A False Claim? Generative AI Investment Thesis in ‘Insurtech’
Thank you to Nick Proud and Danica Bunch (InsurTech Gateway), Peter Tilocca (Zurich Insurance), Timothy Chan (Norton Rose Fulbright, InsurTech), Soohan Kim (Ambiata), Michael Mifsud (Glowpay), Eric Tran (Aura Ventures), The Builders Club Fellows and many more for your thought partnership on this thesis.
- InsurTech overview and market dynamics
- Introduction to the Generative AI opportunity
- 2023 Insurtech Generative AI market map
- Five future market predictions and opportunities
The Australian Insurtech market is digitising and ripe for disruption
The general insurance market in Australia is now worth a substantial $80.9 billion (2022). Despite its size, in the last year, growth has plateaued at 1.6% and the industry has faced challenges with the pandemic, geo-political conflicts and unfavourable macroeconomic conditions (Modor Intelligence, 2023).
In a laggard market, Generative AI and digitisation could unlock innovative disruption and accelerate industry growth.
Before we break down our thesis on Insurtech and make predictions on long-term value, we first need to go back through the looking glass and investigate the core drivers and dynamics of this traditionally “legacy” industry.
Insurance is essentially a safety plan for unexpected problems (policy) in return for regular payment (premiums). At its core, it is a promise to pay and a mechanism to spread risk. Over the last five years, insurance has become an industry which is:
- Heavily regulated, capital intensive and restricted by legacy assets. Traditionally, the industry is one that is quite regimented with strict regulatory frameworks that have led to reliance upon legacy systems. This is driven by push and pull, both to protect consumers from unfair practices through transparency and to ensure the financial stability of insurers. In Australia, APRA, ASIC, AFCA and ICA oversee regulation of this space.
- Increasingly commoditized and competitive, with consumers valuing pricing as the key decision factor over ancillary features. Insurance isn’t perceived as a “loved” product by consumers, and from the insurers’ side, there is a push to reduce costs and maximise efficiency. The emphasis on pricing has increased recently, with the cost of premiums going up significantly especially for home and motor insurance. Hence, managing capital efficiently is of increasing importance to insurers, especially in this interest rate environment.
- Facing digital shift driven by customer demand. Customers used to prefer direct interactions with financial advisors, nowadays digital tools and no-code platforms are now table stakes for insurers. We have seen a shift from direct Insurance, to integrated, to embedded insurance e.g., Smile vs Tesla. An influx of marketplace players and those that offer distribution-focused competitive advantage has since arisen; in the UK ~70% of personal finance is done through aggregators. Australia’s market is still predominantly direct, with a growing opportunity for players to intermediate much like in the UK (particularly for retail insurance, commercial is more regulated).
- Becoming fragmented and incumbents are facing disruption. A focus on product and business model-based innovation like microinsurance, parametric and peer-to-peer products are emerging driven by demand for more customer-centric, data-driven insurance. As a result of this, we have seen the insurance value chain fragment due to synergies between new players (as seen in Celent’s report on disaggregation). Trends such as embedded insurance and verticalization play into this (see Accel’s predictions here).
The opportunity in Generative AI
We have only seen a fraction of what’s possible in this space. There is a lot of opportunity throughout all value chain components of insurance for AI to deliver value. Of note, the Generative AI in Insurance market size is expected to be worth around USD 5543.1 million by 2032 (Mantra Labs, 2023).
This growth will be driven by both push and pull from insurance providers and customers for operational efficiency, better customer engagement and new innovation opportunities (Oliver Wyman, 2023). A prime example of this is in life insurance where the market, “could grow to more than three times its current size. This growth will be driven by addressing under-served and underinsured segments of the market” (Deloitte, 2023).
With these analyses in mind, we centre our thesis research on identifying top areas for investment potential. To do this, we are first deep-diving into the components of the insurance value chain.
The insurance value chain is complex and highly interdependent
Our thesis is underpinned by the dynamics of the complete insurance value chain. The market map below spotlights some focus areas as well as key players in each field. Note that the map is not exhaustive.
The landscape is broken down into B2C providers, marketplaces & brokers; B2B enablement technology; and underlying infrastructure. These three sectors interact in different ways and with different dynamics.
Generative AI Insurtech Market Landscape, Sept 2023
Five future predictions for the landscape:
There are five key future predictions that excite us and on which we form our future investment thesis. Overall, we have found that AI adoption is quite nascent in Australia, and as Nick from Insurtech Gateway attests “is very much in its infancy”.
It is an attractive area to deploy Venture Capital investments where we can see billion-dollar outcomes in Australia alone. On the Global scale, we are already seeing an influx of new players in the market solving anything from point source solutions such as underwriting copilots to AI calling for Insurance claims.
1. High barriers to entry mean “disruptive innovation” in the industry with Generative AI will not happen immediately. Due to the heavily regulated nature of the industry and the relatively risk-averse nature of providers, the concept of “Generative AI as a disruptive innovation to drive value creation” may not be true for some time. Entering the market as an AI native moonshot Insurtech provider, specialising in end-to-end digital journey is a big opportunity but not easy to serve due to these high barriers to entry. Barriers to entry protect consumers from unscrupulous providers and instill confidence in the industry but can also hinder innovation and increase cost. As Tim Chan from Norton Rose Fulbright says, ‘regulation is one of the insurance industry’s biggest problems but perhaps also its greatest asset’. This thesis is compounded by the price-sensitive nature of the customer base
Notably, an exciting area where we may see AI native startups innovating is in core systems. Core systems are platforming, composable CDPs and API enablers. It’s an area where we are seeing movement to cloud, and where there are lower barriers to entry than customer facing services. Our thesis is that large insurers will take on the mentality of wanting complete ownership and access to all their first party data, and modern third-party tooling to help them achieve this, and mid-market will have a higher inclination to outsource tooling due to capacity and efficiency. According to Nick, AI still remains early here, but it is a big opportunity as we are seeing players such as Guidewire and Socotra (no code) begin to enable AI into their service offerings.
This being said, there is room for “disruptive innovation” as we see existing players become AI enabled through using tooling in downstream players in the market landscape.
2. Incumbent startups who are hyper focused on solving single point problems before scaling will be able to win over inhouse solutions and drive innovation. There is an opportunity for startups that can solve a single pain point 10x better than inhouse solutions to deeply embed themselves into the workflows of providers. Our thesis is that by first using a single-point solution as a sales wedge, these players can then scale horizontally across the value chain through cross selling.
Nick from InsurTech Australia attests that they “are excited about AI enabling truly autonomous agents in the insurance value chain. This will free up humans to do more complex tasks and have technology fulfil the repeatable tasks required.”
Although existing larger players may have historical data to train off, and established distribution channels, AI native startups can leverage their leaner technology stacks and shorter feedback loops to train models faster; and hence have an advantage here. Key to this will be developing a “sticky” solution that is deeply embedded into a tech stack. As insurance value chain components are deeply interrelated, there will be great PLG network effects and good reception to cross selling due to interdependencies. Overcoming the hurdle rate required to jumpstart AI startups through “data moats” (see “The Gen AI flywheel of defensibility”), integration capability with a focus on driving action and superior UX are also of focus (see AI Agents thesis here). We have seen point source startups such as Curious Thing AI, who have been successful in selling to T1 insurers, be a testament to this. Notably, there are some larger enterprise companies such as Sedgwick who recently launched Sidekick+ (GPT tool for claims), Cowbell, Kakau and Lemonade with AI advisors beginning to experiment in the area. Despite this, we believe the 10x that an AI native startup could provide is an attractive proposition.
Deep diving further, startups tackling “efficiency” pieces such as summarisation and Q&A are being adopted earlier than more customer facing sectors. In particular, we are excited about the opportunity in broker back-office support as copilots for MGAs / brokers, claims, underwriting and pricing. For risk calculation in particular, this is compounded by IOT devices. We have already seen first movers such as SixFold (underwriting copilot), ChAI and Akur8 (pricing), Zelros (personalised marketing), Functional Finance (operations and back of office) and AI Insurance (core systems).
3. The Harvey AI equivalent (custom domain specific law LLM) for insurance is yet to happen, and when it does it will be a market disruptor. Given a large amount of unstructured data, complexity, and context-heavy nature of the insurance industry, a fine-tuned / custom InsurTech LLM would provide great value uplift to up and downstream players. In a parallel to Harvey AI, according to Nick, the compliance and regulations insurers operate in are the equivalent training parameters. There are grey areas and also areas which are black and white which can be enhanced with AI. This is particularly true in retail insurance, where protections are strong and are added regularly. Soohan (Ambiata) notes that Insurance differs from law as the common corpus is more fragmented (e.g., PDS bespoke to a particular insurance firm), however there is value in industry specific toollings such as compliance, security and document ingestion.
Being able to ingest from multiple systems, having distribution advantage, ability to capture “rule books” (e.g., policy documents, knowledge articles on procedures for claims, underwriting and supplier and customer information) and lastly, UX in the form of targeted inferences / tailored SDKs will be key to winning. Moreover, building with the concept of AI ethics responsibility a security at its core will be essential. Early movers include Roots Automation and Simplifai AI with InsureGPT and InsuranceGPT: “by training models on a deep corpus of insurance specific data, the speed, accuracy and validity of extraction and inference is greatly improved”.
4. There will be a demand for solutions tackling early prevention and mitigation of claims, especially for data as a service players who are focusing on specific product datasets. Service players who aggregate, clean, and derive “intent” out of data to be used to enhance decision-making will be valuable. We believe those that remain domain focused, are platform agnostic and with strong PLG notions will have a right to win. Notably, internationally there is a push from Governments to work with corporations to improve quality of life so there are fewer claims. We are already seeing examples of this collaboration, e.g. Home Honey Insurance discount for sensor instalment, MLC’s World First Partnership with AI guided mental health support. Two exciting areas with growing tailwinds are catastrophe insurance and health insurance which was previously extremely fragmented and traditionally resource heavy e.g., weather prediction, bushfires, COVID-19. We are seeing players such as Sahha focused on mental health in the DaaS layer who will drive this too. Other notable players are ReAsk (weather), Arturo (property), Geolocarta (aerial imagery).
5. Building with Responsible AI and Transparency at the core of solutions will be table stakes for all players — Nick attests that “with ethics in mind, startups should ensure they aren’t hiding their tech in a black box and be open about their models and capabilities. Otherwise, insurers will struggle to engage with them. There are fears of anti-selection bias being exaggerated by AI. This may be true in some instances, but AI can also help by identifying cohorts that are currently deemed uninsurable to actually not be the case via data and machine learning principles.” There are new players in this field, such as Confident AI and Armilla AI who are solving these in a scalable manner, measuring model bias, latency and performance. Players who are first to market and able to have clear product differentiation, particularly with focus on data lineage and transparency will win here. Another area of growing interest in this sector will be regulation of the rise of digital assets for insurers such as marketing and brand compliance. Startups such as Haast look to solve this through AI.
There is a massive opportunity for the Insurance industry to be disrupted by AI and we are at the inflection point.
So, where do you think the big opportunity in Insurtech is and who the lag and lead adopters will be? We would be keen to hear your thoughts or about any companies disrupting the field in the comments.
About Insurtech Gateway: bring the best founders from across the globe with a unique/deep domain expertise, and we provide the insurance knowledge and expertise to break down the barriers to entry into the industry. We create new risk premium pools for previously uninsurable, or new and emerging, risks in society.
- Generative AI in insurance seen to accelerate within 12–18 months: expert | Insurance Business Australia (insurancebusinessmag.com)
- AI no ‘silver bullet’ for insurance — Chubb CEO | Insurance Business Australia (insurancebusinessmag.com)
- Accel — Prepared Mind: The insurtech landscape and five predictions for the future
- Generative AI In Insurance Market Size, Growth Report By 2032 (precedenceresearch.com)
- Why Insurance Leaders Need to Leverage Gen AI | BCG
- Revolutionizing Risk: The Influence of Generative AI on the Insurance Industry (marketsandmarkets.com)
- The risks and limitations of AI in insurance — IBM Blog
- ChatGPT and Generative AI in Insurance: How to Prepare (businessinsider.com)
- Majority of UK insurance executives investing in AI and machine learning (insurancetimes.co.uk)
- Digital-disruption-in-insurance.ashx (mckinsey.com)
- Synpulse | The Rise of Insurtech: Disrupting the Australian Insurance…
- 3 applications of generative AI in insurance — CB Insights Research
- 6 VCs explain why embedded insurance isn’t the only hot opportunity in insurtech | TechCrunch
- Investing in the future: Venture capital funding in Insurtech — The Digital Insurer (the-digital-insurer.com)
- Insurtech Startup Functional Finance Raises $8 Million to Further Expand Operational Efficiency for the Insurance Industry (prnewswire.com)
- The Rise of the Composable CDP | Andreessen Horowitz (a16z.com)