Automation projects never fail; it’s just that they don’t always meet all expectations.
For example, in the case of an automated clearing house (ACH) automation, the processes needed to be rerun, and a subject-matter expert (SME) had to process some of the transactions manually. But in the overall scheme of things, the task, which originally used to take eight to 10 hours a week of SME time, was reduced from anywhere between 30 to 60 minutes, with higher accuracy and efficiency.
The expectation was that after implementing the automation, there would be no need for any human intervention. Even though the efficiency improved by 90%, the expectation was 100%. Even though the ACH runs every Monday and Wednesday and continues to save on man-hours and impacts the bottom line of the balance sheet, it was still deemed unsuccessful.
A new bot needs to be treated like a newly hired employee
Typically, as part of the induction and orientation, an employee gets up to speed with basic functions and responsibilities, but it takes him or her time to become fully functional, and they usually require help from colleagues and/or supervisors to solve issues. A relevant example would be if a new AP clerk has the responsibility of performing an ACH transaction, they would be able to do the most obvious and basic transaction, but in the corner cases, they would need guidance from a supervisor or colleague as they go through these data points experiences.
Drawing a parallel to this example, when it comes to a bot, we need to look at the specific details of its requirements. This helps the bot achieve, say, roughly 80% accuracy, and the other 20% would be achieved based on multiple runs in the controlled environment. Every run may present some nuances that can be learned from and implemented back into the bot. With every bot run, the bot moves one step closer to improving its accuracy.
Key points on how to treat bots like newly hired employees
• Provide emphasis on detailed requirements to serve as an induction program of sorts. A detailed induction program will make the bot more productive and faster.
• Like a new employee, run the bot in the shadow of an expert for at least two to three production runs. This will result in fine-tuning the nuances of the bot, which can then be improved, resulting in higher accuracy.
• Bots should have an ID and the same level of security as a human so that everything can be tracked.