Are you missing the boat on machine learning?

By Anna Game-Lopata –ANZIIF Content Writer | 27 Apr 2017
  • Claims
  • General Insurance
  • Insurance Broking
  • Reinsurance
  • Risk Management
WorkshopPhoto

Over the last few years, emerging technology analyst Nick Teulon (pictured right) has been monitoring the meteoric rise of insurtech and artificial intelligence investment (AI), mainly from venture capitalists in the US and Europe.

Having worked for 90 Seconds, a fast-growing startup, he’d not only witnessed a big disconnect between corporate clients who were talking about innovating and not actually doing it; but also how quickly a small, agile business could shake things up with technology.

In addition, after interviewing 20 ASX 100 C-level executives, Teulon realised they lacked sources for information about what might be on the horizon and had no consistent plans in place for dealing with disruption.

For the love of insurance

With this in mind, Teulon created his own research organisation, FilteredHQ, which tracks emerging technologies and now exclusively provides insurance clients with technology trends research.

‘FilteredHQ only focuses on insurtech because our first customer was an insurer,’ Teulon relates.

‘We fell in love with insurance after studying exactly how it works.’

When he teamed up with machine learning specialist Touchtech to offer his research, Teulon saw an exciting opportunity to build applications that could change the insurance industry.

‘We could also see how the current system is archaic because it’s very human-resources heavy. It’s led by people and a lot of manual processes rather than by technology.

‘Because I have a background in technology I knew instantly [insurance processes] could be automated pretty quickly.’

Insurtech funding

Enter Touchtech

Founded six years ago by Adrian Falvey and Robin Marshall in Wellington, Touchtech discovered a market around app development and grew very quickly to a business of about 30 staff.

After deciding against government contracts, the company explored digital innovation before settling back to software development and expanding to Australia.

CTO Dean Budd, who is based in Touchtech’s Melbourne office, says the business saw the emergence of AI and machine learning and had already identified insurance as one of the markets ripe for disruption.

Getting Teulon on board was a natural fit.

Budd, a software engineer with 20 years’ experience in coaching and leading development projects for large enterprise systems, says AI may sound mystical but it really isn’t (and we’re not about to create human robots).

What is AI versus machine learning?

He says while it’s generally accepted that the terms AI and machine learning are interchangeable, AI is actually a very large field of work of which machine learning is one part.

‘Machine learning is where all the progress is happening,’ Budd says.

Machine learning is the ability to teach machines to ‘think’ or ‘behave’ like humans. Essentially, software is designed to mimic the human brain based on research undertaken on the functioning of neural networks.

Currently the problems that can be solved by machine learning are very niche. The majority of which are limited to the problems humans know can be solved and automating the tasks humans know how to do.

‘For example we don’t know how to teach a machine how to predict the stock market,’ Budd says. ‘That’s a problem we don’t know is even feasible. Anything a bit more complex, like running an organisation, that’s a way off.’

Hungry for data


The reason great progress is being made in automating feasible tasks is that largeamounts of data can be fed in to machines to help them learn how to do them.

‘If a human can perform a task, we have reams of data to tell machines what’s right and wrong,’ Budd says, ‘and if the machine makes a mistake with its predictions, humans can correct it.

‘Machines can now do anything a human can do with one second of thought, which means a whole world of complex tasks that require several small decisions to be made can be automated.’

Driving a car is a perfect example.

Further down the track, data will also be used to predict the next big products by revealing demand.

A simple example happening now is Deliveroo, which started its life as an Uber Eats app competitor which connected restaurants to consumers.

By collecting information about what people want to eat suburb-by-suburb, Deliveroo was able to decipher shortages of certain styles of food in certain areas, then locate their own catering businesses in those areas.

AI funding

What can machine learning do in insurance?

In terms of insurance, brokers who collate a lot of data from the customers in order to make decisions and produce results could use intelligent machines to speed up their process enormously.

Budd says that’s where Touchtech will place its focus from a practical point of view.

While New York-based startup Lemonade is currently the artificial intelligence leader in the area of insurance claims (famously making approvals in three seconds), Teulon says the underlying engine required for the simple contents insurance claims made via the site wouldn’t use much AI.

‘[Lemonade’s AI] functionality would be the face and voice recognition algorithms required to make judgements about whether a person is lying when they make a claim via the app. The claims process itself is pretty straight forward.’

Teulon sees a far more ambitious agenda.

‘If you think about a disaster like the Christchurch earthquake, people are still going through the claims process five years later,’ he says.

‘But in the future we’ll have drones flying over affected sites taking high resolution imagery which will be analysed using image recognition to help the insurer process claims. We’re not there yet, but that’s the way it’s going.’

No more wasting time 

Touchtech has yet to win an insurance commission, but Dean Budd says a successful small project with a talent agency illustrates the possible gains.

The agency, which receives thousands of applications from people looking for talent work such as modelling and acting, had an auto-approval process to conserve the company’s resources.

But as the business added more complex rules into the system, it was becoming less and less accurate.

‘They were aiming for 80 per cent accuracy on auto approvals, but about 60 to 70 per cent of the applications were going through to a person when they should have been rejected,’ Budd says.

‘We built a neural network and trained it using historical data about each element of the application, including which elements should lead to applications being accepted or not.

‘The system is now producing results at 95 per cent accuracy. This has reduced the number of hours someone has to go through the false positives, so the company is thrilled. They’ve gone from less than 30 or 40 per cent accuracy to 95 per cent.’

The 80-20 Rule


Nick Teulon maintains that understanding how the 80-20 rule could work using machine learning for the claims process is the big opportunity for the insurance industry.

‘Eighty per cent [of claims] are probably pretty straight forward, while 20 per cent might be complicated,’ he says.

‘Machine learning is about being able to break that up. You can get a machine to do the 80 per cent that is low level and easy to approve quickly. 

'The 20 per cent that the machine deems more complicated can go through to the junior claims people or an expert. This means the senior people aren’t wasting their time on lower level tasks.

‘That’s where insurance professionals can start now. If they do, in five years’ time their programs will be at the level where the machine does all the claims.’

Cost of missing the boat

If insurers fail to integrate machine learning, Teulon says, they risk losing their competitive edge.

‘If your competitor can improve the claims process by 10 or 20 per cent, they’re reducing churn and improving their overheads.

‘They can pass the savings on to their customers and gain a massive advantage or just enjoy juicy profits for another few years until the competition catches up.

Budd adds that unlike other types of software, if you get a head start with AI then the general consensus is you’ll stay ahead.

‘You’ll start collecting more and more data which will in turn make the machines more and more
intelligent,’ he says.

‘So while competitors might be able to come along and replicate the code or software, that will no longer be where the value is.

‘The value will be in what the software continues to learn, so it will be impossible for competitors to catch up.’

Roundtables on offer

In order to demonstrate the benefits of machine learning that it sees for the insurance industry, Touchtech plans to hold a series of roundtables in Australia and New Zealand.

Teulon says he and the team will talk with insurance executives, initially aiming to ascertain whether they’ve got a deep understanding of the technology and what their fears of the future might be.

‘It’s going to be our job to explain how exciting this can be,’ he says.

‘When it comes to machine learning you’re not impacting customer acquisition.

‘With forty per cent of churn happening at a claims event, machine learning could mean not only reducing churn, but also improving overall customer experience and reducing overheads.

‘It’s a lot easier to save money than make additional revenue. Worst case scenario: nothing changes.’

‘Best case: you could make some serious improvements to the bottom line.’

‘An example is one Japanese life insurer that replaced 34 claims workers with bots. Those headlines should be in Australia within the next 18 months.’

Touchtech is looking at running round tables in May and is currently finalising dates in Australia and New Zealand. Interested? Contact Nick at FilteredHQ.


This article is worth 0.25 CIP Points

  • {{item.ShortMembershipLevel}}

    {{item.OnlineName}} / {{item.FriendlyCommentDateFormat}}

    • avatar{{reply.ShortMembershipLevel}}

      {{reply.OnlineName}} / {{reply.FriendlyCommentDateFormat}}

      {{reply.Comment}}

Report This Comment

Are you sure to report this comment?

Reported successfully.

This popup will be closed in 3 seconds.

Share this Page

Related Articles

  • Robot

    The Robots are Here

    By Anna Game-Lopata-ANZIIF Writer | 27 Mar 2017

    Insurtech leader Dr Catriona Wallace is calling on the industry to experiment with the tech that her company, Flamingo, and others offer to see if it is useful for solving the problems insurers and their customers have.

  • Perry Abbot

    Your Friendly Insurer

    By Anna Game-Lopata, ANZIIF Writer | 10 Mar 2017

    The local incarnation of German peer-to-peer insurtech pioneer Friendsurance is set to launch its first product and CEO Perry Abbott is keen to grow the concept in Australia and beyond.