Webinar: Agentic AI in Insurance: what it is, how it works
Agentic AI: Transforming the Insurance Industry
Transcript of a webinar presentation by Simon Torrance July 2025
Introduction
I've been presenting on this topic frequently across knowledge-intensive sectors, where many believe that agentic AI could be more impactful than the internet, mobile, and cloud technologies combined. Today, I'll explain why I think that's the case, share case studies, use cases, and best practices that can be applied by any company in the insurance industry—whether you're a reinsurer, primary insurer, MGA, broker, or vendor.
Background
Two years ago, the CEO of a large insurance conglomerate asked me to run a think tank on the risks and opportunities of AI for the insurance industry. They wanted to understand the next wave of AI and its impact on business models and customers. I became so fascinated by the topic that I created a new company called AI Risk to help companies navigate this frontier.
My key conclusion: The biggest risk is strategic—being outcompeted by companies that adopt agentic AI faster and more effectively than you do.
Current State of AI in Insurance
Recently, Evident Insights conducted the first benchmark of the top 30 publicly listed insurance groups from Europe and the US. They used a methodology already tested in the banking sector to compare where major insurance groups stand.
What's striking is that none of the most advanced companies in their survey are deploying agentic AI at scale yet. Even the top companies have invested significant sums in traditional AI and generative AI but are just beginning to experiment with agentic AI. This represents both a massive opportunity and threat—because when companies do deploy agentic AI, they'll gain a significant competitive boost.
Understanding Agentic AI
Jensen Huang, CEO of Nvidia (which recently became the first $4 trillion company by market capitalization), envisions a future just five years away with two types of workforce: biological (humans like us) and digital (AI assistants). He talks about potentially having 2,000 times as many AI assistants as human workers.
This dramatically changes the game. With cognitive power similar to human workers available at scale, you can do things that would have been too expensive or difficult before. Agentic AI creates a new workforce that complements and supplements your human workforce—something never done before.
The Agentic AI Maturity Model
We created a maturity model to help companies understand this opportunity:
Levels 1-2: Traditional AI, advanced analytics, robotic process automation, AI chatbots (where 99% of companies are today)
Levels 3-4: Agentic AI (the purple zone where almost no one operates at scale)
The key differentiator: agents in the purple zone are autonomous. They have agency and can do work that previously only humans could do. This creates a nearly infinitely scalable workforce made of software that you can turn on or off as needed.
What Is an AI Agent?
An AI agent has five key elements:
Large Language Model (LLM): The "brain" at the center
Persona: Like humans, agents have roles, goals, and behaviors
Planning: Ability to break down tasks and create processes
Memory: Learning on the job and understanding context
Action: Connected to internal and external systems to create outputs
Agents can collaborate with each other and with human workers. Unlike humans, they're auditable and predictable—we can trace everything they do, which is particularly useful in a heavily regulated industry like insurance.
Case Study 1: Multi-Agent Workflow for Risk Assessment
Over a year ago, we worked with a company needing large-scale investment and risk assessment. Traditionally, they'd hire consultants (often from Deloitte) for six-week projects at significant daily rates.
Instead, we created a team of AI agents in three weeks. The results:
Time: 1 hour instead of 6 weeks
Cost: One-time development, then free to reuse
Scale: Analyzed 50 scenarios instead of just one
Quality: Higher quality analysis than human teams
This demonstrates that agents can do complex analytical work—the future is here, just not evenly distributed.
Case Study 2: Insurance Claims Processing
Consider a simple property claim for a burst pipe. Here's how agentic AI transforms the process:
Customer contacts insurer via email, text, or WhatsApp
Coordinating agent activates specialized agents:
Claims handler agent
Service agent for customer response
Underwriter for policy implications
Agents work simultaneously, reaching confidence thresholds
Compliance agent acts as guardrail
Action triggered: payment sent, premium updated
What typically takes hours or days happens in minutes, potentially with no human involvement, while maintaining full audit trails.
Case Study 3: Agentic Teams (Level 4)
Our most advanced case: A non-insurance company wanted to enter insurance but couldn't hire experts. We created an agentic team that:
Manages all customer care for their product line
Handles all claims
Undertakes product and service innovation
Collaborates on Slack 24/7
The agents include specialized roles like actuarial bots. While not as comprehensive as human actuaries, when combined with other agents, they create powerful solutions. One actuary bot, created in one week, helped dramatically reduce claims ratios and increase underwriting profitability.
Results after 18 months:
70% lower FTE hiring
€1 million saved
Higher Net Promoter Scores
Better underwriting profitability
Higher productivity than human teams
The Digital Twin Innovation
When human employee Charlene couldn't keep up with the agents' pace, we created "Charlene 2"—a digital twin based on her email and Slack history. This twin now collaborates with agents at their speed while Charlene focuses on higher-value activities. Employees were happier because the company performed better and their knowledge was used more effectively.
The Strategic Opportunity
Today's insurance industry is constrained by traditional economics. But agentic AI offers near-infinite operational capacity. Like Jensen Huang envisions, you could have thousands of AI agents doing parts of jobs at virtually no cost.
This enables us to:
Rethink claims, underwriting, service, and fraud detection
Finally tackle prevention ecosystems and protection gaps
Transform fundamental business economics
P&L Impact Analysis
We analyzed a mid-size carrier's five-year P&L forecast. Their baseline (gray line) would satisfy shareholders. But with agentic AI, they could:
Grow faster
Respond quicker to customers
Attract competitors' customers
Even increase claims ratios while improving profitability through expense savings
This exercise effectively engages CEOs, CFOs, and boards—preventing agentic AI from being relegated to an IT project.
Key Lessons Learned
This is not an IT upgrade—it's a business model change. It involves HR, IT, operations, marketing, strategy, and compliance.
Avoid "pilot purgatory." Banks started earlier but got stuck doing endless pilots with minimal bottom-line impact.
Start at the top. Get executives in a room, show them the possibilities, help them understand the P&L impact, and create a hypothesis for a new strategy.
Move fast with prototypes, not pilots. Create real commercial prototypes that demonstrate what's possible.
Don't go step-by-step. You can jump to Level 4 while developing Levels 2 and 3 in parallel.
Build foundations. Ensure IT, change management, and HR foundations are in place for scaling.
Think portfolio. Have multiple initiatives running to move forward effectively.
The Strategic Question
I'll leave you with this:
What could you achieve with this capability?
But more importantly: What could your closest rival achieve with this capability?
The strategic risk of being outcompeted by companies that adopt agentic AI faster and more effectively is real. The technology exists today. The question is not if, but when and how fast you'll move.