How AI Is Affecting Financial Services & Insurance

For David, AI is levelling the playing field between large FS institutions, who have all the capital and customers, and smaller start-ups, who have grand ambitions and are unencumbered by legacy systems or thinking. In other words, no matter the type of company AI is creating a financial revolution by unbundling the bank.

You need only look at the statistics to see that this is the case. David showed that:

• 65% of senior FS management expect positive changes from the use of AI in FS (Forbes)
• 83% of businesses say AI is a strategic priority for their business today (BCG & MIT)
• AI is forecast to drive cost savings of $1.1 trillion in FS industry by 2030 (Bain & Co.)
• AI will contribute $1.2 trillion in value to US financial services companies by 2035 (Accenture)
• There are already 2.5 million US financial services workers whose jobs are already being directly impacted by AI (Autonomous)

It’s clear that AI is going to transform finance, as it will many other sectors. Knowing that, the time is now to look ahead to other issues around AI, not just how it’s being implemented but what people think of it. David claimed that AI in FS and insurance is fast becoming an ethical and trust issue.

For instance, claim optimisation algorithms are able to predict the amount of pay out a customer will be happy with and could therefore offer a lower pay out than average to certain customers. Banks and financial institutions are taking action before legislation has been passed to regulate AI use, further exacerbating this crisis of trust.

Nevertheless, it isn’t all doom and gloom. Autonomous found that AI is and will be utilised across myriad use cases, including: reporting and process automation; conversational AI; personalised CX; KYC/AML; image recognition and labelling; and data processing, among others. In total, these use cases are projected to save FS&I over a trillion dollars by 2030.

It was at this point that Mark took over from David to provide his keen insights, garnered from his time at TIBCO. Mark’s first piece of advice: avoid “puppy dog AI!”

With 11% of senior financial services management trusting their AI models, there is an ugly problem with AI right now.

“I like to think about it sometimes when I talk to customers as puppy dog AI, [which] is still happening in a lot of companies where they’re doing a lot of AI all over the place, like having a puppy in the house, where they’re getting into the paper, chewing up your slippers.”

For Mark, the hype about Auto ML – the ability to automatically generate AI models based on high-level inputs – is misplaced enthusiasm because it assumes that simply by making AI more accessible to any user, you’ll get more effective AI. This can encourage organisations to take on more random use cases of AI that are disconnected from the business.

To prove his point, Mark shared a tree map of analytics by number of users from a real TIBCO client. From the left you can see organisation-wide efforts, but as we move to the right more projects spring up but the number of individuals accessing them decreases dramatically. The trick in avoiding “puppy dog AI” is to keep innovation connected with the rest of the business, like a team sport.

Companies that view AI as a team sport form a collaborative bond between data engineers, data scientists, analysts, developers and business professionals/leaders. In particular, organisations build bridges between data scientists and business users. AI should be created in partnership with business users, not as a separate lab experiment. Otherwise, the puppy will run wild and cause trouble.

Traditional data science assumes the world is stable because they train models only on historical data. But this approach to data science assumes that conditions essentially stay the same, that the same patterns and anomalies from the past will happen again in the future. Mark presented new AI tools that flip that traditional view on its head, like streaming data science.

Streaming data science is different. It’s the difference between looking only in the rear-view mirror and looking through the windshield – you get a bigger picture. Namely, by applying machine learning for applications where the world isn’t stable or predictable, you benefit from a more dynamic view of the future. It takes into account not only what happened last year or last month, but what is happening right now, in real time.

Mark showed us a great example of an insurance provider that has utilised streaming data science to better predict insurance claims following a hurricane in Florida. As the storm travelled up the Florida coast, the AI models are scored against sliding windows of information to predict the number of claims and where those claims will be more severe. He showed how those predictions change in real time as the data changes. This allows the insurance company to more effectively respond as conditions change and engage their clients better: they can dispatch adjusters, for example, to the areas they know will have the heaviest claims.

Debt-Shaming App Launches in China

WeChat is one of the largest apps in the world and is often dubbed a ‘super app’ because of its sheer range of functions, including messaging, social media and mobile payments. It’s also powered by AI, with its latest feature being an in-app addition of ‘Deadbeat’.

This app lets you see if people around you are poor and in debt. It intends to create a social honesty network, but it is effectively a debt-shaming app. 18 million low-ranking individuals have already been banned from flying and a further 5.5 million have been banned from high-speed rail. High-ranking individuals, meanwhile, get discounts and beneficial treatment.

Data Science Models Learn & Get Smarter

Streaming data science takes input from any type of data source and continuously identifies ‘anomalous’ factors that have changed. By alerting analysts, risk managers and operations, these conditions can be evaluated in real-time as conditions change.

24/7 AI & IoT Powering the Home Insurance of the Future

Insurance is seeing great shifts currently. Instead of pricing based on likely behaviour, insurance companies are moving to price policies on actual behaviour. It is a proactive approach to insurance that utilises smart sensors and emergency assistance to revolutionise home insurance.

The idea is that if insurance providers can provide tech that makes gas leaks, water damage and home intrusions less likely, then they’ll be able to pass along those savings to customers with lower premiums.

Continuous, Real-Time Risk Management

By applying streaming data science, models can be continuously re-assessed in real-time as inputs change. Live conditions and predictions can be simultaneously compared to historical conditions. Business action can be taken based on this continuous live view of what’s happening now. Actions that might result include the dispatch of insurance adjusters to the right place at the right time, or even simple customer connection and awareness based on predictive signals.

AI Tackling More Diverse Problems, Like Perfume Creation

Perfumery is an ancient art, but Olfactory is disrupting the space with AI. IBM and Symrise are working on next gen perfume and fragrances using a combination of historic data and concoction types aimed at specific demographics.

Perfumers are working alongside AI apprentices to create thousands of formulas and novel combinations. This effort brings together human expertise with machine intelligence to create novel, fine fragrances. Advances in AI, meanwhile, are helping us to introduce unique new fragrances.

AI-Driven Predictive Pricing Signals for Gaming and Betting

Little known fact: 80% of the world’s sports betting activity happens after the game starts. As such, odds change in real time throughout the game, taking into account an unimaginable amount of possibilities – a gargantuan challenge for bookies. But TXOdds has another solution.

TXOdds is a specialist in online gaming data for predictive. The company applies streaming data from live sporting events throughout the world in real-time to predictive models. Bookies and bettors use the feed this produces to make real-time decisions on odds.

AI for Real-Time Image Analytics

Any FS&I firm can innovate with AI-aware streaming video and images. For instance, one solution monitors traffic information on-the-fly, providing real-time updates on traffic flow, accidents, pedestrian footfall and more. FS&I firms could utilise this information to identify trends that could help inform financial decisions and insurance policies.