Why RPA implementations fail

Robotic Process Automation hype is at an all-time high, with an alphabet soup of providers and offerings. Industry watchers are declaring RPA a must-evaluate technology, and some are heralding the start of the next industrial age. Software providers on both ends of the automation/AI spectrum scramble to incorporate new capabilities and meet in the middle, Given all that, one fact is often lost – many RPA implementations actually fail.

Failure frustration anger user man worker
Thinkstock

Robotic process automation (RPA) uses software code, often called “bots,” to record and replay repetitive software application interaction tasks that are normally done by employees in an enterprise, as an example, inputting order management data into an enterprise resource planning (ERP) system and returning confirmations or exceptions by email.  And the beauty of RPA is that it does not require expensive and time consuming enterprise application integration efforts, as it essentially interacts with various applications the way humans do – through the screen user interface. As such, it offers large opportunities for companies to boost efficiency and free up staff for higher-level tasks.

In 2017, it is hard to find a company that is not already implementing RPA or seriously thinking about getting started soon. The hype is at an all-time high, with an alphabet soup of providers and many offerings. Industry watchers are declaring RPA a must evaluate technology, and some are heralding the start of the next industrial age. Moreover, with artificial intelligence (AI) becoming more of an enterprise agenda, boundaries are blurring as software providers on both ends of the intelligent automation to artificial intelligence spectrum scramble to incorporate new capabilities and meet in the middle – the sweet spot of the market.

In all of that momentum and energy, one fact is often lost – many RPA implementations actually fail. Five years into the industry, there are few examples of strong success, among the thousand plus enterprise robot deployments. Alex Edlich and Vik Sohoni, senior partners at McKinsey & Company, find, “several robotics programs have been put on hold, or CIOs have flatly refused to install new bots.” Looking across the implementations, it is possible to recognize patterns in ones that were successful as opposed those that are not. The path from experimentation to industrialization is mapped with many lessons learned for enterprises starting the journey. There are three broad takeaways to reflect upon:

1. Don’t trade the entire tool box for one screwdriver

Successful implementers holistically think through the entire plan for digitization, not just focus on the one RPA component. A typical automation effort requires a set of complementary component technologies that together address the entire requirement; planning this early and fully is a leading indicator of downstream success. For instance, often neural net-based text extraction capability will need to be implemented for non-positional optical character recognition (OCR) to be able to use free-form documents. Similarly, integration into telephony systems or messaging or Simple Mail Transfer Protocol (SMTP) engines are often needed to automate a process fully.

Also, without the larger plan in place, teams can struggle through the RPA implementation to get to the end of the project only to realize that it is actually the beginning of the next relay to apply machine learning to this new set of digitized data. And once that is done, the next relay can become then applying conversational AI agents that can use next best action (NBA) insights from the machine-learned patterns to automatically answer incoming queries. What started as a quick sprint can transform into a long multistage relay race. With the right planning and comprehensive design from the beginning, each RPA implementation would deliver a much better outcome.

2. Design is key to a strong foundation for success

Projects that are built on deep and thoughtful design almost always come out ahead. Many implementations get so focused on getting bots up and running that they dive headfirst into software configuration and neglect the preliminary step of design. As an example, most teams unfortunately spend more time discussing which RPA software to use rather than debating which processes to automate. Analyzing the operating model and designing the target end-state, reviewing the end-to-end processes and picking the right subset to automate with RPA, and designing the automated processes to interlink back into the existing control points on each end are just as important – if not more – as configuring the actual implementation to the unique step-by-step process requirements.

Change management is the other critical but often overlooked piece of the puzzle in design planning.  The reality is most enterprise processes do not exist in a vacuum – they act upon and as a next step to processes that trigger them – and provide output as an end result into other processes that take on the follow-on actions. This means the handshakes with the upstream and downstream processes often change with RPA, and planning for, and consistently working through, the change management aspects is critical to making RPA successful.

3. Governance is single largest driver of value accretion in RPA

Many companies think that once bots are set up, they will just run in the background and operate autonomously. Nothing could be farther from the truth – the reality is bots need constant management and maintenance over their productive life. Let’s examine this a little closer. Most financial institutions that implement, for example, 4,000 bots over two years will likely pick two software providers so as not put all their eggs in one basket. More often than not, each software will have a couple of versions up and running at any one time, given scale and deployment footprints. Then there are security patches and other updates happening discontinuously across all these different versions. Finally, processes also can change weekly; interfacing applications get updated regularly; data formats evolve constantly. With these thousands of moving parts at play, a strong governance with a command and control hub is the only way to deploy RPA at scale and automate mission-critical processes that keep the lights on.

Proper command and control also addresses misses in design; for instance, when the original design does not integrate enterprise password policies, and passwords subsequently change, the bots can no longer log into the application and the automation comes to a halt. In these cases, governance dashboards can bring visibility to the issue and put a spotlight on quickly addressing otherwise hidden choke points.

In summary, RPA is a valuable automation asset in a company’s greater digital journey and can deliver great results if implemented well. Most RPA implementations though have not delivered the returns promised, and a careful review reveals some important leading indicators of success in RPA. Organizations that combine RPA with broader set of digital tools appear to realize value much faster than others. Implementations that bake in deep design and broad change management into the approach benefit from stronger business outcomes across the end-to-end process. And governance, more than any other line item on the project plan, extracts real business value from a set of automations over time. Enterprises looking to embark on the RPA journey can use these learnings to avoid the pitfalls that otherwise weigh down RPA success.

This article is published as part of the IDG Contributor Network. Want to Join?

NEW! Download the Fall 2018 digital issue of CIO