Risk Insider: Mark Watson

Specialty Insurance: The Domain of Humans

By: | December 11, 2017

Mark Watson has served as President and CEO of Argo Group since 2000 and has been with the company since 1998, when it was Argonaut Group. He can be reached at [email protected].

Despite warnings by thinkers no less admired than Stephen Hawking and Elon Musk, humans are scrambling to deploy the power of artificial intelligence in every discipline. After dawdling for decades, the insurance industry itself is now eagerly experimenting with technology. Online, automated offerings for coverage in standard lines appear weekly, with the new proof of AI savvy being the lowest number of questions an insurer must ask a policyholder before underwriting a risk.

But what of non-standard lines? What of specialty? Will the role of the specialty underwriter soon be obsolete? Not in my opinion. Here’s why.

Insurance contracts are either transactional or negotiated. In transactional business, where efficiency and scale are critical, technology can shorten the time and lessen the cost of clearing, rating and assigning business.

In some consumer transactions, the whole process from quoting to binding can be automated. But here’s the catch: the risks that can be managed in this category must be static, with little or no variance from year to year. Such risks are generally small in size of both premium and payout, confined to well-defined silos (life, auto, home, pet), and immune to changes in weather, politics and the economy. That’s what makes them transactional.

Negotiated insurance is a people business.

Negotiated risks, on the other hand, are unique. They tend to be large. They are influenced by a customer’s finances, the state of the economy, the hunger of the competition and a dozen other factors. Changing conditions influence the kinds of exposures needed by that customer in any given year. In such situations, end-to-end automation is seldom useful. It is practically impossible to establish underwriting rules where a system can understand, discuss and then deliver customized responses.

Enthusiasts will offer that until Deep Blue beat Garry Kasparov in 1997, we said exactly the same thing about chess. True.

The wonder of artificial intelligence is that it can learn to match or better the human process of judgment. But data crunching, no matter how nimble, requires data. Bridget van Kralingen, senior vice president of IBM Global Business Services, contends that 80 percent of the world’s data is out of reach of traditional computing systems, and that security concerns may keep it that way. The specialty underwriter’s deep-domain experience, therefore, may remain an advantage for some time.

Negotiated insurance is a people business. Specialty underwriters learn to interrogate a risk in part by assessing the character of a company, the quality of its management and the reputation of the partners with whom its executives work. I question whether software alone will do that adequately.

Will artificial intelligence be able to assess the impact of weather six months out on the profit margins of a fixed-price contract? Or notice unsound assumptions in a cashflow projection? Or read non-verbal cues to determine the likely success of a CEO’s plan for growth in an emerging market? Unlikely.

In my view, specialty insurance will long be a domain where human intelligence will be just as important as artificial intelligence. Will specialty underwriters be replaced by computers? Not any time soon. Chess players may be a different story.

More from Risk & Insurance