There’s an adage in the world of fitness and nutrition, ‘you can’t out train a bad diet’. Similarly, in the world of processes and systems, it’s fair to say, ‘you can’t out automate bad data’.
Data is Fuel
Data is the fuel that processes run on. If data is not timely, complete, trustworthy and accurate, the process must have steps in it to correct those shortcomings. These steps are often less obvious in human driven processes as humans can integrate the mundane with the complex at the same time with relative ease. Judgement calls and computation can intermix in what seems like a single action.
Think, for example, of a payment calculation based on earnings, if it appears that a school teacher is only earning $400 per month (or $40,000 per month). The human who is running the calculation will identify this as strange and double check before processing further. If you’d like to automate this kind of judgement call, you have several options, including:
- Build a complex system of matrices to match human experience across a large data set
- Invest in an AI processing engine to learn what “looks right”
- Revisit the processes that got you untrustworthy data in the first place
Often the last option of revisiting the processes will get you to a place that works out less expensive, more robust and more scalable than the other two.
Connect the Core
In assessing automation options, I’d propose that your first step is to ensure all the process inputs are correctly maintained in appropriate systems of record. The systems of record must also be connected in an accurate, complete and timely manner. Where data is not up to standard, manual processes must be introduced. Those inadequate processes will chip away at your automation goals.
Barry Duffy is a FINEOS Claims Product Manager. Over the course of 2019, he’ll be working with the FINEOS Practice Programs to define new standards in industry processes to facilitate further automation and user experience improvements across claims and absence management.
