Recently we hosted our sixth robotic process automation (RPA) roundtable. In 2017 when we first hosted a roundtable on this topic, RPA was the most practical available tool for automating tasks. One of the primary drawbacks noticed however, was that it wasn’t very intelligent. Today, suppliers and organizations are addressing the intelligence process by adding artificial intelligence (AI) and machine learning (ML) to them. Now that practitioners have been working with some of the more advanced tools for adding intelligence as they continue to develop their RPA programs, they have valuable lessons to share.
In our most recent roundtable, practitioners largely from Financial Services and Insurance talked shop about program best practices that they’ve learned along the way using RPA and some of the newer intelligent process automation (IPA) technologies. In this post we offer highlights from this real-world discussion of RPA and IPA in practice.
Considerations for Team Structure
Some of the participant automation teams are structured by technology type, whereas others are structured as a hybrid – bringing in both line-of-business and IT subject matter experts.
There was significant discussion around the benefits of centralization. Centralized approaches make it easier to get processes into production more quickly. Practitioners also expressed that a center of excellence with a variety of technology expertise is an ideal solution. This overarching group can handle initial triage, decide on the best approach and then route to the appropriate areas for execution of the solution.
Regardless of whether your organization takes a formal centralized approach, practitioners still advocate that a variety of stakeholders be involved when designing solutions. It’s easy to wear blinders based on your specific area of expertise, so having the perspectives of different technology experts and vested business owners will make it less likely that you miss critical considerations when designing your solution. This approach can also improve the customer/user experience because internal customers become part of the solution design.
What About the Citizen Developer?
Some vendors are touting the usability of their products for a wider group of users – beyond developers or IT professionals to technically-minded business users often referred to as “citizen developers.” First it’s important to agree on a definition for the citizen developer. One of the definitions offered during the roundtable that resonates with our experience is that a citizen developer is “someone who’s not in the technology group, but has a logical mind for configuring solutions. At the most basic level, citizen developers are configuring – not coding.”
What we heard in the discussion was that citizen developers are still on short reins – not only because of the complexity of the work, but also because allowing citizen developers necessitates more governance. What does this look like in practice? One participating organization created a process optimization group that allows citizen developers to ideate and even begin proof-of-concept planning. IT however takes projects the final mile.
Incorporating Multiple Automation Technologies Into One Overarching Process - Still Pie in the Sky?
Practitioners hear frequently that ML, AI and the new IPA solutions utilizing these technologies, are going to make their work lives so much easier. But given the huge investments required, there need to be some very practical use cases that meet their needs in Financial Services and Insurance. Here are a few examples that surfaced in the roundtable discussion.
RPA + ML for Automation of Donation Verification Research
One practitioner was dealing with a manually intensive process that verifies donors are not receiving personal benefit for donations made through their charitable investment accounts. Before re-designing the process to include automation technology, an employee had to manually research to confirm no personal benefit was received. Stakeholders brainstorming a solution realized there was enough historical data in the system to solve for this, though. Now, the organization uses RPA to retrieve historical data, uses ML to interrogate that data, and finally uses RPA to update the workflow. The result is that now 70 percent of these donations can be verified without manual intervention.
AI, ML and RPA for Not-in-Good-Order (NIGO) Resolution
One of the use cases our team at Doculabs has seen is in the resolution of NIGO items in production capture operations. When an exception is identified, more data is extracted from the NIGO documents if necessary. Then that data is provided to an RPA that fetches relevant data and answers from multiple systems. If the RPA is smart (with AI and ML), it will resolve the issue or propose solutions, which the human can choose from.
RPA and AI for Some Customer Communication
Another use case we’re seeing is the combination of RPA and AI to improve IVR and chat bot responses.
Program Metrics Used for RPA and IPA
Some of the quantitative metrics practitioners are reporting include:
- Total number of processes automated
- Number of cases moving through those processes
- How much manual processing time the bots have saved
Practitioners are considering the qualitative dimension as well, measuring Value Hours. This is how much time is given back to the business as the result of the automation, weighted by quality improvements.
Planning for Resiliency
There was significant discussion about resiliency. Some practitioners have realized the hard way that many of the apps that bots interact with aren’t necessarily stable – which creates huge headaches when it comes to maintaining bots. One very sound piece of advice that was offered is to thoroughly research the stability of the apps involved during the vetting phase as you decide what processes are appropriate for automation. Another great suggestion offered is to go around an app UI whenever possible, and connect through an API instead. That’s a great way to launch an automation more quickly and with greater resilience. Finally, practitioners encouraged an annual review of each process, which can be helpful for both maintaining resiliency and for governance.
Continuing the Conversation
What happens in the production trenches offers critical lessons that can’t be learned from vendor demos, test environments, and sometimes not even proofs-of-concept. As enterprises continue to rely on RPA and IPA, it’s important to learn from experts in the field, including your fellow practitioners. If you’re a current practitioner, planning ahead for a first implementation of automation technologies, or have general questions about the technology, reach out and let us know. We’ll work to include you in the conversation.