The timing coincides with the launch of our own Big Data project. Behind the scenes at Dataplan we have lots of IT boffins but we have hired a data scientist to help us develop, test and implement machine learning and AI solutions for our payroll solutions.
How can AI be relevant in the world of Payroll you may ask. A few years back we would have said exactly the same. What could AI mean for us? Banks of robots typing data into computers removing the need for skilled and qualified payroll staff.
When you start to look at machine learning there are some very obvious practical applications. Machine learning is about finding patterns or algorithms that are trained across big sets of data. Once identified these algorithms can be used to predict future outcomes.
Of course payroll processes historically do look at patterns and AI does exist in payroll. Classic AI that already exist are traditionally based on predefined rules. For example AI may identify that when someone’s pay has changed by more than x% it could be an unusual transaction and merits further investigation. Our skilled payroll professionals enhance this process through their own manual testing of payroll data.
A humans ability to see patterns in data as potential anomalies are where the real intelligence happens and where a set of predefined rules beyond classic AI are limited. Recent developments in AI see neural networks that are capable of running through huge sets of data and genuinely learning how to reach conclusions rather than the more directed approach of classic AI, opening up a whole world of possibilities where we expect to match and exceed a human’s ability to spot patterns in data.
One specific example of a pattern we are particularly interested in is pay distribution across payroll. By looking at pay distribution at multiple levels within a pay period, historic pay periods and across similar businesses we can build predictive models that highlight pay patterns across an organisation.
So what you may ask? Using the above algorithm we can identify anomalies within payroll against and beyond traditional exception reporting models. This way of modelling data will allow us and our customers to better understand the components of a payroll and whether the payroll and changes to it are within a sophisticated set of norms. It will also provide much greater management information about the payroll itself in the context of previous periods, previous years and the industry sector itself. There are many other areas where we expect to be able to identify patterns and algorithms across big data sets.
These are changing times in the world of payroll and at Dataplan we are excited to see what AI can bring in the same way that we saw how machines can learn on their own to play games such as the Atari classics and Go in the Joy of AI program. Our Big Data project will run across 5 million anonymised payroll records and use technology identify patterns well beyond those that a human could possibly expect to see.