When Artificial Intelligence initiatives fall short, the blame is often placed on a skills gap. But there's more to it. Does your organization prioritize these three foundational AI pillars?
How to get started with Robotic Process Automation (RPA): 4 steps
Robotic Process Automation tools may make the how to get started easy, but you must still figure out the what, where, when, and why. Consider these steps - from identifying RPA opportunities to measuring success
It’s relatively easy to get started with robotic process automation (RPA), except when it isn’t.
Various RPA software tools promise to help you launch a bot quickly, sometimes with little to no coding skill needed to get it up and running. While these platforms shoulder some of the heavy lifting when it comes to the “how” of RPA, you and your team will still need to figure out the “what, where, when, and why.”
[ Want a primer? See our related post, How to explain Robotic Process Automation (RPA) in plain English. ]
Get started with RPA: 4 steps
That can feel a bit mystifying at the outset. Where do you begin? How do you identify and implement the first phases of an initiative that can become both sustainable and scalable? How do you measure success or failure? We asked several RPA leaders those very questions. Here are four tangible steps to consider when getting started with RPA.
1. Don’t start small with RPA: Start micro
It’s conventional wisdom when starting from scratch with a new technology: Start small. With RPA, the wisdom needs some modification: Start smaller.
“It sounds obvious, but most people get this wrong: Automate the small, easy processes that you do a lot,” says Antony Edwards, COO at Eggplant. “Too many people start by trying to automate the really long, gnarly process that only happens twice a month. This is hard, takes a lot of time, is brittle, and has low value.”
In fact, “small” might not be quite the right scope when you’re just starting out. Instead, consider “micro RPA,” which Edwards describes as “automating specific, small actions that take two to five seconds” to complete.
“This is more robust and will get you going,” Edwards says. “Then build up.”
It can be useful to think in terms of single tasks rather than large, end-to-end processes when you’re getting going with RPA. In fact, RPA on its own can’t always necessarily tackle end-to-end processes in their entirety, particularly if there are any steps in that process that require creative or complex decision-making, or that are otherwise not a good fit for RPA.
2. Look for high-value RPA candidates
As Edwards points out above, that byzantine process that only runs once or twice a month probably isn’t your best fit for RPA, and it certainly not a good candidate for square one.
Instead, look for tasks where you see (and can articulate to others) a valuable business outcome via automation. The driving question of your evaluation process shouldn’t solely be: Can we automate this? Rather, if the answer is yes, then you need to ask: Should we automate this? You can spin up all sorts of variations of the same, such as: What will we gain by automating this? If you can’t find a clear answer, that should give you pause.
“Many organizations have a lot of ideas for automation candidates, but only a few of them ask for the real business value,” says Tom Thaler, senior product manager, ARIS at Software AG. “This prevents successful RPA implementation and can burn all potential to grow.”
We’ve got an in-depth article for you on how to identify RPA opportunities. As a starting point, though, map the “micro RPA” approach to highly repetitive steps in a larger process that are currently causing a lot of pain.
“It’s essential to focus on the real weak points and bottlenecks of the end-to-end processes right from the start,” Thaler says, adding that established techniques such as process mining or business process analysis (BPA) can be useful ways to reliably smoke out bottlenecks.
“This allows you to not only identify good automation candidates, but also to predict the effect of automation on the business outcome,” he says.
3. Capture the process at a keystroke-and-click level
RPA is essentially software that can automate certain computer-based tasks that would once have required a person to do manually with a keyboard and mouse. So, if you want to create a robot or bot – the popular terms for this kind of software, not to be confused with physical robots – you’re going to need to show it how to do its job.
“You then need to capture the process with all details on a click level, as this is required for implementing the robot,” Thaler says.
RPA bots on their own (unlike machine learning tools) can’t figure it out over time. So shortchanging this phase could render your project DOA. The fundamental necessity of this phase can create a different issue, however, one IT that leaders will want to keep in mind: Based on experience, Thaler estimates that around 70 percent of RPA project resources are spent on this stage.
“This phase may become a bottomless pit,” Thaler says. “The reason for this is you’ll always have bias and [missing] information if you [only] talk to people, which leads to a number of workshops and iterations until everything is covered.”
Certainly, talking with your team is a good way to identify bottlenecks and other pain points. Thaler points to task mining or process mining as a more complete approach to RPA discovery, one that doesn’t solely rely on anecdotal input.
“[This] can prevent bias and allow you to learn from the daily work of your employees,” Thaler explains. “From there, it’s easy to implement the first robot – and also the second, third, fourth, and so on.”
It’s important to note that RPA won’t magically improve a broken process; it will just automate – or attempt and then fail to automate – a broken process. Process optimization needs to happen in the real world before you implement RPA.
4. Define how you’ll measure RPA performance
As with most IT initiatives, your RPA program is more likely to founder or outright fail if you have nothing in place to measure its progress. This is not something you should figure out months or years down the road.
“Build an enterprise business case to determine ROI for your RPA efforts,” advises Chris Huff, chief strategy officer at Kofax. Huff has seen success, at Kofax and previously while implementing RPA while leading Deloitte’s public sector robotics and intelligent automation practice, with a four-pillar framework for a holistic business case. It should cover:
- Strategic alignment to the larger enterprise strategy
- Workforce impact
- Operational metric impact
- Financial profile/impact
Huff is also a proponent of the Center of Excellence approach for ensuring long-term RPA success. In your baby-steps stage, you should at least be planning to capture your results in a meaningful way.
Workforce impact alone is a good motivation not to skimp on this step early in your project. Automation anxiety is real. Communicating quantitative and qualitative RPA results will be key to alleviating it, as well as for getting the kind of buy-in necessary to grow the program beyond its infancy.
“One of the most important things to do when the robots are in place is to capture and communicate the benefits achieved,” says Thaler from Software AG. “This includes collecting all data that is relevant from a business perspective, calculating and visualizing corresponding KPIs, and presenting them to the management. This is the key to a sustainable management buy-in, ownership, and budget to grow with RPA.”
[ How can automation free up more staff time for innovation? Get the free eBook: Managing IT with Automation. ]