IPA = RPA + MACHINE LEARNING

What’s the difference between the terms ‘RPA’ and ‘IPA’? The simplest way to think of it is that RPA tasks are extremely rule-based. They’re typically very high-volume and performed at all times of day. These are jobs where quality is important and tasks follow strict parameters, and if those parameters change, manual intervention is required. IPA adds machine learning to core RPA to create a smarter form of automation that is great for rote tasks but can also learn and improve as it goes along.

RPA is very formulaic and it will fail if the environment is full of exceptions. Machine learning, on the other hand, can predict and handle such exceptions, which allows IPA to deal with certain grey areas. For example, optical character recognition won’t work as well with handwritten text and bots cannot help in differentiating a handwritten zero and a letter O, but IPA has the capacity to account for discrepancies of this kind.

To get a desired outcome for your business, determine which technology applies where, and then bring in the right teams for the specific business issue.

IPA AND RPA: WHAT ARE THEY GOOD FOR?

The key advantages of process automation tend to be in cost savings and efficiencies. Cost is either the number-one or number-two desire, but people also want to know how they can reduce human effort involved in a process and how to upskill teams to perform value-add tasks. From the evidence we have seen through our engagements, accuracy improves by up to 99% because a bot won’t make a mistake if its parameters are set accurately.

Many of the tasks for which process automation is most effective are clerical. Think of accounts payable, where a person will take invoice information, such as a purchase order, and match it to what’s in the system. A bot can do that in a systematic way that yields high levels of accuracy and handles high volumes. Another is reporting. An individual might sit at a laptop, open an application, go to the right place in the application, copy data, open Excel, locate the correct field for the data, and paste it in. A bot can do all that but far more quickly and with far higher levels of dependability and accuracy.

But there are more advanced use cases. Consider email processing, where we can train systems to extract content and context from messages to find meaning by analyzing and parsing text. For example, when a utility consumer sends in a meter reading, that information can be quickly and accurately sorted by a bot to go to the right place.

Most automation assists and augments a human agent. Chatbots and virtual assistants can help a customer to check an account balance or support insurance claims processing. Finance and accounting are often a great place to apply IPA initially because tasks are highly systematic and rule-based, but any transactional, rule-based process with high volumes is an equally good candidate.

GETTING STARTED

Firstly, don’t simply pick up a random sub-process to automate, but instead identify a larger process-driven area of business, such as the front office and take a task such as processing emailed customer queries. Consider it from the point of view of a business problem such as the need to accelerate service issue resolution in a call center, and then see what you can do to improve the process you already have in place. You may find you are able to eliminate unnecessary steps or tweak to optimize current working practice.

And only after that should you look at what IPA and RPA can do. This will usually be to accelerate a set of high-volume, repeatable manual processes that are probably currently error-prone, or use machine learning to deal with processes where data science and analytics are relevant. With IPA you can make these close to error-free through a series of systematic steps. That’s the beauty of IPA: it’s a way to combine software robots with cognitive technologies that mimic the way humans think and behave–the best of both worlds.

Ideally you would perform an end-to-end assessment in order to discover which processes within the business would benefit from IPA and RPA. Then you’d look at each one of these processes individually, including who works on it and who manages it. But while this is the gold-standard approach to understanding your business’ suitability to process automation, it isn’t realistic in all cases. That’s why we often perform proof-of-concept exercises to show that automation works in a specific environment, rather than immediately focusing on its capacity for cost reduction.

CHALLENGES AND MISCONCEPTIONS

Most of the C-suite has got the buzz over RPA and IPA. In a number of cases, business has taken the lead as well, as automation gets more embedded in the digital change. But even where the automation activity has been led by line-of-business heads, organizations have to involve IT deeply and ensure that areas such as security and compliance are considered and that dependencies can be handled.

In our experience, one common misconception about IPA and RPA is that they can automate 80% of an entire process. This miscalculation can lead to mismanaged expectations and flawed ROI calculations. Try to be realistic about what automation can deliver: when supporting a process that involves a few workers, it might not yield a massive benefit. For example, in an accounts payable process, we’ve seen that using RPA for checking invoices against the system can reap savings of 60% to 90%. However, the net impact on the overall end-to-end process is a saving of 20% to 25%. But when applied to more extensive processes that involve many people, the saving in time and cost and efficiency improvements can be substantial.

IPA and RPA should not be applied to automate a broken process. To do so will simply make a bad process faster. Another mistake clients sometimes make is to select an automation product as part of an enterprise-wide strategy and then to drive an automation agenda based around it.

Such examples demonstrate why it’s important to review all processes end-to-end before implementing automation. You also have to think collectively about platforms, technologies and ecosystems to take a broader view on how they will combine most effectively. For example, a bot has the same IT requirements as a human being: it needs to be permitted access to the right apps and must go through security checks. So, before moving it into production, it needs a login ID and password.

Sometimes issues arise because teams are not in sync. For example, one company developed a bot for a small process and worked with our consultant to put it into production, but unbeknownst to the consultant the app being automated was soon to be retired–wasting effort. Ensuring that all relevant teams and stakeholders are involved will avoid this sort of confusion.

COMMUNICATING THE PLAN

Change management and communications are critical at all levels. Business case documents for upwards reporting will help show what’s working and what’s not, reinforcing the argument for investment. A successful communications program highlights inefficiencies and the benefit of removing them, such as unproductive office hours when people could rather be deployed to do value-added work, take training or engage in new tasks. It’s also important to involve those who work with the processes in the design of the change program. These teams often have the best ideas.

Allay any fears about job losses. Automation aims to assist agents–not replace them–so that they can be more efficient and focused on work best handled by humans. Here, it’s important to have a plan and budget for training, upskilling and reskilling.

SUMMARY

As with any early-stage technology, process automation is changing quickly. By adopting the ideas outlined here and staying on top of future best practices, companies can more efficiently handle rote tasks, redesign and optimize processes, and put their people to better, more productive use.

As you explore and embrace the many developments in process automation, remember these key points:

• The technologies and benefits of process automation are growing fast. Organizations should already be assessing how it might apply to them–or risk being left behind by faster-moving competitors.
• Bringing in process automation experts can help. By enlisting the support of third-party experts with the skills to use this new generation of tools, firms can move more swiftly than those that go it alone.
• Thinking holistically saves time. While seeking quick wins might be tempting, avoid automating processes that are already suboptimal. Assess the entire process before considering how automation can help; automation can’t make a badly designed process good but it can provide speed, consistency and accuracy.
• Build a team that has broad skills. Process automation might ultimately touch many elements of the organization, so invite all relevant stakeholders to join the team to avoid potential roadblocks.
• Be realistic. Don’t believe that new technologies are ‘silver bullets’ that offer a universal panacea. Automation can accelerate and improve process elements, not entire processes.

Our thanks to Jeeshu Ganguly, Senior Director, Cognizant Digital Operations, for his help in compiling this edition of The Big Think.