AI technology is best at work when it’s improving existing processes rather than trying to totally transform the entire business model, according to a panel at the Global InsurTech Summit 2019.
Enthusiasm towards AI and machine learning technology is rising at a feverish rate, as is the same with the InsurTech space. The amount of funding in the InsurTech sector is testament to this. Last year, $3.1bn was invested into the space. The panel discussed the role AI and machine learning will play in the future of insurance.
The panel consisted Rcapital director Emma Bell, Shift Technology co-founder and chief science officer Eric Sibony, Zurich Insurance group head of artificial intelligence Gero Gunkel, Prudential global head of artificial intelligence Michael Natusch and Ushur co-founder and CEO Simha Sadasive.
Chairing the discussion was Emma Bell, who pitched the question, “Are you looking at AI from an incremental change perspective or a total business model transformation?”
First up to the microphone was Zurich Insurance’s Gero Gunkel stating that current focus, in his firm at least, was being put more on improving existing processes. Traditional incumbents are littered with old systems and manual processes all ripe for AI technology to come in and improve. It would be a too tough a task to replace everything in one go, let alone for it to work smoothly first-time. He said, “I just think it’s better to focus, at least for now, on the new opportunities within the existing business models is a good opportunity to learn and to build up skills. Whereas, with a lot of the new business models, I am not sure for there’s always a first move advantage just because you’re the first one doing it and very often also it means you’re the first one trying a lot of things that don’t work out. In the end when you’ve worked it out others can copy that.”
Gurnkel’s opinions were met with a general agreement on the panel. The clear reason for the consensus was linked to the idea entire model disruption is just too ambitious at the moment. AI technology is only in its infancy and has yet to even begin to see its true strengths. Time is needed for the solution to evolve and it is far easier to focus the technology on a specific use case, perfect that, and move onto another area.
Shift Technology’s Eric Sibony said, “I think the disruption will come with time. I don’t know if we can talk about disruption yet but in ten years it won’t be the same as today. But it cannot be from one day to the next. In our case, we’re focusing on claims processing. We started with fraud detection and now we’re focusing on other areas to automate claims. Something that needs brainpower is the problem, you need to define the problems to know where you want to go, and this is usually done step by step. And so, over time with the given years, when we look back then we will realise a big disruption has taken place, but it won’t be from one day to the next.”
Fears on whether AI will mark the end of humans in the workplace never seem to disappear. In truth, it is not unwarranted, AI can revolutionise how processes are conducted, either automating data gathering, generating insights or just handling multiple processes in near real-time. Humans simply cannot match up to this skill set. However, not all human processes can be removed – AI simply serves as a way to improve the lives of these employees.
Sibony said, “The black box aspect of AI doesn’t work in insurance fraud detection because fraud handlers need to know why a case is suspicious in order to do their actions and so we’ve built our solution around AI but specifically to solve this problem.” The fraud detection within the claims process is integrated within the process and humans will just select suspicious claims for further investigation. This action would need additional information from the real world or third-party players to decide whether the claim can be accepted or not. He stated that this process cannot be replaced entirely by AI and instead, needs to be layered over the top. it’s not enough to alert a suspicious claim, you need to know why it is suspicious.
Shift Technology is an AI and SaaS technology developer which helps insurance companies better detect fraud. Its services analyse and score claims, giving a real-time alert to a fraud handler if there is a suspicious claim made. This helps fraud detection teams speed up the number of suspicious claims they can analyse and pass judgement upon. The InsurTech recently closed a $60m Series C round led by Bessemer Venture Partners and is being used to improve product development and expansion in the US and Japan.
AI’s greatest accolade is that it opens the playing field to everyone, he added. A startup can then use this to their advantage by tailoring the AI solution towards a specific use case.
However, Michael Natusch of Prudential believes there are pros and cons on both sides of the market for leveraging the power of AI. A technology startup can more simply integrate an innovative AI solution because it is not held back by legacy systems, something of great envy to Natusch. This also works against them, as they do not necessarily have the data to back it up. A notch in the belt for financial institutions is the sheer amount of data they have, both from historical and current customers. An AI solution can only be as good as the amount of data it has access to, and the more it can use, the better the quality insights or services it can offer.
Data is really the key to AI and without it, the technology is rather redundant. Ushur offers automated customer engagement solutions, powered by AI to simplify backend processes. Insurance firms can use this service to automate their claims processing across endpoints, including claim filing, claim estimates, claim status, short-term disability and FMLA. Access to data for this is essential, not just existing databases but also information accumulating from customer interactions, such as keyword search, data retrieval, topic interest and commonly asked questions.
The company co-founder Simha Sadasive said, “The data science and the accuracy levels of these models are predicted by how much data you actually feed to these models. So, we basically rely on customer data to improve the model but that doesn’t mean you’re starting on ground zero. We build basic models from various different industry dictionaries that are available and that’s a starting point that we work off of it. And we leverage for the specific use case, we leverage the insurance company’s data to improve the models and improve the accuracies of our algorithms.”
Although startups and incumbents can leverage technology to different strengths, Natusch still agreed that AI should be used as an improvement to specific services rather than disrupting everything – whatever the size of the company.
He said, “I think incremental improvements are better for our bonuses individually. But, I think for our businesses, we need to think, experiment and work very hard on building the kind of change that enables us to compete with an environment that will be dominated by what we see emerge out of China for instance. And I think that’s a very different competitive field than what we’re used to in Europe at the moment.”
The panel began with Rcapital’s Emma Bell referring to an article on the FT which stated around 40 per cent of Europe’s AI startups do not use AI. As a final piece of advice to the audience Natusch suggested anyone looking to enter the market, don’t be one of that 40 per cent.
Copyright © 2019 FinTech Global