The robots are coming and, if the headlines are to be believed, jobs as we know them are doomed.

But forgive me if I’m not running for the hills – or reconsidering my choice of industry – just yet.

In fact, quite the opposite.

Unless you live in a news-free bubble, you’d be aware advances in machine learning and Artificial Intelligence (AI) are likely to drastically change the jobs of the future.  Some claim between 30 and 80 per cent of the jobs we do today will become obsolete.

The insurance sector is seen as fertile ground for this ‘invasion’ - given its relative lack of disruption and innovation to date - with the claims space in particular considered ripe for the taking.

The Guardian was one of a host of international outlets that picked up and ran with this story from Japan recently:

“A future in which human workers are replaced by machines is about to become a reality at an insurance firm in Japan, where more than 30 employees are being laid off and replaced with an artificial intelligence system that can calculate payouts to policyholders.”

Insurance newcomers like ‘Lemonade’, meanwhile, are adding to the hype with promises of using technology to pay claims within 6 seconds.

Yes, significant changes as a result of technology are inevitable. But does this mean the roles we know today will disappear completely?

No. And conversely, the opportunity for new and improved roles in the coming years looks more promising than ever.

Here are four reasons I don’t see robots taking the job of claims managers in the area I work in; life insurance.

1. Technological change happens slower than we think

Consider the fact that as much as 80 per cent of all life insurance claims are still submitted as a hard copy form.

Forget about a robot doing anything, we haven’t even digitised the  form! 

This is despite the technology to submit forms electronically having been widely available for more than 20 years.

Most would agree an online form is the superior channel for the customer, the insurer and the environment, yet it is still not common practice.

Why? Well technology is not the constraint; it’s the ability and willingness to change the existing process.

The considerable cost and effort needed for change requires evidence of a significant return on investment - and it’s not really there. 

The life claims sector has an even bigger challenge: Even if we can get a claimant’s information online or over the phone, will we get doctors and other third parties to use an e-channel to submit documents? 

This will remain one of the biggest challenges in the industry, despite the technology required being basic and available.   

2. It’s about gathering more than deciding

A supposed benefit the robots will be able to bring is making an accurate decision instantly.

Lemonade’s six-second claims decision is sold as a game-changing experience. 

But there is a significant flaw.

Making a quick decision is easy.  We can say ‘yes’ to any claim the instant it is submitted if we decide to but, of course, we don’t.

We accept that to make an accurate decision there needs to be some due diligence.

As an example, 95 per cent of group Total and Permanent Disability (TPD) claims are accepted. It’s simple to demonstrate that paying 100 per cent of claims - and dismissing the claims team’s  - is not cost effective, and not good business.

Putting cost and effort into ensuring only valid claims are paid is a value-add exercise.  The challenge, therefore, has always been to gather the necessary facts to make the right decisions. 

So robots must tackle the real issue - getting information from medical parties, employers and other stakeholders.

This is a much bigger challenge than simply making a decision at speed.

3. It’s not worth the risk

There is already work going on to build models that can accurately determine the likelihood of a claim being declined, based on available claims data. 

If that 5 per cent discrepancy mentioned above for TPD claims can be detected by the robots , there is obviously a benefit to be had.

I’ve tried myself to build machine learning models using 10 years of claims history to see how well those 5 per cent of invalid claims could be identified. 

I’ve failed, dismally. I never really had a chance. 

How good does my model need to be before I can say the productivity benefit it will provide (by replacing the case manager with the robot) outweighs the risk of making an incorrect decision?

The answer is: between 99-100% accurate.

Why? Because any incorrect decisions that could cost on average $250k will quickly dwarf the resource cost savings.

Further studies of declined claims show root causes are often so complex that a machine model would struggle to get anywhere near 80 per cent accurate, let alone 99..

In summary, there is just not enough productivity benefit to be had to justify taking the chance.

4. The 80-20 rule means it doesn’t make sense

Of course, robots don’t have to do all the claims. You could have a hybrid.

Perhaps the robots can process the simpler claims and leave the costly humans to focus on only those claims where their advantage makes sense. 

This is already happening.  Triaging low-risk claims is something our robots can assist with.

But how much benefit will this deliver? That is, how many case managers will it eliminate? Probably not many. 

The 80-20 rule fits pretty well in the life insurance claims space. 80 per cent of claims received are, in fact, pretty straight forward - low-hanging fruit that could possibly be automated, albeit with some risk.

Unfortunately for the robots, these cases are also pretty simple for the humans, and that 80 per cent of claims volume only accounts for 20 per cent of the resource effort. 

So even if every one of these cases was automated, only 20 per cent of resources could be saved.  And the likelihood of automating that 80 per cent is low. 

The risk of incorrect decisions of an ‘auto-accept’ process is such that most businesses would only be comfortable with as much as 30-40 per cent being robot approved.

Optimistically, this would eliminate 5-10 per cent of the claims manager resource costs.  Not a great return for a significant investment and increase in risk. 

For that reason, it’s not something that we would expect to see jobs being lost to in the near future.

So am I cynical enough to believe that the machine learning and AI will not have any impact on life insurance claims processing?  Absolutely not. There are huge benefits to be had. 

My point is replacing the case manager should not be the goal.  It’s simply not worth it. 

Improving the performance of the case manager, however, is where the future lies. 

Take a group of 10 case managers doing the same job and measure their output and there will be plenty of variation. 

Not all claim managers are created equal, it seems. 

The potential benefits of having those least-effective managers improve – to the level of the best - are substantial. 

Artificial Intelligence – the robots - might just be the best tool to close that gap.  By identifying what it is that makes some managers better than others.

So I suspect the robots are coming, but we have little to fear and plenty to gain.

ENDS