I can’t help but notice that while RPA (Robotic Process Automation) is normally considered to be a good companion of the Business Process Management Automation, it can easily become an adversary of a quality BPM. Many business functions subject to RPA are notoriously inefficient process-wise. And while it is very tempting to replace slow human beings with fast robots, the bots will learn and inherit all inefficient and ass-backward ways of doing things from people without questioning anything (yes, the best AI is not that sophisticated yet).
It may seem like not a big deal (we are still executing faster and cheaper, aren’t we?), but replacing a human with a bot makes the process execution very opaque and process inefficiencies free from critique and revisions – a robot just does what it was told or learned to do.
“Companies may quickly get caught up in automating the wrong processes, non-optimized processes, or even too much of a process” (UiPath on why RPA deplayments fail)
UiPath talks about implementation and everybody tends to talk about implementation and how to make a success, because RPA has a great promise of ripping the cost benefits from the get go. But about its later stages? How sustainable a good RPA deployment is?
Obviously, it is not advisable to automate a bad process whether through a BPM App or an RPA. But an RPA has the most potential of turning even little process inefficiencies into big problems in time. Why? Because an enterprise entertains changes on a regular basis. As difficult as it is with humans, adopting those changes by bots is likely to be even more challenging, especially when changes are not managed proactively or effect RPA’d functions indirectly and inconspicuously. A target function being a result of Machine Learning is not like a script of a traditional screen scraping, and cannot be easily visualized and modified according to changing needs; it will require re-learning or more learning So even a very well implemented RPA is a time bomb unless properly maintained and proactively educated.
Also, contrary to other means, learning from a human directly tacitly discourages documenting. But if things go wrong, robots go crazy and no knowledgeable humans left to take over the failing process execution, the process documentation seems to be the only fallback for the under-performing technology.
RPA vendors warn that an RPA adoption is a “journey” and requires planning, good foresight as well as maintenance. Nevertheless, the emphasis is still on deployment. The aforementioned concerns are somewhat muffled, even though they are so well understood in App Dev and Business Process Management practices.
Bottom line: Robotic Process Automation is in its hype but it must rely (like nothing else) on rigorous Business Process Management discipline and continuous improvement in its broad sense to yield sustainable results. RPA promises immediate gains in efficiency and productivity but being new, its long term effects are not well understood and can lead into significant problems caused by the opaqueness of RPA mechanics and likely lack of attention to keeping it in sync with business changes.