War of the worlds, the advent of AI


Artificial Intelligence (AI) is the start of something huge in the financial services industry and beyond. We are just not sure if it’s for better or worse.




“Across the gulf of space, minds immeasurably superior to ours regarded this Earth with envious eyes, and slowly, and surely, they drew their plans against us.’’ – H.G. Wells, War of the Worlds (1897)


Japanese insurance firm Fukoku Mutual Life recently replaced 34 employees with a guy called Watson. IBM Watson, a computer built by – you guessed it – IBM. Thirty-four may not seem a significant number, but the scalability of Watson should send a shiver down your humanoid spine.

Watson’s only job is to calculate payouts to premium holders, but he’s so damned good at it that his hiring was a no-brainer. Watson will increase productivity by 30 per cent and save Fukoku roughly $1.2 million per year, not bad returns considering the company invested $1.77 million on the guy. It’s easy to imagine why he would cut an unlikeable figure in the office.

It’s not the first time workers of flesh and blood have been replaced by automated steel-and-gear labour. In the automotive industry, robots have long been working the assembly lines and today perform critical functions in almost every sector imaginable, be it in healthcare or textiles. But only lately has the realm between the physical and the sacred been breached. Engineers are building robots to emulate the human brain, which enables them to make decisions and perform tasks that transcend mere brute strength. “Optimists believe computers will surpass the functions of the human brain,” says Jacques Steyn, head of the School of Information Technology at Monash South Africa. On the other hand, “Realists say emulation is an analogy, not a simile.”

If you side with the former, computers are winning and have been for some time. In 1996, supercomputer Deep Blue beat world champion chess player Gary Kasparov at his own game; then Watson, yes the same fellow at Fukoku, whipped the reigning champion of the Jeopardy! game show; and last year, Google DeepMind’s AlphaGo trounced the world go champion at the ancient 2 500-year-old Chinese game.

Currently, a computer can’t outperform humans in an open and dynamic system, explains Steyn, but we are no match for them in closed process systems, such as games like Jeopardy!, chess, go or claims processing, for that matter. By this reasoning, computers can replace some functions of an underwriter, but human underwriters are much more than just a digestive system for data.

More than a numbers guy

“Underwriters who go a step beyond strict rules sometimes take risks based on gut feel and greater contextual understanding,” says Steyn. For the most part, computers can’t do this. The human appetite for risk is based on trust – a foreign concept to a computer.

There is, however, no denying the efficacy of AI. Where it takes a human broker, say, a week to process a claim, a computer can finalise it in seconds and, in some instances, milliseconds. Last year, another overachiever, AI Jim, employed at US insurer Lemonade, broke the unofficial record for the fastest paid insurance claim ever – taking all of three seconds.

The company’s claims bot was named after Jim Hageman, chief claims officer at Lemonade. “Real Jim is by far the more experienced claims officer; his alter ego, on the other hand, is never off the grid and can handle billions of operations in one second. They make a formidable team,” the company says in an article on their website.

The true power of AI will be the ability to simulate a personalised product for the customer, says Neo Mathe, value engineer at SAP Africa. “Insurance will be able to design a personalised marketing campaign that will ultimately lead to quicker sale cycles to the customer.”

It’s not just brokers whose jobs are in the crosshairs. Medical professionals, journalists, accountants … the ripples will radiate across the entire labour pool. As Chris Barry, CRM executive at Mint Group Technologies, says, if the IBM technology at Fukoku proves to be successful the floodgates will open.

So are we entering an era of the broker as persona non grata? The majority of the insurance industry doesn’t seem to think so and continues to tout human capital as just that – capital. A study by consultancy firm McKinsey & Company revealed that very few jobs would be entirely replaced by AI in the next decade, but it will affect almost all jobs to some degree. The research also showed that demonstrated technologies could automate 45 per cent of “activities people are paid to perform”, and that 60 per cent of occupations could see 30 per cent of their operations automated with the technology available today.

Brokers and other professionals serving a commodity-type function should see it as an opportunity to build on their skill set. “If the human role only requires going through certain steps and effectively checking boxes, the likelihood of them facing AI replacement is high,” Barry points out. For instance, AI can efficiently manage the basic tasks of asking the normal policy review questions, taking the answers and translating that into information required for decision making or advice. But AI will never be able to forge a true relationship with a client and inspire comfort.

When steam-powered engines replaced the many jobs of man during the Industrial Revolution, it led to the existence of entirely new fields. Alternatively, as renown philosopher Daniel Dennett and computer scientist Deb Roy explained in a paper, it [the advent of AI] is comparable to the Cambrian explosion where new life forms were created but as many destroyed. “It was an arms race,” Dennett recently told The Guardian.

“The key is to understand where the differences lie and how to empower the current workforce to elevate themselves,” says Barry, “but AI is the logical next step.”

The McKinsey report estimates that US mortgage brokers, for instance, spend as much as 90 per cent of their time processing applications. AI-driven verification processes for documents and credit applications could reduce this number to 60 per cent, which would free up advisers to focus more on consulting clients rather than routine processing. It would favour both the company and client. AI should be seen as superhuman interns that lap up the dirty work. It gives companies the opportunity to strategically redeploy underwriters or brokers into positions where they can put their intelligence and insight to good effect, intimates Karin Kruger, associate director of data and analytics at KPMG South Africa. “Humans in the industry will always have a role,” she adds.

Recent research by KPMG revealed that claims processes for a client was reduced from 35 days to 15 minutes end to end, and also led to a 15 per cent reduction in claims spend due to less interference during the claims process. The technology has a place in the operations of a company but should be tested in a fail-safe environment, explains Kruger.

Businesses need to allow time to perfect the data, the algorithms used and come up with smart ways of utilising this data. It’s what Discovery did with their Vitality programme. “They started on a telematics journey, which is ultimately machine learning,” she says. And it has paid massive dividends for them.

Fukoku’s use of Watson is just one part of the job tasks of an underwriter, Steyn of Monash emphasises. For those values on the edge, AI is not ready yet. These systems can be programmed for self-learning, and context-dependent content requires specific variations on algorithms – but these are constructed by human brains. “Depending on the extent to which an underwriter’s tasks can be formalised in strict rules, a computer might be able to beat a human,” says Steyn.

The AI trend can even be used to train staff. In one study that KPMG was involved in, a bank used AI to emulate a top employee who has a knack for knowing which customers are considering opting out of their contract with the bank and, more importantly, a knack for persuading these disgruntled customers to stay. After interviews with the employee to figure out exactly their actions and words when dealing with these customers, engineers programmed the data into a computer. Today, this AI robot is used to train new staff and check performance management. New human staff are essentially competing with the robot and no one has lost their job, neither man nor machine.

The various sectors in Africa are still trying to wrap their heads around AI and the effects it will have on human labour, not even mentioning the sticky topic of labour laws – specifically in South Africa. But if companies can harness AI and analytics to make smarter, faster decisions and combine it with the power of human capital, favourable times lie ahead. “It puts people in a different position and companies will have to think differently about their workforce. The good ones [employees] will stay,” says Kruger.

Sci-fi flicks rarely follow a plot where humans and machines cohabit in peace as there’s no drama, no flair. Oren Etzioni, CEO of the Allen Institute for Artificial Intelligence, recently told Wired magazine, “One of the reasons I don’t like the discussions about superintelligence is that they’re a distraction from what’s real.” History has far too many examples of promises made on false starts, says Steyn. “Will Watson replace underwriters? It’s too early to tell.”

Thirty-four employees were replaced by a robot and it made headlines around the world, but imagine the plight of the postman with the arrival of email decades ago or of taxi drivers when self-driving vehicles are put to work very soon. These evolutionary changes occur every day right under our noses and it is up to companies and employees alike to keep up or be left behind. The good will stay, as Kruger explains.

Also, in the movies humans always beat the machines or aliens. Then again, we’re the ones writing the script.