AI: Virtual assistance
Written by Graham Buck
The prospect of robots replacing humans for some tasks has moved from science fiction to reality. Graham Buck considers how artificial intelligence and robotic process automation is impacting business
Even by the its usual standards, the Daily Mail’s headline on Bank of England governor Mark Carney’s speech in early December was particularly alarming. “Robots to steal 15m of your jobs, warns bank chief” the Mail informed its readership. Although other reports were less alarmist, the Bank governor did indeed suggest last month that many professions were set to be “hollowed out” as a result of the rise of robotics and artificial intelligence (AI) and that professions such as accountancy could even see the need for humans eliminated.
Certainly for the banking sector, whose members are attempting to fend off competition from a host of fintech start-ups steadily encroaching on their territory, using so-called ‘robo advisors’ for services such as wealth and investment management makes sense. HSBC was among the first off the starting block; 18 months ago the group announced that its global workforce would be reduced by 10 per cent – around 25,000 employees – by the end of 2017, and it was allocating a US$4.5 billion budget for transforming its operations, of which US$1 billion would be earmarked for accelerating digital and automation programmes.
Other banks, such as Switzerland’s UBS, have decided that robo advisors are the way forward. In November, UBS launched its online wealth manager SmartWealth, which offers clients with as little as £15,000 the opportunity to access its investment expertise.
Before its introduction, the investment threshold was previously £2 million. Belying the perception that banks are slow to change, the new platform was developed in just one year.
Many tech industry analysts expect that the year ahead will see many more firms across the financial services industry move beyond pilot schemes and begin automating processes, using AI and robotic process automation (RPA) tools. Digital business consulting and technology services firm Synechron has dubbed 2017 “the year of the chatbot”, with banks and insurers introducing computer programmes to interact with clients and conduct conversations. Royal Bank of Scotland (RBS), which started phasing in AI for its small business operations after a successful two-month trial in early 2016, added an online chatbot dubbed ‘Ludo’ last month to answer customers’ queries.
“By combining hands-on service with some element of automation – alert notifications and auto-diallers, for example – financial institutions can manage customer relationships easier and with a degree of personalisation,” reports Synechron. “Expect to see a whole range of customer service and virtual assistant bots going live in 2017 to enhance banking, trading and insurance interactions.”
Indeed, the first week of January saw Japan’s Fukoku Mutual Life Insurance announce 34 redundancies, with employees being replaced by an AI system based on IBM’s Watson Explorer, which according to reports incorporates cognitive technology that can think like a human.
The technology is able to digest the contents of medical certificates, as well as unstructured text, images, audio and video and use the data to factor in the length of hospital stays, medical histories and surgical procedures when calculating payouts to policyholders. Among other benefits is an estimated 30 per cent increase in productivity, with Fukoku anticipating a return on its 200m yen (£1.4 million) investment within two years, with ¥15 million annual maintenance costs more than offset by savings of ¥140 million.
It’s easy to imagine the technology steadily – and swiftly – extending to a whole range of underwriting, claims handling and fraud detection activity. So how concerned should insurance and risk professionals be about the potential of AI to replicate their roles? Airmic’s chief executive, John Hurrell, reports that the association has yet to conduct research into the topic but keeps abreast of developments. From the insurance industry’s perspective, he sees the primary focus of AI and robotics currently on personal and small commercial lines, with potential to extend to the middle and upper end of the small to medium enterprise (SME) market.
“Airmic generally represents the UK’s largest organisations – most members have a global turnover of over £1 billion – and this sector has completely bespoke insurance requirements,” says Hurrell. “Every client has a different risk/underwriting profile and policy architecture, which lends itself less readily to the AI and robotic trends developing in other areas.
“That said, our members are benefiting from much improved access to data, including benchmarking data, but this is being used to improve risk management rather than to commoditise the insurance products available to them. The market is motivated to preserve the bespoke nature of underwriting in this sector firstly to respond to customer needs and secondly to avoid commoditisation of pricing.”
Limits to adoption
Among the rapidly-growing fintech companies working with the banking sector in developing robo advisory services is Italy’s Objectway, whose chief product officer, Georgios Lekkas, reports that AI is automating a growing number of tasks and thereby enabling smaller teams to handle a steadily increasing number of customers and cases.
In the middle and back office, simple types of automation such as straight-through-processing (STP) have been steadily improving efficiency for decades. The result
for professionals employed in these fields has generally been beneficial, allowing them to focus on more value-added tasks rather than taking away their job.
“Today’s robo-advisors do not, in the vast majority of cases, qualify as AI since they are programmed to execute algorithms pre-determined by their programmers,” says Lekkas. “The ‘robo’ word caught on in the press, because the concept of machines replacing humans captured imaginations. But it is a case of simple automation similar to STP, albeit much more visible because it involves the client relationship.
“As an example, under increasing price pressure, AI could be used to deliver services to the retail and mass affluent client segments to keep the business cost-effective. It would be employed at scale to serve such clients with algorithms of increasing sophistication that exhibit adaptive behaviour, replacing the simple robo-advisors of today.”
However, Lekkas doesn’t envisage AI taking over completely. He believes that there is a threshold, above which banks and other financial service firms will still prefer to employ human advisers who can better understand their clients’ needs, and manage their reactions.
“Machine learning algorithms cannot explain their mechanisms and be understood by customers. In fields such as private banking, the cost of failing to meet client expectations is very high, so we expect that, in this kind of businesses, employees will continue to do interaction-rich jobs, performed with massive support of computer systems and algorithms necessary to cope with the ever-increasing amount of data.
“The same goes for decision-making roles such as insurance underwriting or lending. They could be completely delegated to AI – or still be mediated by professionals supported by AI, according to the customer value and type of business.
“While AI could significantly improve financial institutions’ risk management, through more in-depth assessment of risk in portfolios, it brings about uncertainties and distrust related to the possible misuse of this technology, in terms of concerns surrounding the security, privacy and quality of data.”
In addition to ‘chatbots’ for engaging with customers, expect to see the growing use of other systems such as voice digital assistants (VDAs), virtual agents and natural language processing (NLP) to pick up programmable tasks, says James Hall, CEO of RPA and AI solutions provider Genfour. Hall is one of the optimists, suggesting that the new technology won’t make humans redundant, but rather free them up for less mundane tasks than trawling through reams of data.
“Like any new technology or process, automation must deliver results,” he adds. “The beauty of RPA for financial services firms is that a process can be automated in a few weeks, while the results in terms of completed transactions and freed-up human resources are visible immediately. That’s in stark contrast to the typical core system transformation projects that take years to map out, much less execute.” So if 2016 was the year of the AI/RPA pilot, 2017 will be the year of adoption.
This article was published in the January 2017 issue of CIR Magazine.
Download in PDF format
Click here for more interviews and analysis
Contact the editor
Follow us on Twitter