+44 203 808 8299
In a world of digitization and self-service channels, shrewd organizations are racing towards headcount and cost reduction to create competitive advantage – potentially signalling the end of the white-collar worker.
A small industry of consultancies armed with Robotic Process Automation (RPA) software are turning up at brick and mortar businesses promising huge headcount reductions. Yet in their flurry of buzzwords such as Artificial Intelligence (AI) and Machine Learning, there is scope for confusion. I am often asked to demystify the relationship between RPA and AI. For example, “Will RPA help reduce my cost of compliance?” or “Can AI replace my analysts and investigators?”
The hard truth is that key workers can, and will be, replaced by Artificial Intelligence, and it should be at the expense of RPA when reviewing the best options for your business.
To start, let’s lift the veil on some of the mystery around what these terms actually mean.
What is RPA?
Fundamentally, RPA tools interact with user interfaces intended for humans to perform tasks against an organization’s existing IT systems. Effectively, they are “scraping screens,” and then automating the entry of data based on some basic rules they follow, in the same way as a human would. So, they are robots, such as those that replicate workers on a production line, just without a physical presence. Generally speaking, however, they are not “intelligent workers.” They may use some basic AI or machine learning to help them mimic or reverse engineer the rules that the humans are following, however, they are not exhibiting AI in the same way that other decision systems do. A good example of RPA would be KYC checks, where robots pull together multiple sources of data access external systems to collect information, and then interact with multiple internal systems to complete the KYC related forms and steps.
Will Workers be Replaced by RPA or AI?
Then there is the question of whether you can replace white-collar workers who make more complex decisions; the more expensive resources that can often be bottlenecks to business objectives if their expertise is scarce. The answer is yes, through the use of AI. However, don’t assume you can just throw a “deep learning super computer” at the problem; you have to capture the expertise of the scarce resource you are seeking to replicate, and combine this with machine learning techniques.
Consider, for example, making a decision regarding a potential Anti Money Laundering (AML) alert. What would an expert user do? They would probably look at the business and the flagged transactions the business has made, and see if these are consistent with their historic behavior.
For example, take the case of a fruit exporter in Latin America making a payment to a mining supplier in Africa with connections to the Middle East. They need to ask: Why they would do this? Should we also also be concerned that a major shareholder in this supplier has a directorship of a second company where the other director is a politically exposed person (PEP)?
This seems like complex analysis, but the right type of AI can make thousands of such decisions an hour. How? By joining all the data sources together to get a view of the business, linking up transactions and directors, using peer group analysis to understand what normal relationships are, using historic cases to identify high risk geographies, and applying risk factors such as PEPs and sanctioned individuals. These are smart data science techniques, but this time informed by our very best AML investigators.
One could argue, and I would certainly be one of them, that through the augmentation of machine learning, coupled with the ability to join and analyze vast quantities of data, an “artificially intelligent robot” could replace high-end workers such as investigators, inspectors, auditors and underwriting credit decision professionals. However, this is a generation beyond traditional RPA, and is in the realm of AI.
How do the “Digitization Professionals” do it?
As governments seek to break down traditionally closed markets and introduce free competition for the better good of the consumers (PSD2 in banking being a good example), the markets are opening up to the “Digitization Professionals” of this world such as PayPal, Google, Uber, AirBnB and Amazon.
Unencumbered by old IT systems designed for paper-based or phone-based interaction with customers, these “Digitization Professionals” are embracing AI, combining it with big data, and using it wherever they can to remove the need for human intervention.
They are setting the bar for low cost of operation, customer acquisition and customer service. Essentially, they have bypassed the need for “sticky tape and plasters” approach of RPA, and are embracing AI at the heart of their systems.
So, what does this mean for RPA?
As organizations enter the race to the bottom for headcount and cost, there is a limited window of opportunity to give yourself a head start by spotting, and taking advantage of, RPA ROI opportunities.
After all, ripping out and replacing big IT systems can’t happen overnight, so as bizarre as it may seem, spending money now to put in an IT system to talk to another IT system through a clunky UI interaction can actually make sense.
However, this must not be at the exclusion of striving as quickly as possible to the end goal of removing repetitive human based tasks by design. Furthermore, to win the race it will also be critical to introduce a programme of intelligent worker replacement in parallel.
So, take a critical look at your RPA program: is the ROI case strong enough that you could you skip straight to the intelligent worker AI solution, such as in the AML alert investigation example? Finally, make sure you are making the most of your big data, representing it in the right way, and applying the appropriate flavor of artificial intelligence treatment.