This month, we had the pleasure of hosting an insightful webinar on the AI revolution within the financial services industry, co-hosted with Insurance Times and EY. The three speakers (shown below) explored the true meaning behind AI, differentiated between fads and the real benefits, explored what successful implementation looks like, and dove deep into the regulatory and ethical challenges and considerations.
In this article, we’ll revisit the key themes and learnings that emerged from the fireside chat and resulting audience questions from our esteemed three-person panel:
This is due to two main reasons:
“The explosion into public consciousness has bled into the C-suite,” as David Carmalt explained. “They’re beginning to ask questions around AI and whether it could be right for their business, what is hype and what is reality.”
From a corporate perspective, this is the first time we’ve seen a major move at a C-suite level to engage in these discussions, and at such a wide scale. Senior leaders within financial services are beginning to understand the impact that data science can have on the bottom line – whether driven by improvements in customer engagement, process automation, risk pricing, reserving, fraud detection, or other areas.
In short: the surge in interest is not a passing trend; it signifies a broader awakening to the potential of AI.
AI is already making a significant impact in insurance and reinsurance. Pre-trained models and accessible AI technologies have paved the way for real-world applications.
As [EY] noted, “Now you don’t need to train the models. They’re available on a nice user interface for everyone to use, and it’s a massive step forward.”
Some applications already used within the industry include:
When it comes to implementation, knowing the expected ROI is just the first step. The next step is taking stock of your business and understanding true areas of improvement.
As in any industry, implementing AI presents a set of challenges and even potential risks alongside the abundance of opportunities.
Understanding the tipping point
When discussing these challenges, an audience member asked how to navigate the ‘tipping point’ around the cost of implementation vs. value creation.
Dewald Botha, Director of AI Solutions, explained how all investments in AI must come down to ROI, plain and simple. He expanded, “You have to do that assessment really well, and to know the exact cost going in and the return expected to be received. If that isn’t clear, it’s difficult to justify an AI project. For me, the tipping point is cost vs return.”
Selecting the right use cases for your business
There are thousands of use cases where AI could have an impact on your business, but what may work for one firm may not be right for another.
When it comes to implementation, knowing the ROI you expect to generate is just the first step. The next step is to take stock of your business and understand true areas of improvement, not just copying what your competitors are doing.
It all depends on what you’re underwriting and your competitive landscape. “It depends on the lines you’re writing, how evolved your business is, and whether there are strategies that you may be looking to adapt,” David expanded. “AI could really change your business… but you have to base the decision as to how and where to use AI on tangible needs and expected results.”
When it comes to finding your specific business need, Apoorv summed up succinctly: “My biggest advice is to start with the biggest, the lowest common denominator – what are your biggest pain points? From this, you usually get the biggest return.”
Ensuring data quality
Ensuring data is of high quality and suitable for AI model training is crucial. A “platform mindset” is key, said Dewald when answering a question about the dangers of implementing too quickly.
“You need to keep in mind that whatever is developed should be capable of adaptation and to be used for years to come… As long as the foundations of the platform are correct from the start, you can keep building and developing whilst new use cases arise and as technology evolves.”
As with any project, ensuring alignment among all stakeholders, both internally and externally, is essential for successful AI implementation.
“We find the most hurdles in implementation when there’s a lack of clear alignment. This could be over the use case itself, data which is in the wrong place, or people are arguing over data and technology ownership,” David commented. “Having clear alignment enables companies to execute with good momentum, iterate quickly whenever needed, and continually sense-check that the solution being built works for all stakeholder groups. “
The regulatory environment and transparency
An increasingly important stakeholder to consider is the regulator. As a result, most solutions or products which are built should be done so with one eye on existing regulation or the likely evolution of AI regulation.
Most financial services firms answer to multiple regulators, whether within a single jurisdiction or across multiple ones. This is obviously a challenge in that AI regulation is not homogenous across jurisdictions or indeed in some countries, across sectors. That being said, most major financial services firms have long dealt with the complexities of multiple regulators, and so are typically used to managing these challenges.
One critical aspect of AI technology which all regulators agree on is the need for transparency, and therefore building solutions which allow interrogation by risk teams, boards and regulators will continue to be extremely important.
Bias and fairness
An important note made by Dewald is that “Bias in models and data is almost always present.”
For example, a motor insurance company may have historical data which reveals men had more accidents than women. As a result, men would be charged a higher premium. Being aware of how your data is working is key: “You have to always be aware of it and be able to explain it. Then you can ask yourself – is it ethical to discriminate against someone based on Risk Factor X?”
The implementation of AI in financial services does carry risks, but these risks should not deter organisations from embracing AI. As Apoorv noted, “The risk is not taking advantage of the opportunity.” The true risk lies in not harnessing the power of AI to drive innovation and efficiency.
Organisations need to strike a balance between risk management, compliance, and innovation:
In conclusion, the rapid growth of interest in AI has naturally created some hype around the subject, but the genuine applications within insurance and reinsurance have the potential to be transformative in some or all areas of the insurance value chain. These, of course, have to be considered alongside the challenges in implementation, ethical and regulatory considerations, and the rapidity of the technology evolution which can (if not executed with a long-term “platform mindset”) render certain solutions redundant.
Organisations that navigate these themes effectively can harness the power of AI to heighten competitiveness and drive innovation in the ever-evolving insurance and re-insurance landscape.
Other articles that may interest you