The cost for a robotic processing automation (RPA) bot averages between $5,000 to $15,000, according to an oft-cited industry number. Bots can cost much more, however. In fact, when Wells Fargo started its automation program with unattended bots in 2018, the price tag for a bot was as much as $50,000, and Gartner told Bank Automation News that RPA initiatives range from approximately $70,000 to $100,000 in the first year of operation.

How can banks make every dollar count, when it comes to RPA? BAN spoke to industry leaders, consultants and practitioners about bot best practices and came up with five recommendations.

1. Find a high-level executive sponsor

This is a common best practice for success in any technology project, but big banks found success in 2020 when their senior management championed automation, said Amit Kumar, vice president of financial services for RPA provider UiPath.

Chief executives at some of the biggest U.S. banks have recently cited their focus on automation, even in their earnings reports, Kumar said. Three or four years ago, that wasn’t the case, he added.

“About 100% of all the managers that have high-use automation, they have either the CEO or CFO as a champion,” Kumar said.

2. Define the process before you send out the bots

Banks should define and then evaluate a process before automating. As Gartner analyst Nicole Sturgill told BAN, unless banks evaluate the process they want to automate, they could simply automate a bad process — and do so at scale.

Look at ways to improve the process before automating, suggested Karen Reichle, vice president of customer success engagement at RPA software company Nintex. In most cases, this leads to a better process even before the automation is in place, she said.

“Always look at the process, look for areas where you can make it better and then do your automation,” Reichle said.

Read more: How to assess value in robotic process automation.

Process discovery can help achieve this goal, said Beji Varghese, who specializes in working with financial institutions as a partner with the private tech and risk consultancy Guidehouse. Process discovery can help determine what humans and the systems involved are doing, he added.

3. Involve employees in the automation

One emerging best practice for deploying bots is to couple the bot with a human to partially automate a process, said Kumar.

At one investment bank, junior bankers were spending 70-90 hours each week creating investment reports for clients, Kumar said. By deploying bots to automate the repetitive aspects of production, such as extracting data and moving it into a PowerPoint template, the bankers were able to produce customized reports in fewer hours, he said.

The largest banks are also putting the technology into the hands of the employees like relationship managers, branch bankers, investment bankers, analysts and business teams, Kumar added.

4. Practice bot governance

Banks should govern bot access in the same way they would govern individual access, suggested Nintex’s Reichle. One bank told her that it had designed a bot that had unlimited access to “do anything.”

“Even if you have a bot, which is very common that has its own login credentials, you still want to attach a person to that bot … so that you have accountability,” Reichle said.

Another way banks should govern bots is by putting mechanisms in place to exit and restart a bot “gracefully,” suggested Varghese. Otherwise, a bot may stop — without even providing an error message or instructions about how to reboot.

5. Test the bots before you deploy

“Test, test, test, test,” said Rajiv Garg, associate professor of information systems and operations management at Emory University’s Goizueta Business School, when asked to recommend one best practice for deploying bots.

“If you don’t test, you will make mistakes, be it being discriminating or making an error which will lead to larger losses,” Garg said.