Bank Five Nine CEO and President Tim Schneider generously shared insights from Bank Five Nine’s experiences with AI process automation.
At Bank Five Nine, we use a combination of Blue Prism, Python, and JSON as the foundation of our robotic process automation. Each of these tools has its strengths and weaknesses. Some automation uses a combination of all three.
What specific banking processes has your institution automated using AI-driven robotic process automation, and what impact has it had on efficiency?
When we first started our automation initiatives, forty-eight (48) use cases were identified as potential candidates for automation. The following automation processes are in production:
Monthly Auditing
• Run monthly reports from Encompass and upload to StoneHill site/portal for auditing to meet GSE and government loan program requirements in Encompass. File and maintain selected loans in Encompass for access by StoneHill.
Use dates and data from loans in Fiserv Premier to automate emails to borrowers
• First payment details/reminder
• Subsequent Payments, as due after a certain date
• Escrow activity notification (pre and post)
Update Customer Information in Premier
• Read Activity reports provided by 3rd party and perform maintenance in Navigator
Operation Support Maintenance Tasks
• Closed Accounts
• eStatement Setup
• Missing Endorsements
Non-post
• Review report and post to an existing account or return maintenance
Rendering Statements
• Automation of the rendering process
COLD Commit Bot
• Commit all COLD reports parsed into Digital Imaging. The bot emails stakeholders that the process has been completed and that reports are available for the day.
EBC Parser
• Reviews text files from Fiserv that list login attempts for Business OLB users and counts any user with 12+ login attempts.
Check Image Data Extraction
• Review OCR data collected from Director Imaging Automated Indexing and exports it to a file location. This data is compared against the Teller Check report to assign a check number (from the check image) to the payee/remitter/amount data. This is
then compiled and sent to MoneyGram. The purpose is to add the Payee name so MoneyGram can lower the risk associated with cashier’s checks.
Zelle Balancing/Transfer Reports
• RPA Blue Prism, Director workflow, and Python scripting are used. Reports from our online banking system are exported by a bot. Director Imaging exports a text file from Fiserv. A Python script then compares the two reports to identify Zelle transfers that were made by customers after the Fiserv cutoff. These transfers were causing our GLs to be out of balancing due to the difference in cutoff times between Fiserv and NCR OLB.
HMDA Score Update
• RPA BluePrism bots generate a report from our LOS monthly. The bot then goes into each loan on the report and updates the data based on the generated report.
Loan Maintenance Review
• We use BluePrism to export an autogenerated text report from Fiserv and use an input file with several keywords to search for maintenance that needs to be reviewed. There are two separate bots for Commercial and Retail loan operations teams.
Online Banking Activity Reporting
• RPA Blue Prism is used to export documents for archiving and retrieval in Director Imaging. This is a daily process. We maintain records for failed logins, eStatement sign up, enrollment, new transfers, etc.
Loan Document Export for Disaster Recovery
• RPA Blue Prism is used to export documents for archival in Director. The bot processes closed loans daily.
Reg CC Review
• Bots use Director Import PDF Processor and Workflow to route checks to Operations Support to place holds. Operations uses options within the workflow to approve, deny, or void the Reg CC requests. Branches receive auto-notifications from workflow based on the activity of Operations Support
Debit Card Review
• Auto-review common maintenance types to eliminate the need for manual review on all items. From an efficiency perspective, the processes that we have automated have greatly reduced the number of hours associated with that process. People still must review the output to ensure that it is correct and troubleshoot the automation if it errors off. We reduced the tasks, not the responsibilities.
What lessons did your bank learn during the implementation of AI for process automation? Were there any unexpected challenges or benefits?
Automation is not a create it and forget it process. Automation is an iterative process. Continuously changing and evolving as technology and business processes change. Data quality and process mapping are key to success. Having people who know why things are done the way they are is so critical to success.
Which repetitive tasks at your bank have been the easiest to automate, and which have proven more complex?
The tasks that have been the easiest to automate are the ones with processes that are clearly defined and whose outputs/outcomes are known. Example: Did the customer specify that they wanted an e-statement? Yes or No. Any process that, while clearly defined, has multiple outputs/outcomes and triggers other processes based on those outputs/outcomes has been the most challenging.
Has AI-driven automation improved accuracy in your bank’s operations?
Yes and no. Data quality is critical to success. The automation is only as good as the information that we feed it.
What advice would you give to other community banks considering AI for process automation, particularly in balancing technology adoption with maintaining a personal touch in customer service?
Focus on customer needs: Internal / External. Ensure that the automation addresses the specific needs and expectations of the customer.