Robotic Process Automation (RPA) is the use of intelligent software to perform repetitive tasks that would normally be done by an employee. While it has been promoted for years, our clients are now seeing the tremendous process improvements and cost savings this technology offers.   

While there are over 60 companies currently offering RPA, only 8 have 2018 annual revenues above $20 million, which is the minimum we recommend our clients to consider. None of these technologies have the capability of performing in every business function. Some are good at front office applications while others are good at back office applications. Some have built in functionality for cognitive ability while most do not. Each must partner with a third-party vendor for higher-end cognitive capability.  Our top 8 RPA vendors in descending revenue order include:

  • UiPath
  • Automation Anywhere
  • Blue Prism
  • Nice
  • Pegasystems
  • NTT-AT
  • EdgeVerve

In our client example, we have a healthcare agency reconciling a million consumer applications with multiple insurance carrier enrollment files. Due to mismatches in names, dates of birth, and addresses, they experience a 5%-10% failure rate. Those failures are placed in a work queue for people to manually analyze and remediate using external source systems and then the applications are resubmitted. 

As you pursue your own RPA project, it is important that you follow these five steps to ensure your success:

Step 1: This technology does not apply itself to every business application. You need to be savvy and select the appropriate process to automate. This technology is great for business processes that are high-volume, performed manually and rule-based. In our case, the enrollment reconciliation process met all three criteria and was thus selected. If you select a process that is low-volume, your teams will spend too much time configuring the ‘bot to perform the task and you will not see the ROI on the technology. Likewise, if the process you select is extremely complicated and does not follow specific rules, your teams will not be able to adequately code those rules into the ‘bot and your project will fail.

Step 2: Analyze your process and determine the business case. This is a critical, albeit familiar, industrial engineering step. You need to capture each step of the business process and make any necessary improvements before you automate it. This is not terribly different from the process redesign efforts that businesses did during the ERP implementations at the turn of the century, but the scope is smaller and the tools are better. You can use a tool as basic as Visio, or you can use a more sophisticated tool, like UiPath’s StepShot. 

Have your team select the tool, analyze the process and capture the time, effort, cost and decision details used. Once this is documented, you now have your benchmark to use to determine if RPA technology is beneficial. The output from this step should be detailed process flow design documents, decision criteria, and heuristics.   

Side note: – Do not underestimate the effort required to alleviate employee concerns that they are about to lose their jobs or their identities. This is very real and can kill your project. Show these stakeholders that this technology will allow them to elevate their analytical skills and perform more value-added, fulfilling, work.

Step 3: Next, you will need to select the right RPA software for your business application.  If your business and IT groups have ever conducted technology vendor evaluations, this will be similar in many ways. The three key elements you need to cull this vendor list will include:

 1) Is this business function front-office, or back-office? This is the big differentiator we have seen in the vendors because no vendor appears to be best at both. 

2) Will this be a stand-alone function, or will you need to integrate cognitive ability?  

3) Will you need to interface with an external system? Some vendors are architected better than others for this requirement.

In our client’s case, we had a back-office function that needed to interface with a federal source system to perform the processing and it had a low-cognitive requirement. We selected Blueprism for this specific application and it worked well. 

Step 4: Configuring the ‘robot’ (software) – These tools do not require programming skills, but technical knowledge will make your project roll out smoother. We found a classic, cross-functional team that included a Blueprism specialist, a software engineer with oversight of a systems architect, a subject matter expert and a business analyst worked well. These are fast-prototyping projects with timelines measured in weeks, not months.  

Step 5: Fast-Paced Test and Revision – We found that the most effective technique for rolling out this technology was to build-test-revise-test and then pilot the ‘bot with a limited amount of data and then use a human cross-check to verify accuracy. 

We are seeing our clients reach accuracy of 100% in transactions, and reductions in FTE hours of 25%-60% for a given function. In addition, since these ‘bots can work 24/7, we are seeing through-put increases approaching 300%. 

Previously, we had analysts performing this work. Now instead of reconciling, they are able to analyze the data patterns and evaluate trends.


Arkenstone Technologies

For more information, contact:

William Estright

(724) 513-4945