Robotic process automation: is it really scalable?

Image: Shutterstock (For illustrative purposes only)

There is so much written about Robotic Process Automation, it promises to be the biggest change lever in mid and back office operations seen in the last few years. Indeed, organizations have leveraged this well to derive productivity and quality benefits. Most value propositions refer to productivity i.e. cost take outs which in a benign revenue environment is pretty welcome. That said, I actually sense a drop  in the interest levels of the initiative in financial services – initial hype seems to be wearing off, proof of concepts don’t seem to be scaling fast enough and value realizations are slow. What’s going on?

I am going to focus on sharing some experiences (read scars on my back) and conversations in the global Financial Services space that reflect this concern.

Productivity uplift is low, we only saw some incremental benefits of faster processing but no benefit to the end customer – in isolation a UI (user interface) based automation agenda will probably not provide end to end solutions. We experienced this with a global bank, wherein automations in their payments business gave a good productivity uplift but did little for their corporate customer’s end experience till we looked at the process in an end to end manner and redesigned it.  In another instance a team at a call center spent many moons automating customer requests. What was really needed was a unified view of experience across call and email channels to drive a better experience. The first proof of concept solutions focus on some quick wins to declare victory. As the program scales, end to end design is important to derive maximum value.

The myriad tools in the market and their claims are befuddling – I am a big proponent of the fact that there is no one single panacea solution out there and UI automation alone will not solve all process situations. Let’s take the mortgage business as an example in the United States. We did extensive work with a mortgage servicer to address the compliance requirements of a heavily paper driven business. Documents have been digitized now but you still need extraction engines backed by machine learning and a deep domain understanding to solve for this requirement. It took a while for us to code in every single nuance of a loan boarding audit at the time of service transfer. The initial effort with the tool self-learning on extraction from mortgage documents is yielding good results. So in this case, it was an OCR engine, a workflow, a machine learning engine and UI automation working together.

I don’t know how my robots would behave if applications or third party environments changed – This is a real and valid concern for many and we had some interesting learnings on failure scenarios and rogue robots. To the human eye a blank field is reason enough to pause processing but to a robot it could be a ‘null value’ with its own processing logic unless controlled. The other fascinating thing we had to solve for was the co-existence of a robotic and manual workforce. That led to the creation of a control tower with one of our anchor customers where a set of robots look out for changes in the IT applications and systems to manage the robotic workforce and seamlessly transfer work exceptions to a manual workforce – robots managing robots.

Who owns this initiative – is it the CIO or the COO? This is an interesting one, I have heard cheers from ops leaders calling this a revolution. Empowering them to change and automate processes without going through hoops. I have heard a CIO saying that the power of redesign of the bank has been handed over to ops. Yet at the end of it all, there seems to be a software development lifecycle here too – design, documentation, development, architecture, testing, implementation, maintenance etc. So what’s new? I believe in the first two odd years, the initiative will continue to be owned primarily by the IT teams. After all we in operations have to up-skill ourselves on technology too. The maintenance and easy change management of the solutions will be our confidence drivers to take complete ownership of this initiative. Progressive organizations have brought some sort of a CoE together to gain a consistency of approach and implementations across Lines of Business.

What lies ahead? I think the RPA initiative is the beginning of a transformative journey that will leverage artificial intelligence to solve business problems and requirements. There are some interesting use cases that we worked on using our AI engine. A key issue in the scalability of the automation engine is the effort needed to map processes and define automation opportunities. The AI engine used the ‘system footprint’ to chart out the happy path of the process through the systems, and voila! We had an automation shortlist. Another one in ‘fraud investigation’ helps analyze spend patterns of individual customers and reduce incorrect holds due to legacy alert engines. The ability to analyze huge amounts of data in real time can help reduce the volume of alerts in itself and improve false positive ratios thereby redirecting capacity to find real fraudsters.

While there is much out there that seems to be holding back from blazing poster child productivity gains, it will be a pity to see the efforts of ‘operations entrepreneurs’ lose steam. In my opinion, the trick lies in getting started meaningfully within lines of business with a balanced focus on cost outs, improving business metrics and customer experience.