Following my initial review of how financial services firms are getting on with artificial intelligence (AI), I was interested in finding out more about adoption of AI and related technologies such as RPA and the technical challenges companies currently face, as well as near-future evolution at sector organizations.
According to the study Digital Pinnacle Enterprises by analyst firm Everst Group, of the 55 financial services organizations polled, 16% adopted AI most effectively while 89% of those had already invested in AI in some form or other.
The most common AI uses in the sector were for sentiment analysis for marketing, personal finance virtual agents and financial and advisory virtual agents. Another Everest study of 12 property and casualty insurance companies last year showed that 29% were running AI pilots and 50% were seriously considering it.
RPA is much more pervasive than AI in the sector: Everest data shows that banks and financial firms account for 40% of the RPA independent software vendor market. In the insurance study, 93% of the sample had already deployed RPA - by comparison, some 29% had implemented AI. There is a number of ways in which RPA can support automation in financial services, according to Sarah Burnett, research vice president at Everest Group. She says:
I have seen a variety of operational challenges addressed with RPA that in turn can lead to successful business outcomes, such as dealing with an influx of applications for financial services as a result of new government legislation, a problematic process that led to high levels of staff attrition in a banking offshore shared services center.
Accelerating customer on-boarding is another key benefit. Citing anecdotal evidence, Burnett mentions a bank that has reduced this process from 16 days to just 9 minutes with RPA.
So where in the RPA technology adoption curve are most financial services firms today? According to Burnett, sector organizations were early adopters of RPA and Everest Group data suggests that they account for almost 40 percent of that market, with many setting up RPA "centers of excellence" and tapping into AI-based technologies to expand the scope and scale of process automation.
As organizations evolve in their RPA strategies, most challenges are related to going from a proof of concept to full scale roll out. According to Burnett, this is why it is important to build proofs of concept to be realistic rather than too simplistic. Conversely, there are also concerns related to the scale of these projects. Burnett says:
Over the years, as deployments become larger, enterprises will need to invest in good robot control solutions for very large-scale and mixed vendor RPA deployments or they will lose track of what they have and what the robots are doing.
The issue here is that there aren’t many solutions that would provide views of robots from multiple vendors on one single dashboard, according to Burnett, who points out that only a few service providers offer these types of control environments enabled with AI. She adds:
This would mean a high-level view of groups of robots operating across the enterprise with drill-through capabilities to see the status of individual robots when needed, and the ability to stop, start or reboot failed robots and address errors and exceptions.
Other technical considerations for financial services firms using RPA will typically include selecting which platform to use and what features it offers around aspects such as security, auditability, robot controls, version control, ease of use and maintenance. In addition, software licensing is another aspect to be considered. Burnett adds:
RPA robots login to business systems the same way that people do and so, for example, they may need SAP or Oracle licenses. This sometimes clashes with human resources policies that dictate new licenses are only bought for new staff.
The fact that RPA systems can process a lot of transactions very quickly may also result in a slowdown of enterprise systems, says Burnett. This means that getting the runtime environment right is important - and many organizations overcome these issues through trial and error.
Speaking of human errors, there is also the cultural aspect of challenges often faced by financial institutions working with RPA. According to Burnett, getting buy-in from the teams that would be affected by robots and senior executives to drive the automation agenda in a collaborative and supportive way is a major sticking point. She adds:
Getting IT on board from the very beginning and addressing project funding, ownership and governance can also be a challenge.
Going forward, Burnett predicts that in a year's time, addition of AI to RPA platforms will be more commonly seen, in examples such as self-diagnostic healing robots that do not have to wait for a human to fix exceptions.
In addition, further developments for RPA in the coming months foreseen by the analyst include more process-related analytics and data mining that will enable RPA systems to recommend what processes should be automated next.
There is a lot more being done on robot controls as well. I expect to see new solutions emerge offering specialist RPA for banking and insurance software or extended libraries and bot stores for these sectors.
As the cost of building RPA POCs drops considerably, it may be tempting to sit back and watch the efficiencies brought by automation multiply. But as in most emerging technologies, there is no relaxing for financial services firms when it comes to RPA and this is especially true at the moment of taking trials to full production. While the business case for using automation to take on repetitive tasks is clear, further developing RPA can introduce various technical challenges and longstanding business practices and culture can be slow to change.
The combination of RPA and AI - despite being potentially a lot more costly than standard RPA - is looking like the new normal in financial services automation. But is there such a thing as a journey from "dumb" robots to the smart, self-learning robots? Is AI imperative if RPA is to realise its full value? Burnett and other experts argue that the transition from RPA to AI is a natural step but it is not a required step. That is because both technologies address different needs, but very often complement each other - so businesses really need both.
Vendors understand this and are adding more AI capabilities to their software, so organizations are going to get cleverer robots anyway. It will be interesting to watch how financial services firms, the early adopters of automation, take on developments such as RPA as a service enhanced with various flavors of AI to extend the scope and scale of automation initiatives.