Many have proclaimed that 2017 will be ‘the year of AI,’ with a 300% increase in investment in its technologies predicted compared with 2016. Start-ups with AI at the core of their business in particular will be looking to benefit from this surge in investment capital, as they did last year when 550 raised $5 billion in funding. While efforts at embedding human intelligence into machines has a long history, recent years have seen a rapid evolution of the technologies captured under ‘artificial intelligence’, spurred on by decreasing costs in computing power, advances in memory capacity and cloud computing, and the necessity borne of the generation of evermore quantities of structured and unstructured data. Naturally, the ability to quickly and automatically process this data and derive actionable insights is creating a lot of excitement across multiple industries, yet there remain a number of obstacles to AI adoption; many business leaders cite the lack of a defined business case or required skills, or the need to first modernize internal data management platforms.
You want to automate what?!
Advances in AI are also fundamentally challenging assumptions about what is automatable: repetitive, rules-based processes are no longer the sole focus as AI opens up the possibility of automating activities that require knowledge and experience. Predictably, this has led to anxiety around the wholesale replacement of employees by machines, anxiety that can in part be challenged by a shift in focus from the automation of occupations to the automation of activities, where processes are transformed and roles are redefined rather than made obsolete. That machines can complement the capabilities of employees, affirming the value of expertise while enabling them to dedicate more time to higher-value work, will soon become clear, first in sectors such as financial services, where automation will be mostly software-based. For many, then, and in contrast to well-reported events in Japan, the near future will be defined more so by the arrival of co-bots in the workplace rather than the sudden and dramatic disappearance of colleagues.
What’s so special about insurance?
Data rich and characterised by routine processes, the insurance industry is ripe for transformation using artificial intelligence, and its potential cuts across all lines. The ability, for example, to generate a continuous stream of insights through the automated analysis of both structured and unstructured data could provide a clear view of shifting customer needs at a level of granularity not possible in the past. However, in order to arrive at this point, insurers must align themselves with AI’s rise from a cultural, technological, and organisational perspective: carriers will require an integrated and fully digitized environment, the capacity to gather and effectively manage data from a variety of sources, and a culture of innovation guided by senior leaders attuned to the pace of change and transformational potential of AI. In terms of automation, systems that simply ‘do,’ such as RPA, have delivered efficiencies in the back office across the insurance industry, providing a scalable way to carry out administrative tasks, for example, in data entry and notifications. 2017 will see a greater emphasis placed on mapping a route to systems capable of learning, harnessing technologies like machine learning for their capacity to analyse large amounts of data from various sources and feed into both processes that are dynamic and unbound by rules.
Insurance, the co-bots are coming
Machine learning is attracting such attention in the insurance industry, as well as in other sectors, because it dramatically reduces the cost, and increases the quality of, prediction. Considering that tasks can be broken down into four parts (data, prediction, judgment, and action), ML relates to one input of automation, and as the technology matures the predictions of humans will increasingly be displaced. However, employees in the insurance industry will be in a position to complement prediction by exercising their capacity for judgment. This means, for example, that while ML will lead to the automation of standardized underwriting across auto, home, commercial and life insurance, it could also be deployed alongside employees in the identification of emerging risks. Data analysis could be combined with human judgment as part of an iterative process that would see feedback and reinforcement given to an ML component tasked with analysing trends and identifying risks, the significance and growth of which would then be monitored by employees with sufficient domain expertise.
Shifting the focus to the automation of activities rather than occupations throws up some interesting questions: should we really be preparing to bid farewell to underwriters, or should we instead be thinking about how their roles might be redefined as the predictive power of AI increases? Could the role of the underwriter become more sales-oriented given their extensive domain expertise and deep product knowledge? While both opportunities and uncertainty abound, exploiting technologies like ML will require strong leadership guided by a strategic vision oriented towards becoming a truly data-driven business. C-level insurance executives should be encouraged to recognise the increased importance of data insights relative to domain expertise and the potential of artificial intelligence to work alongside, and enhance the performance of, their companies’ employees.