answer is not quite yet. We see a lot of potential for the technology to help employees understand their paychecks and get answers to common questions to free up experienced payroll staff to work on more complex problems.
Most companies will be better served with AI Chat Bots that are backed by large and reputable accounting or payroll companies that have the resources available to ensure that the chat bots are trained on accurate data and can maintain the training dataset over time.