Rapid advances in neuroscience and artificial intelligence are converging to present enormous opportunities to reimagine how life insurance is designed and delivered to members.
The Investment Magazine Insurance in Super Summit in Sydney on 23 July heard that two issues related to the human brain – diagnosing and treating neurological disorders; and mimicking how human intelligence works in machines – offer the prospect of better treatment and management of illnesses, and an era of more personalised cover, including a greater degree of preventative treatments, and targeted communications to improve engagement.
AI and neuroscience overlap when AI is brought to bear to support neurological diagnoses. Neuroscience Research Australia chief executive officer Professor Matthew Kiernan said it is “totally transforming the field”.
“MRI scans; diagnosis; putting in huge volumes of 100,000, half a million patients; global networks of genetics; this is all AI-derived, and I think we’re going to be the lucky recipients,” he said.
What it means for insurance is “probably much better outcomes”, Kiernan said. He pointed to a UK study where blood samples have been collected from about 55,000 individuals over a period of about two decades.
“In that period of time, a thousand of the 55,000 developed Alzheimer’s disease,” Kiernan said.
“Now they’ve gone back through all of the blood samples and put it all through AI – 65,000 blood samples – and they’ve worked out four…particular proteins, put them together, that diagnoses Alzheimer’s disease. That is currently being developed now as a diagnostic test.”
Testing for Alzheimer’s today requires an amyloid PET scan and a lumbar puncture to detect amyloid in the cerebrospinal fluid. The costs of the diagnostic tests – not including additional costs for neurology reviews – run to between $5000 and $7000.
“If you could send someone from the GP practice to [the pathologist] and they do that blood test, it’s about $100,” Kiernan said.
“So that’s going to transform the diagnosis, accelerate it and work out who should be getting amyloid monoclonal antibody therapies. That’s just one example.”
A role for super and insurance
Kiernan said there is a clear role for the superannuation and insurance industries to support research and the development of new diagnostic tests and treatments, which tend to start out being very expensive before they’re scaled up and the costs come down.
“The cost for the insurance and the hospitals is massive for the community,” Kiernan said.
“You’re the brilliant minds, this is what you’re going to have to come up with, because the bottom line is a lot of these therapies are incredibly expensive,” he said.
“We already pay massive overheads and we contribute to the healthcare of the community. But I suspect that we’re not going to be able to cover all of these costs in the long run. So how do we do that? Well, I think this is where you’re going to be coming in.”
The diagnoses and treatments described by Kiernan are becoming possible in part to rapid advances being made in AI to find meaningful patterns in masses of data and its improving ability to make accurate predictions based on the data.
Ajay Agrawal, Professor of Strategic Management at the University of Toronto’s Rotman School of Management, said that AI is at its core a predictive technology.
Speaking by video link from Canada, Agrawal said prediction is, put simply, using information you do have to generate information you don’t have, and AI is getting exponentially better at using existing data to make accurate predictions. The real trick is restating the questions that need to be answered as prediction problems.
The implications for the early detection, treatment and management of diseases are obvious for the insurance industry, which is undergoing an evolution of its own.
It’s moving (a cynic might suggest) from a business model built on finding reasons to deny claims, to one seeking more and varied ways to help individuals claim and then recover, undergo rehabilitation and, where possible, return to work.
At the same time AI is improving patient outcomes, it is helping insurers understand better the risks they face when underwriting. And that comes down to being able to make better predictions.
“We can boil insurance down to effectively three predictions,” Agrawal said.
“The whole industry, prediction one is the marketing prediction of which customer should I target with which products?
“Prediction two is, what’s the likelihood of a claim – so, predicting the risk. And prediction three is, when a claim comes in, is the claim valid or not valid?”
Reducing uncertainty
Agrawal said a lot of processes and systems, or “scaffolding” in insurance is there “because we have so much uncertainty around these things”.
“Once we get much higher fidelity predictions around those three key points, we can do a fair amount of redesign to how the insurance industry works,” he said.
Even before an insurer’s actuaries get to work interpreting the AI-derived results of the work done by scientists such as Kiernan, its predictive powers can be brought to bear in three areas, Agrawal said:
- Engaging members: Personalise communications; predict the preferred communication channels and timings for each member; ensure members receive relevant information at the right time.
- Financial planning assistance: Predict each individual member’s future financial needs and goals; provide tailored advice and projections to help members understand the importance of decisions they make regarding their superannuation; encourage proactive engagement.
- Customer sentiment analysis: Predict member satisfaction on issues based on their past interactions; allow funds to address concerns promptly; improve services and make members feel valued; make members more likely to engage with their fund.
“For those in the room that are focused on healthcare and insurance in particular, you can think of this, the insurance industry, as largely underwriting risk, and the risk is the likelihood of a claim – so, predicting the likelihood of a claim,” Agrawal said.
He said there are many areas where “high-fidelity prediction can be used for reducing health-related claims”, including predicting individual patient risk; reading images and detecting anomalies (such as cancerous cells); predicting the most effective treatments. On a business level, AI can play a role in verifying the validity of a claim, and in fraud detection.
“Many of these are areas where we already were using sophisticated statistical tools for predictive analytics and AI just gives us a much higher fidelity prediction capability,” he said.
Think beyond underwriting risk to managing it
Agrawal urged the insurers and super funds at the summit to “think about even leaning further into it” and to move past the traditional insurance approach of just understanding and underwriting risk, to proactively managing it.
“Now that we can make these very high-fidelity predictions at much higher frequency intervals, we have much greater opportunity to use that for managing risk,” he said.
[In] life insurance, that rather than just predicting the likelihood of a claim, in many cases, the insurance company will know more about the risk and the potential factors that could reduce risk, than…the person themself.
“My sense is that the future of this industry is using the intelligence and the data that the industry has for predicting these risks not just to underwrite them and price them, but to reduce them.”
Agrawal said the industry must also consider what it really means to engage with members, not get caught up too much in the overwhelm of data collection and analysis, and focus on the outcomes the AI needs to be trained to achieve.
“Rather than thinking too much about the data – other people will sort out the data – is the issue of what does it mean to engage,” he said.
“Why do you want to engage them? And what types of engagement are better than other types of engagement? How do you measure engagement performance?
“The thing to remember about machine intelligence is it is the first tool in human history that learns from use, we’ve never had a tool, that the more you use it, the more it learns.”
This field is fascinating and brings hope to all of us, that cures for diseases can be accelerated.
Where it will stall, is when the data starts telling the real truth about genetic weaknesses amongst ethnicities who are prone to disease simply because of who they are.
The Insurers will want to be able to protect their profitability by making reasonable actuarial calculations based on medical and scientific studies, then charge accordingly, which is exactly what happens now if you have a medical condition that creates a loading or exclusion at an individual level for better quality retail Life and Disability Insurance cover.
Behind the scenes there will be roadblocks thrown up if too much accurate data is accumulated that could be seen as setting up a dangerous precedent by, for goodness sake, telling the truth.
What these brilliant people who are developing AI, to bring to life rapid improvements in the treatment of disease will need to be protected from, are the loud wailing minorities who have done nothing positive to society and whose sole function is to denigrate innovators so they can protect their own backyard of self importance in their own little minds.