Insurance’s AI Moment
In my last post, I explored how AI is transforming the nature of work and opening up new opportunities to build.
Insurance is one of the industries I highlighted where full automation is rapidly replacing roles previously augmented by AI.
Insurance workflows that began as simple rule-based systems have quickly evolved into sophisticated AI models capable of solving complex challenges across the insurance value chain.
Today, the mainstream adoption of AI is pushing customer & user expectations towards more intuitive systems that automate manual processes. LLMs are a key driver of efficiency, accuracy, and better customer experiences in insurance workflows as they streamline underwriting, personalize policies, detect fraud across claims, and optimize coverage pricing.
As technology expectations shift from legacy software to modern day automation, the broader market is growing in parallel. 2024 was the first year where the premiums written for P&C insurance firms exceeded $1 trillion while software spend came in at just under $30B.
The world of insurance is changing and the time to build is now.
Where We Are and How We Got Here
The original generation of vertical software companies serving P&C insurers emerged at the internet’s inception. Duck Creek Technologies and Guidewire were both founded in the early 2000s and emerged at a time when there wasn’t even a cloud standard for insurance software. These companies were able to become a core software system of record for every part of the value chain from policy management to billing (i.e., payment, invoicing, disbursement) to claims management and beyond.
These core systems gained a significant advantage by becoming early systems of record. As they captured and centralized core policy data the value of their platforms compounded over time. This advantage grew as they expanded into additional lines of business and added complementary modules such as claims analytics, distribution management, and customer service. By owning the policy data and layering on high-value functionality, they became deeply embedded and indispensable to their customers.
Areas We’re Exploring Today
From my last post, I wrote:
For decades, the insurance claims process has relied on a high volume of human coordination: file intake, damage assessment, policy verification, customer communication, and payout. It’s a deeply operational role, but one that has long been constrained by inconsistent documentation, subjective judgment, and bureaucratic lag.
Given how (relatively) easy it is to embed AI into products, insurance claims has emerged as the holy trinity of automation readiness:
Structured data flows (policy documents, claim forms, repair estimates)
High repetition at scale (millions of similar claims every year)
Clear outcomes (approve, deny, pay X)
As AI takes the bulk of claims processing, human-labor will be increasingly confined to edge cases: complex liability issues, litigation exposure, or emotionally sensitive claims (e.g., life insurance).
Today, we’re meeting builders that are taking aim at established incumbents, looking to reinvent existing workflows with AI, and even rethink the role of brokerages altogether.
3 Areas in Insurance We See as Ripe for AI Reinvention
Below, are 3 areas where AI will drive improved outcomes across the insurance space:
P&C Customer Communications
P&C insurance is a highly regulated industry, with strict standards governing customer communications and the exchange of documents. Adjusters often spend up to four hours a day drafting claims notices which could otherwise be spent actively managing claims or acquiring new clients. We’ve spoken to many teams that rely on communication templates that are meant to save their time, but the reality is building them is labor-intensive. Creating a single template can take 40 hours and it’s common for large insurers to maintain 200+ templates. This creates a significant drag on the speed and efficiency of communications with both new and existing policyholders.
Still, the challenge goes beyond drafting messages. There’s also the issue of communication compliance. Missing a required recipient in a notification can trigger costly penalties, making precision in communication a critical operational concern. As a result, there’s growing demand for modern software solutions, and a wave of startups is rising to meet it. ClaimSorted, for instance, is an AI-native Third Party Administrator (TPA) that insurers can outsource claims operations to, offering automated fraud detection, regulatory compliance, and claims adjudication often delivering payouts in minutes. Similarly, Kyber helps brokerages with document-heavy workflows to automate regulatory communications for personalized and compliant communications.
AI-First Brokerages
The brokerage model that is built upon phone calls, PDFs, and human-led quote gathering is being reinvented by AI-native entrants that automate everything from customer intake to quote comparison and policy binding.
Historically, brokerages have played a high-touch role in the insurance process, but AI is enabling a new class of players to offer faster, cheaper, and more personalized insurance procurement at scale.
Now, modern digital insurance brokerages like Marble are bypassing traditional brokerages and leveraging AI to analyze coverage gaps and recommend personalized policy adjustments across carriers and taking hefty commissions. Others are exploring underserved & speciality areas of insurance like professional liability insurance.
Today’s AI-first brokerages are rethinking distribution, risk assessment, and customer relationships from the ground up and shifting how trust is built in a digital-first insurance world.
Catastrophe Insurance
Catastrophic events like wildfires and floods are growing more frequent and severe, pushing insurers to improve how they model risk, process claims, and manage reinsurance. Traditional CAT insurance underwriting is rooted in historical data and manual assessments can’t keep up with today’s climate volatility.
AI is changing that. By leveraging real-time data from satellites, drones, and IoT devices, insurers can now assess risk, detect damage, and price coverage with far greater speed and accuracy.
Kettle, for example, uses deep learning and 100+ data sources to underwrite wildfire risk in California with real-time precision. Federato helps carriers optimize CAT exposure by simulating portfolio risk and guiding dynamic underwriting. These tools reduce uncertainty and build resilience in one of insurance’s most volatile sectors.
What’s Next?
The next frontier of insurance is still up for grabs with a number of open questions yet to be answered:
Business Model: For one, is the right business model a pure play vertical software that sells to brokerages or one that leverages AI to become a broker itself?
Moats: Data moats also become increasingly important as AI application companies have to fight for incremental revenue and the defensibility needed to retain that revenue over time.
Recording SME Knowledge: How do we digitize and store data that previously only lived in the minds of different employees?
Fr-enemy GTM: Is the path to success about partnering with established players like Guidewire and Duck Creek, or is there a real opportunity to disrupt and replace them?
New Carriers: Will we see new carriers emerge to serve riskier populations and entities given macro tailwinds (climate change, lower credit scores, etc.)?
If you’re building in insurance and have ideas for how to reshape the future of work, I’d love to chat.
👋 I’m a principal at Work-Bench, a $160M enterprise seed venture capital fund based in NYC. I lead pre-seed & seed rounds across AI/ML, Developer Tools, and Enterprise Applications. I also run cofounders.nyc, a cofounder matching community. If you’re building in enterprise, I’d love to chat.