Harnessing AI and tech to serve millennials better 

South Africa-based digital insurance provider Pineapple is leveraging technology to give it an edge in delivering motor and all risk insurance products to the younger demographic . Their important disclaimer: Technology is important, but the customer comes first. “We start with the customer and determine what their pain points are before seeing how we can address these using technology,” says Marnus van Heerden, Pineapple’s co-founder and CEO.

The insurtech has ingrained AI technology across its business to improve efficiencies, take customer engagement to the next level and make insurance more understandable. To improve outcomes, Van Heerden advocates for a combination of easy, seamless user interfaces and well-trained staff alongside AI: “It is not just an AI solution or technology or a key person,” he says, “but using a combination of these to solve customers’ problems."

At the application level, AI is deployed to recognise photographs of items that customers wish to insure, and to pre-select appropriate types of cover. At the backend, the digital insurance provider uses self-developed AI models to assist with pricing and risk prediction, while in the customer service space, large language models (LLM) help staff to structure appropriate responses. Pineapple takes a dual approach to AI, buying off the shelf or developing in-house.

The company says there are excellent solutions in the commoditised AI-backed API space – especially in areas such as image and voice recognition – where it makes sense to buy rather than develop. The big differentiator is the volume of data that large technology brands can use to train their models. “But if we need a bespoke solution and we find limited online capability, we are comfortable building our own end-to-end AI models,” van Heerden says. Such solutions are typically indicated in niche areas like fraud detection, risk management and pricing.

Launched in 2018, Pineapple has since built up extensive experience of AI adoption and implementation. “Off-the-shelf solutions that give you an accuracy probability score allow you to manage your risk upfront,” van Heerden says, “they pretty much do what it says on the tin once you have implemented them.” The challenges arise when developing an LLM to assist in quoting or underwriting: “You have to put in place full processes to review that model as if it were a human interacting with customers,” he cautions.

AI has proven excellent in the relatively simple motor and all risk cover categories but does it scale into complexity? Pineapple’s CEO reckons the exponential pace of LLM development hints that complex insurance engagements will be “easily handled”, provided context and sufficient training data are available.

Hardware development stands out as the principal driver of ongoing AI improvement. “They have advanced the speed of AI computation by a factor of a million in just over six years, and the latest chips are around four times faster than those used to train well-known AI-controlled chatbots,” van Heerden says. “If you extrapolate that over time, these models are going to get to a point where they can deal with incredibly complex tasks at scale.”

The case for AI is compelling but also a bit frightening. As LLM models improve, van Heerden envisages a disruptive scenario in which insurance providers will set up a single environment and feed the model everything there is to know about the business. “Once you get to the level of generative AI, you will be able to ask the AI model to implement strategy and eventually give it a marketing budget and an instruction to maximise its underwriting result within a compliance and regulatory framework,” he predicts.

For those who have grown up with rapid technological advancements, particularly in the Generation Z and younger demographic, the integration of AI in business processes is a prospect that is both fascinating and daunting. This generation is well positioned to understand and leverage emerging technologies but must also be aware of the potential risks. They will have to navigate and shape the future of industries heavily influenced by AI, while advocating for balanced, ethical approaches.

Regulation provides a useful counter to unchecked AI scaling. “AI could completely disrupt entire industries globally but in the insurance provider context it will be more difficult for a global disrupter to take shape, due to the sector being regulated on a per country basis,” van Heerden concludes. Getting the funding necessary to capitalise an insurance provider or insurance solution stands out as another potential friction point for startups entering the space.