- What is the potential of robots and automation in financial services?
- Will this potential be harnessed and how long will it take?
- What might the impact of robotics be on employment in financial services?
The financial services industry is a prime candidate for robotics since so much work is rule-based and transaction volumes often justify investment in automation. Here’s a summary of the three main robotics options that organisations could investigate:
So-called ‘robo-advisors’ use algorithms to automate investment management. This makes widely available the kind of personalised financial management that has until now only been available for high net worth individuals. These tools automate every aspect of the financial management process from capturing a customer’s requirements (financial goals, risk appetite and investment preferences) to portfolio allocation, automated share trading and reporting.
The ‘robo’ aspect of robo-advice has been somewhat over-played because the algorithms that lie behind robo-advisors are developed by humans, without the machine learning that hedge funds and others often employ to enhance quantitative and algorithmic trading systems.
2) Robotics Process Automation (RPA)
Robotics Process Automation (RPA) software takes the place of a person and automates clerical tasks – for example, streamlining account reconciliation by transposing data from one system to another. One of the principal attractions of RPA is the speed with which it can be deployed since it is just layered over existing systems, with processes and organisation left relatively unaltered. At the next level, unstructured data and machine learning come into play through Enhanced RPA which has been adopted in areas such as complaints handling and simple insurance claims. IT help-desks use the same technology in the shape of virtual customer agents that ‘chat’ with customers, providing an automated response where possible and escalating to a human for more complex exceptions.
3) Intelligent Automation, Cognitive Automation or Artificial Intelligence
There is a lack of standard terminology at the next stage of robotics which reflects the fact that these technologies are for the most part still at the R&D stage. Here the distinguishing factors are the ability for a system to harness huge volumes of data to generate its own rules, to evolve, and to predict outcomes. Risk assessments, whether for fraud, lending or insurance, are highly suited to these approaches. In banking the frontier of these technologies lies in investment management where start-ups and a few large banks are joining together several technologies to provide an end-to-end solution that represents artificial intelligence: speech recognition captures a customer’s query, feeds this into a robo-adviser and chat-bots give a ‘spoken’ answer, while behind the scenes solutions are progressively improved through machine learning.
This points to the future direction where to exploit the true potential of robotics and AI, banks will need to look beyond simply automating their current ways of doing things. MIT’s Erik Brynjolfsson suggests:
Will the potential be harnessed?
The second of our three considerations is whether or not the potential will be harnessed and how long it will take to get there. For example, Sir Humphry Davy invented the carbon arc lamp in 1802 but it took a further 80 years for the first large scale central power stations to light up Holburn Viaduct in London and Pearl Street Station in New York in 1882. Likewise, some 15 years ago I was COO of an AI company but it has taken massive increases in computer processing and data volumes to push AI forward to the next stage of development.
In our own business process outsourcing operations (encompassing HR, F&A, card processing, loan management and cheque processing) we have been applying robotics extensively for several years now, with consistently high increases in efficiency, 30% plus being typical. When we look at financial services institutions, put simply we don’t see robotics and other automation technologies being applied to anything like the same extent. Why is this?
In our experience of applying robotics, scale is critical. Even where one finds efficiency savings of 30% – 50% the business case sometimes does not stand up. Many of the tasks that are ripe for robotics exist in small pockets of 5 – 10 people. As a result, the cost of implementing the change can outweigh the benefits. Not all organisations have the required scale.
A second phenomenon that we see in financial services is that investment in automation seems to get squeezed out. The twin areas of investment focus are regulation and the digital customer experience. Both are must-dos: one to ensure compliance; the other to keep up with the competition. This makes the back-office and consequently robotics the poor relations. The irony is that automation supports both compliance and a better digital customer experience by reducing cycle times, enabling real-time customer services, removing scope for manual error and capturing audit trails. If looked at in a more rounded context more business cases for robotics would stand up.
When it comes to robotics adoption, the biggest barrier, however, is skills. Educational institutions have not yet redirected training from ‘traditional IT areas’, such as web-development, to robotics so there is a profound skills shortage. But it’s more than that; because robotics is a new technology the business and IT managers that might sponsor a robotics project rarely understand what robotics and, even more so, artificial intelligence can and cannot do. The opportunity to apply robotics is often overlooked; or conversely, solutions over-reach with too much expected too soon from over-ambitious AI solutions where RPA or Enhanced RPA would have been more suited as a starting point.
The impact of robotics on employment in financial services
The third consideration is whether Mark Carney’s predictions will come true in terms of employment in financial services. The problem with predictions of mass unemployment owing to automation is that they are simply not borne out by past experience.
An interesting case is the deployment of ATMs in the USA. The first ATM was installed in 1971. By 1984, 42% of US families had ATM cards. The vice chairman of Wells Fargo duly predicted that branches would have “few, if any, support staff members”. What actually happened is that the number of tellers in the USA has shown a slight increase. The average number of employees used to operate a branch went down from 20 to 13, so it was less expensive to operate a branch which in turn led to an increase of 43% in the number of urban branches. Moreover, once transactions were diverted to ATMs, staff were free to focus on relationship-based selling so tellers were upskilled and became central to bank growth strategies. Meanwhile, the digitisation of payments and reduction of cash handling in branch has seen an upsurge in advisory-only branches. The irony goes further in that the latest feature in many ATMs is a video link to a live teller.
This pattern of displacement and reskilling has been repeated across industries. For example, desktop publishing and word-processing may have reduced the number of typesetters but resulted in four times as many graphical designers being employed as there had been typesetters. Stepping back, this point is obvious since so many more people are employed now than before the machine age that came in the wake of the industrial revolution.
Indeed, a study by the McKinsey Global Institute found a strong correlation between productivity (driven by technology) and employment in the USA between 1929 and 2009. There can be a period of readjustment, with a lag in the time taken for the economy to pick up the slack (and unfortunately it’s not always the same people who are retrained and redeployed; at an individual level there can be losers), but even if this 80-year period is analysed as a series of individual ten-year periods there is a full correlation between productivity and employment.
What futurists, such as Brynjolfsson, might say in response is that this time it is different. Potentially, digitisation and platform economics have brought a fundamental disruption in the rules that have applied until now, for the simple reason that a digital platform can scale more easily than a business model based on people. Instagram famously had only 13 employees when it was sold for $1 billion. Closer to home in the financial services sector PayPal, still nowhere close to scale, has more than two and a half times as much revenue per employee as Barclays.
For sure, the industry is changing shape but this is what progress is all about. I suggest that optimism and curiosity should characterise our approach to evolving technologies. Panic has no place on the agenda.