by Greg Simpson

The many hands of a robot

Oct 10, 2018
Artificial IntelligenceMachine LearningTechnology Industry

Deploying robotic process automation for modern applications.

Greg Simpson and others discussing at the Innovation Station
Credit: Synchrony Comms

There is a lot of discussion in the industry around AI and the coming robot invasion. I’ve talked a great deal in my blog about some of the different forms of AI including machine learning and chatbots. One of the rapidly growing areas of AI today is robotic process automation. Robotic process automation, or RPA, usually focuses on the highly repetitive tasks that don’t require as much cognitive processing or human judgement, but rather are often “boring” tasks.

Some years ago, the focus in IT was the implementation of large scope (and hence large dollar) systems for projects like enterprise resource planning. These were expensive, long term projects. In many of these cases, humans became the glue between the various large scale system—people would pull data from system one, massage it in excel, upload it into system two. It was cheaper for the humans to absorb this role as facilitator between the many different large monolithic systems than to automate it. In other cases, humans needed to “check” things… compare, contrast, report. Again, humans were called on for these more glue like tasks. RPA is about redistributing these mundane tasks to robots. Let’s focus our human experience on doing the difficult cognitive work and reduce the time we spend on rote tasks.

In the world of AI, if machine learning is the robot’s brain, RPA represents the hands of the robot. Examining data, moving it around, and enforcing controls. Forrester estimates that the RPA market will be $2.9 billion in 2021.

At first blush, RPA sounds a lot like the many digitization efforts businesses have embarked on over the years. The difference is simple; the technology has matured to the point where bot development is fast and more proficient at handling more complex interactions. This real power goes beyond lower cost rote robotic automation; however when you start to combine machine learning models with RPA, suddenly the brain (machine learning) can provide direction, enabling the hands (RPA) can do more complex, cognitive tasks than before. This confluence of traditional digitization and AI drives smarter automation solutions.

In my view, here are some of the things you need to consider before you embark on your RPA journey.

  • Controllership is key: Well-designed robots can deliver real tangible benefits to almost any company. Poorly designed robots have the potential of wreaking havoc with your data, processing errors faster than ever before. Pick an approach, create a COE, and take the time to get your controllership right up front. It will pay off in the long run.
  • Understand the human implications: Automation is often implemented as a productivity play, and this often means impact to people’s jobs. Do you need to re-skill workers? How do you plan to retain the IP for the work that the robots don’t do? Automation is just one step of the journey that involves your entire workforce – including humans.
  • Capture the benefits: Usually, RPA only automates a fraction of a role – a few tasks. This can lead to a company implementing many robots, but not capturing the true benefits. Sometimes adding this efficiency into the system is all that is needed for a rapidly growing company, but if the company is looking for cost out, they need to understand how to deal with fractional headcount savings and think holistically about re-engineering the process, not simply implementing “a bot.”
  • Bring the brain into the mix: You can greatly improve your benefits when you start to combine AI technologies together to provide a level of cognitive thinking to the robotic process automation. A machine learning model may simply direct the robot down one of a few paths, but this is a significant step forward from executing the same steps every time.
  • Use your technology architecture skills: Use your COE to create “metabots” to do common tasks. Don’t program basic things like logging into a core system into each bot, leverage metabots to speed your bot creation the same way you leverage microservices to speed your development.

In summary, RPA isn’t a magic bullet. It requires the same diligence as many technology projects. However, when executed properly, today’s RPA technology combined with a little cognitive AI can deliver significant benefits. Remember, you are essentially introducing a robotic workforce, but once again, this robot workforce augments the human workforce – it doesn’t fully replace it. 

Reach out to me and let know your robot workforce stories!