Robotic process automation (RPA) technologies are a cost-effective and non-invasive way of automating repetitive business processes. Apart from achieving process automation, RPA also digitises manual processes in the organisation and captures valuable process data. Applying modern data analytics tools to this RPA-generated data can help you unlock further process insights and gain an unprecedented understanding of your organisational structures and workflows, simulate process changes, and identify precise process improvement opportunities. This approach removes subjectivity and ‘gut feeling’ from the evaluation of manual processes and enables businesses to make transformational change decisions to accomplish process targets and strategic goals. Let’s take a closer look at what’s already possible.
First a quick recap: as we explained in the introduction to this blog series (Robotic process automation: the basics), RPA is a software solution that can automate ‘robotic’ (very repetitive) human processes that run on a computer. It’s also a cost-effective, quick and non-invasive way of digitising manual processes without having to replace or modify existing software. RPA is already a megatrend in the large banking industry, enjoying great success in the marketplace, not least because it’s cost-effective and yields a rapid return on investment. In addition to low cost and rapid implementation, which are great incentives in themselves, organisations that are able to unlock the value of the data RPA generates can stand out from their competitors by being truly transformative.
Automation = Digitisation = Big Data
Data is the most precious asset in this digital age, and businesses have been seeking ways to capture, enrich and generate value from their data. Based on our Big Decisions Survey 2016, 92% of global businesses are already somewhat or highly data-driven. However, if a business process is manual, there is little or no data to be captured and then used by decisionmakers.
From our experience with RPA, the software not only automates processes, but digitises them as well. Automation equals digitisation, which enables big data collection from those previously manual processes. We’re going to look at three of the most promising data analytics approaches that can benefit from the data captured by RPA. In ascending order of technical sophistication they are: process mining, process simulation, and machine learning.
1) Process mining
Until now, the methods for understanding business processes and seeing how they’re running have been fairly basic. You could look at key performance indicators and process documentation, for example, and conduct interviews with users. But KPIs are too high-level to get a comprehensive picture, process documentation will only tell you what the process is designed to look like in an ideal world, and interviews are often subjective and limited by the interviewee’s knowledge. Once a process is automated by RPA, process mining technologies can be deployed to visualise the entire process on the basis of data generated by the RPA technologies, enabling a much more insightful understanding of the process than any of the traditional approaches.
For a start, RPA creates a trail of incredibly detailed data that spans processes that were previously manual, fragmented and subject to human factors such as the wellbeing of the staff performing the task. A good example would be an accounts payable process, which typically consists of a number of steps, some manual and some automated: a clerk sees an email with an invoice attached, enters the invoice details in the system, requests that their superior approve the payment, enters the payment into the online banking system, then passes it on to another approver who reconciles all the supporting documents before finalising the payment. Under the old, pre-RPA set-up, the data on this process could have come in a variety of different forms, and you would have had to look in different places to find it: email inboxes, the procurement system, and an online banking system. If one of the employees involved in this process was preoccupied with another task or unwell, the data collected might not be a true reflection of the normal process. RPA, by contrast, generates data for every step in the most efficient way, giving you a seamless, detailed data trail that can be used to mine and visualise the whole process.
As always, pictures speak louder than words, so check out this example of the same process mapped in four different countries:
All figures: Procurement process benchmark, X-territory
The process mining technology literally visualises the actual process − not in an ideal world, but in reality. You can calibrate the tool to discover what’s happening most of the time in the processes, or to look at just the outliers. You can visualise user interactions from the highest level right down to identifying individual transactions in the process. You’ll have an actual representation of the process that allows you to see at a glance what’s working smoothly and where the bottlenecks are. From the examples above, it’s immediately apparent which territory will be the role model when it comes to optimising the process. Equally, you can localise individual tasks in the process chain that are consistently weak and address them specifically.
Another interesting point is that the data provided by RPA doesn’t just enable you to visualise processes, but also gives you fascinating and valuable insights into how the people involved in the process are interacting with each other and with the RPA technology. Used intelligently, this type of data analytics combined with RPA can help you take the robotic components out of people’s work and free them up to channel their human effort and intelligence more productively. Using the accounts payable process as an example, process mining can show process seasonality, enabling the business to identify opportunities for better production planning to flatten the season – a solution that would not have been evident without the ability to harness RPA-generated data.
For more details, check out the blog by Rafael Accorsi, PwC’s Process Intelligence Leader.
2) Process simulation
From visualising a process using data generated by RPA, it’s only a short step to running a simulation program using the process data to identify process dependencies and simulate the concrete impact of different scenarios. When you’re planning changes in your processes (for example team restructuring, using a new supplier, introducing new software, etc.), you can use process simulation to predict the impact of the changes by modifying specific variables in the simulation model. In other words, you get to simulate a process change before actually making a major investment.
In a manual and complex business process, it can be very difficult to work out the consequences of even small changes. So businesses often rely on gut feeling when making process changes. This frequently leads to new and unexpected process issues, for example bottlenecks in downstream tasks. Because bots (developed from RPA) will execute processes strictly on the basis of predefined rules and faithfully record every activity, a simulation model developed with RPA data takes the guesswork out of process change decisions. Do we want to switch to a provider in a low-cost region? Run a simulation. Does it make sense to acquire Company X? Use RPA-generated data to model the impact of the purchase. If we change production process or achieve a 5% improvement in speed, what will be the dollar impact on working capital? Ditto. Sometimes relatively unspectacular changes can bring about significant bottom-line improvements. Process simulation on the basis of RPA data makes it easier and less risky to implement the changes that count.
3) Machine learning
If you’re tempted by the potential of RPA data as a basis for process mining and simulation, machine learning is where it gets really exciting.
A simulation program is designed to answer ‘what-if’ questions by modelling future processes based on scenarios. This means that process simulation on its own is very useful when you already know what you want to change in the process. Machine learning comes in when you’re looking for ways to optimise the process but don’t know exactly what you want to change. Machine learning can be used to ask open questions, for example about how to make the current process run faster. There are various machine learning algorithms that are good at telling you what elements in a process (for example supplier, material types or time of year) can drive what kinds of outcome (for instance lower process cycle time). If you can feed RPA process audit trails into various machine learning algorithms, you can get prescriptive solutions to improve your processes.
For example, machine learning might come up with the suggestion that ordering material X from supplier A in the week of Christmas instead of the first week in January will result in a 50% improvement in order fulfilment in January. You could change the RPA robot setting in line with this suggestion to make sure orders to the relevant suppliers are placed during Christmas week, while your staff are on vacation.
Apart from its ability to generate simple correlations, machine learning combined with today’s computing power is increasingly capable of identifying unknown relationships within multiple business processes. For instance, it can potentially correlate procumbent processes with sales processes to analyse directly what supply chain management actions need to be taken to improve sales. All these complex analytics can be made possible using the comprehensive and detailed audit trails kept by RPA technologies.
RPA is already used in machine learning. At PwC we have brought our machine learning and RPA teams together. The result is a bot we call the Data Science Machine. The Data Science Machine harnesses the concept of RPA to autonomously experiment with a library of machine learning algorithms with different process data elements (features) to identify the best prediction methods. So it’s an RPA bot that can learn the data generated by another RPA bot from actual business processes. It might sound like science fiction, but it’s already a reality.
In a nutshell
Robotic process automation on its own is a rapid, low-cost and non-invasive approach to automating existing processes. But there are even greater business benefits to be had if you combine RPA with state-of-the-art data analytics and machine learning technologies to visualise, simulate and autonomously identify improvement opportunities in your processes. So after achieving early process improvements such as reduced process cycle time, increased throughput capacity and cost savings via RPA, you can further leverage the data it generates to truly transform your business.