Redundancy, dear Watson

Mentioning Artificial Intelligence (AI) prompts complex reactions in humans. These reactions are often tinged with fear about redundancy, and the reporting of AI is often in a controversial setting. This is perfectly illustrated by recent events.

On 26th of December last year, Fukoku Mutual Life Insurance Company announced 34 workers were to be replaced by an AI system. Fukoku Mutual is a Japanese insurance company specialising in providing insurance and annuity products in the areas of nursing care and medical insurance. The story certainly made the headlines. Zurich Insurance Group’s COO made an announcement the previous month that it was bringing AI into its injury claims unit. Redundancies were not mentioned and the story did not gain the same traction. It is clear that the redundancy angle is what drove the headlines in the Fukoku Mutual story.

Both announcements concerned the introduction of IBM’s AI platform, “Watson”. The Watson software enables document processing using artificial intelligence in roles that previously were performed by humans, often more slowly and with greater error rates. IBM labels Watson’s technology as Cognitive Computing. Watson is famous for having beaten two of the best (human) contestants in the Jeopardy! general knowledge TV gameshow in 2011.

In technical terms, Watson brings natural language processing, information retrieval, knowledge representation, automated reasoning, and machine learning technologies to the field of open domain question answering. In layman’s terms, what Watson is doing is understanding written or spoken language, reasoning with it and responding to it. To a certain extent, Watson “understands” what it reads and can make links and draw conclusions based upon this understanding. It can do this extremely quickly. IBM claims that Watson can read 200 million pages of text in 3 seconds. Additionally, it will not forget what it has read and it can learn from its own conclusions.

IBM is pushing Watson technology across a number of specific areas, most notably the medical sector with Watson Health. However, IBM is also targeting business analysis and decision support for banking and finance. IBM’s Watson web presence already extends into the banking and finance sectors.

Fukoku Mutual and Zurich are both looking to employ Watson within injury claims departments, presumably leveraging Watson Health’s medical know-how. Zurich has revealed their experience is that the review time of documents of up to 100 pages has dropped from 58 minutes to five seconds.

The saving in man hours is clear. However, whilst Zurich saw an opportunity to reallocate work hours to much more value-added tasks, Fukoku Mutual, chose to cut headcount. Redundancy fell amongst a department of shorter term workers with five-year contracts which the firm “will not seek to renew”.

The headline claims that human workers are being replaced by AI are, as is often the case, somewhat more nuanced than they first appear. In both companies, Watson is being employed to remove process bottlenecks in time-intensive, document-based work. In both cases, the work is time-sensitive as customers are awaiting an insurance pay-out. In both cases, final pay-out decisions will be taken by claims staff acting on preliminary analysis by the new Watson systems. The humans have not been completely removed from the process by any means, but the previous bottlenecks have been cleared.

IBM provides Watson as a series of pay-as-you go services which any systems developer can access. However, the specialised knowledge and skills required by cognitive computing would currently be beyond the competence of generalised systems developers. It is still an expensive resource to develop and deploy.

The power of systems that can analyse data in the way that Watson and other similar offerings do cannot be dismissed. Any area of office work which requires the reading and interpretation of bulk paperwork, whether it be medical scans, company reports, sales data, technical research, appropriateness tests, suitability documentation, legal case papers or injury claims, is likely to fall within the remit of such systems.

GRC Insight has talked about the application of AI in areas such as robo-advice or other areas of professional work. In our engagement with financial advisers and wealth managers, we have seen a range of reactions to the label ranging from cautious interest to scepticism and even outright fear.

In assessing human reactions to AI, much depends on how the technology is presented and what claims are made for it. There is a need for acceptance by both the service provider’s user staff and by the customer too if they are expected to interact with an AI system.

However, looking across other industry sectors, it seems inevitable that such a powerful support resource will soon find its way into larger retail facing work processes very soon. Where such technology provides a competitive advantage, it will be adopted, and adopted at ever increasing speed.

AI applications are increasingly impinging on everyday life. Cognitive processing technologies are increasingly being used by people everyday outside of work. We access them from our phones and in our homes in the guises of Alexa, Siri, Cortana, Google, and similar technologies that select music, tell us information or turn on the lights or heating in response to typed or voice commands. The application of those technologies to our work environments seems strange and artificial until we get used to it. However, within the last 30 years, the same was said of fax machines, visual display units, word processors and desk top computers, not to mention e-mail and the internet.

The headline worry that cognitive technologies will simply replace humans across the office is undoubtedly premature. These systems will be brought in to support and enhance productivity. Humans are simply not that good at reading and remembering every linked concept in a series of 100 page documents. Machines are now capable of doing that. However, machines cannot relate to another human being, understand emotion or display empathy. In many cases, the people are very likely to be redeployed to do these higher value “human” things that machines cannot do – be involved in business relationships with customers and colleagues.