Breathing new life into RPA

You can’t move for people talking about artificial intelligence and how it can be applied to virtually any aspect of how we work and live. It’s the digital technology buzzphrase of the era, even if we can’t quite agree whether it’s an already existing technology or the stuff of science fiction.

 

There are, however, more mundane technologies that have already produced jaw-dropping results but for some reason are seen as, well, a bit dull in comparison.

 

A case in point is robotic process automation, or RPA – a type of automation that has been around for more than a decade now. Being a technology that automates rules-based, repetitive (and therefore boring) tasks, it has always struggled to capture people’s imagination in the same way AI has.

 

A strong track record of freeing up the human workforce

 

However, there is no denying that RPA has already proven its merit. By automating routine tasks such as customer data entry into core banking systems, document collection and verification, RPA has reduced the client onboarding times of banks from days to hours – or even minutes.

 

RPA bots can extract relevant data from loan applications, automatically verify documents and evaluate loan eligibility in a fraction of the time manual processes require.

 

RPA has also revolutionised supply chain management by enabling the continuous monitoring of inventory levels and preventing stockouts by leveraging enhanced demand planning through gathering sales forecasts, market trends and weather data.

 

Rather than celebrating its achievements, however, experts tend to focus on RPA’s shortcomings and its supposed imminent demise. Being rules-based, RPA can’t deal with exceptions, lacks the ability to adapt or learn and can only be fed structured inputs.

 

Some dismiss RPA as a shortcut to automation, which only businesses resisting the in-depth transformation of their core system opt for. Indeed, RPA leverages the user interface to automate tasks without any deeper system integration, where the RPA layer simply adds an interface on top of the legacy application.

 

RPA in action resembles an invisible employee that clicks a mouse, opens websites, taps information into databases or copy-pastes data from one app into another. Similar to other digital transformation projects, RPA adoptions had their fair share of failure, with 30 to 50 per cent of them reported not to have met the business’s intended objectives in 2019. Early in the technology’s lifecycle, only 13 per cent of companies were deploying RPA systems, which soared to 74 per cent seven years later in 2022.

 

Although the scaling of RPA systems originally presented serious challenges, in a 2022 survey intelligent automation company Blueprint found that the top 10 per cent of the 400 middle and top-level leaders from companies in North America and Europe with 1,000 to 10,000 or more employees had 251 automation in their estates, while the average number of automated processes among all respondents stood at 120.

 

A new lease of life

 

But while the rise of gen AI has sparked conversations about how it will inevitably make RPA obsolete, what we are actually witnessing is the cross-pollination of (relatively) old and new technology.

 

As cloud-based RPA platforms such as UiPath, Automation Anywhere or Blue Prism are designed to run on major cloud providers, their platforms can integrate AI tools into RPA systems. These integrations give RPA a new lease of life, with capabilities that previously constrained them and made them inferior. Augmented with NLP and ML, they are able to identify patterns and extract information from unstructured data sources such as images, audio and video.

 

AI can power them to learn from interactions, as well as make decisions in real time based on patterns and predictions. AI-driven technologies also enable RPA to understand context and perform sentiment analysis.

 

These intelligent RPA-based solutions will be capable of finding scattered, unstructured data and feed that into rule-based automated systems to carry out routine tasks, handle exceptions or even make informed decisions.

 

Mind over matter?

 

The analogy that describes what RPA and AI bring to the table when combined is that of the body and the brain, where the former – with its ability to interact with various systems and data sources to perform pre-defined tasks – is the workhorse, while the latter is tasked with decision-making and complex problem-solving.

 

While AI can guide RPA bots through processing complex tasks, RPA handles all the repetitive tasks, including the collection and preparation of data that can be fed into AI algorithms for intelligent analysis. 

 

AI-enhanced RPA solutions can also serve as alternatives to native AI deployments in a number of use-cases, such as predictive maintenance, personalised customer experience through sentiment analysis, or fraud detection.

Given the high level of RPA adoption, thousands of companies may go down the path of infusing their bots with AI-capabilities to make the most of their original investment.

 

Another business case for integration is that RPA is easier and quicker to implement than pure AI models. It can also seamlessly connect to legacy and disparate systems without APIs, which constitute a substantial part of AI deployment expenses.

 

Purpose-built AI systems based on neural networks, where AI is pervasive throughout the whole architecture, are naturally more advanced than AI-powered RPA. But their excellence at automating tasks and processing sometimes comes with high risks – and at a cost that small and medium-sized companies in particular can feel is prohibitive.

 

So for companies that have already invested in RPA, or for those that can’t afford to take an early plunge with advanced AI systems, tapping into RPA injected with AI capabilities – (embedded AI) can represent an easy win, or even a force multiplier.