Derek Szeto believes the next phase of insurance innovation will be driven less by flashy consumer-facing AI and more by the invisible infrastructure reshaping how insurers, brokers, fintechs, and digital platforms work behind the scenes.
As co-founder and CEO of Walnut Insurance, Szeto is helping companies embed insurance into their customer journeys under their own brands, with programs that can launch in 30 days or less. It is a model that has gained momentum as fintechs and other scaled digital platforms have matured their data, customer pipelines, and ability to present relevant offers at the right moment.
Szeto recently joined the Canada Fintech Symposium panel InsurTech & AI: Reinventing Risk, Claims & Customer Experience, where industry leaders explored how artificial intelligence is changing everything from underwriting and distribution to claims automation and customer trust.
In a discussion with Fintech.ca, Szeto discussed where AI is already making a practical impact in insurance, why embedded insurance adoption is accelerating, how API-driven distribution can reduce friction for customers, and why data security, transparency, and consumer value must remain central as the sector evolves.
Your session at the Canada Fintech Symposium focused on how AI is reinventing risk, claims, and customer experience. From your perspective, where is AI already having the biggest practical impact in insurance today?
DS: It’s where you don’t see it as a consumer, it’s what the insurers, brokers, insurtechs, and other partners in the ecosystem are using AI to help with coding and workflows.
Coding is mostly self explanatory, you can just get a lot more done more efficiently with the maturity of the platforms and harnesses.
Then in the back office, there’s a TON of paperwork still flowing through the insurance industry and it’s why if you look at the SaaS and AI startups and their websites they are often targeting the insurance industry as one of the key segments.
Walnut Insurance enables partners to launch embedded insurance under their own brand in 30 days or less. What has changed in the market to make that kind of speed and flexibility possible?
DS: The platforms have gotten naturally more mature. It’s been maybe 5+ years for most of the scaled fintechs for instance and they’ve gotten to the point where they are really good at managing their data and have the right pipelines in place to put relevant offers in front of the right consumers. It’s natural to start with a core value proposition, do it well and then look for other places to add value. Inevitably, it’ll get to embedded insurance.
We’ve been around for over 5 years now as well, so we’ve also gotten more mature. We have more programs that are ready to go, ‘off-the-shelf’ including travel, tenant, credit card insurances to name a few. And strong insurer relationships where we can get new programs up and running quickly.
This is where the AI coding agents come in to put it all together. We have public API documentation, and are starting to develop ‘kits’ so it’s easy for developers at our distribution partners to apply their AI to the code and can get from start to finish in what could just be a few hours even.
Embedded insurance has been discussed for years, but adoption now seems to be accelerating. What makes an embedded insurance offering successful for both the partner and the end customer?
DS: Ultimately, it still comes down to consumer value. For insurance, it’s some combination of the right coverage, convenience, and price. If we put tenant insurance in front of a customer for instance that meets the lease requirements during a digital leasing flow, they can pick the amount of contents coverage that matches their own stuff they want to cover, and our partners are able to help the landlord get the proof of insurance they need to hand over the keys and get them in the door sooner – that’s a win for everyone.
How are API-driven distribution models changing the way consumers discover, purchase, and interact with insurance products?
DS: We sometimes say that nobody wakes up and is looking forward to buying insurance. And the spray and pray approach of TV, billboards, and sending everyone the same digital message actually creates lots of noise. APIs and data give you an opportunity to put the right product in front of the right customer at a time that it’s relevant.
If you’re at the international airport, you could get a message reminding you that you could need travel insurance and just get it right there – that can’t happen without APIs.
AI-driven underwriting and real-time risk assessment promise faster and more personalized coverage. How do insurers balance personalization with transparency, fairness, and trust?
DS: This is still early but in some ways AI is just a tool like the tools that have come before it. You certainly cannot let the AI run wild and build whatever model it wants – you will need evals, guardrails, test data to ensure it is fair. The insurer and underwriters ultimately still own the results.
On the flip side, ChatGPT, Claude, Gemini, etc. actually is an amazing consumer tool to help end users understand what they previously would not spend the time to for example read their full policy. Consumers actually get an opportunity to not just trust the policy but use AI to help verify to have a much higher level of baseline insurance understanding in a few chat prompts that wasn’t possible before.
Claims automation is often framed around speed and efficiency, but customer experience is just as important. What does a better claims experience look like in an AI-enabled insurance model?
DS: There will be more parametric insurance and AI will be part of the enabler of that. Parametric insurance is programmatic, trigger based insurance – which also means it can be instant for the customer. That’s most common for example with weather for crops and flight delay where data is public. Before AI, parametric insurance still existed but it really needs definitive data.
With AI claims automation, if you fast forward you could have AI make these decisions on the claim without human intervention against ‘fuzzy’ data. You can manage risk by only having it handle approvals as one way to implement. And if you could approve more claims instantly customers will certainly be more happy with their claims experience. Insurers will also benefit from lower cost to process claims so it’s another scenario where the application of AI can be a win-win.
As insurance becomes more data-driven, what regulatory or trust considerations should insurtech companies and their partners be thinking about early?
DS: Security of data is the first stop for trust. SOC2 Type 2 compliance is something we get asked about all the time. That is table stakes for insurtechs that want to leverage data for insurance.
From a regulatory perspective, insurance is about pooled risk and helping customers in time of need. That’s the lens you want to apply to the use of technology and AI… how does it help the customer get suitable access to products? Get the value they paid for when it comes to time of claim?
Looking ahead, how do you see embedded insurance evolving over the next few years, particularly as fintechs, retailers, employers, and digital platforms look to add insurance into their customer journeys?
DS: It’s going to accelerate. If you think about where we are with AI coding today vs. 6 months ago. Imagine that applied to all the other elements of the business – operations, legal, finance, customer support, etc. etc. It’s all going to accelerate. We’re all going to have AI agents helping us do our jobs.
As consumers, we’re going to have AI assistants. Try OpenClaw, HermesAgent – Google this week announced Gemini Spark. It’s coming and faster than we expect.
Be the first to comment