Digital agents equipped with artificial intelligence (AI), known as Agentic AI, are likely to revolutionize the entire insurance industry, according to experts who participated in a workshop on this topic during the Journée de l’assurance de dommages (P&C Insurance Day), held in Montreal at the end of March.
At the beginning of the workshop, entitled L'IA agentique : la puissance opérationnelle des grands, à la portée des PME (Agentic AI: The Operational Power of Large Companies, Within Reach of SMEs), Jean-Sébastien Dessureault, a professor in the Department of Mathematics and Computer Science at the l’Université du Québec à Trois-Rivières (UQTR), gave a brief presentation explaining the difference between generative AI and agentic AI.
“I like to compare generative AI to the emergence of a genie. You rub the bottle, the genie pops out. The genie answers our questions, and when we're done asking questions, the genie goes back into the bottle. That's all. Agentic AI is more powerful than that. Our genie will emerge from the bottle, but we will give it a purpose, we will give it resources, and above all, we will give it time to accomplish its purpose.”
He compared generative AI to a tool that answers a specific question or solves a given problem. Agentic AI, he says, is more powerful than that. “From the moment AI takes actions, has a goal, and is autonomous in its environment, it becomes a much more powerful entity, far surpassing that of a tool. A hammer doesn't take initiative. An AI-based agent will take initiative,” explains Dessureault.
He adds: “AI isn't a fad; it's here to stay.” Just as speed limits are enforced on the road network, Professor Dessureault believes that guidelines will need to be put in place for the use of artificial intelligence.
A step forward
At the start of the discussions, Willie Savard, founder of Tchat N Sign and president of Altura Financial Group, said he believes that the OpenClaw platform is a giant leap forward.
He praised Peter Steinberger, creator of OpenClaw. “This guy is a solopreneur. He developed it all by himself. And now it’s been acquired by Nvidia.”
Samuel Bélanger-Lamoureux is co-founder and co-president of Cemiar, a firm specializing in property and casualty insurance technology. He notes that many companies try to make themselves more attractive to investors by promoting their product with an AI-related label. “But it’s often just a buzzword,” he says.
Property and casualty insurance “is a highly regulated and complex industry, where operational processes are not simple,” he continues. Using OpenClaw to develop agents in this sector “will take some time and will need to be used in a very controlled manner,” according to Bélanger-Lamoureux.
At Cemiar, they have been working for two years on developing the Cassie platform, an agent-based project. The product has not yet been commercialized due to the complexity of the processes to be integrated.
“We've been working on this for two years with our best engineers, people from Microsoft who work in AI, and we still haven't released the product because it's so complex. The idea of someone starting to create an agent on their own to operate in the insurance industry, I think there's a much greater risk—even a catastrophic one—to your data,” he said.
According to Bélanger-Lamoureux, it will still take time before agent-based AI can be used on a large scale by companies without specialized engineers.
For SMEs
Workshop facilitator Philippe Le Roux, Director of Digital Growth for the Insurance Journal Publishing Group, asked: Does agentic AI offer small and medium-sized enterprises (SMEs) the opportunity to use digital tools without having to invest a fortune in technology and labor?
In the context of property and casualty insurance, Cemiar specialists are asked by their clients to find solutions to improve the efficiency of certain operations. “What we saw initially: we were presented with 10 or 15 operational challenges common to everyone,” explains Samuel Bélanger-Lamoureux.
“The best solution for small firms, at the lowest cost, is algorithms, software scripts. We often start with AI and, with a lot of engineering, we manage to turn it into a script,” he adds. In some cases, the solution involves an agent that requires significant neural network capabilities, as it is entrusted with a large number of requests (prompts). “When data is not organized in heterogeneous systems—emails, activities, documents, fonts—where text needs to be interpreted,” agentic AI becomes necessary, he explains.
“Smaller firms generally have smaller budgets and less expensive tools, which in turn mean less connectivity and fewer application programming interface (API) capabilities. Their processes are less standardized,” he says.
Bélanger-Lamoureux suggests this to small firms: “If you want to develop velocity and momentum—it’s essential in technology to believe in it and show your teams what the technology brings—start with quick, efficient, low-cost wins with minimal risk. Over time, we’ll be able to tackle the challenges that require agentic technology. But I wouldn’t start there.”
Compliance
Willie Savard points out that the requirements of Bill 25 in Quebec and its federal equivalent mandate that all SMEs better protect personal information. “We’re all in the same boat. We have to be compliant when working with client data. We have to make sure we’re doing it the right way,” he says. The arrival of OpenClaw offers SME leaders the opportunity to solve a specific problem by developing their own solution, without resorting to large multinational engineering consulting firms, according to Savard.
“There are things to put in place, and standards to establish. And it's true that we don't always need the massive AI machine that's on American servers; sometimes an algorithm or small local programs can handle simple tasks. It's much more efficient and faster, and it protects data better. But we'll quickly move towards multitasking different elements. That's what's usually difficult for a large company to implement. Because when you talk about a small, five-person, solopreneurial firm, we'll compare it to a large company with five thousand employees; their needs for product development are completely different,” he said.
With his team of three programmers, one of whom specializes in artificial intelligence, Willie Savard developed his own agent, Luge. In just a few months, this tool has become a work colleague that performs tasks in full compliance with regulations.
The user can ask it questions in natural language or write a query. The agent "is able to connect to your emails and text messages. It is able to connect an electronic signature and perform actions."
“An SME is capable of implementing a functional system to retrieve information from its own systems, perform a cross-analysis of its specific needs, and add a personal touch to how it responds to its clients,” Savard points out.
The agent was designed with barriers and safeguards, granting it access to company data that is not available on the web and should not be. “Everything we do in communication channels and transactions ensures that the shared information is blocked, analyzed, and processed in compliance with regulations,” Willie Savard explains.
Will not replace advisors
Ultimately, AI will not replace advisors, but it will eliminate the clerical tasks assigned to assistants, he acknowledges. Advisors can then focus more on building relationships with clients and delegate the preparation of relevant documents and follow-up to the digital agent.
In property and casualty insurance, adds Savard, a broker starting their own firm eventually needs to hire an assistant to relieve them of tasks they are less interested in or skilled at. “You can make a mistake when hiring someone; you have to train them,” he says.
By using AI, the user explains the task to the agent, and they carry it out. If it’s done poorly, it’s because it was poorly explained. “You can quickly have a team working for you, a team you’ve specialized to meet your needs,” explains Willie Savard.
Security
While acknowledging that AI increases brokers’ distribution capacity, Samuel Bélanger-Lamoureux reminds us that the regulatory burden and compliance should not be overlooked. “It’s an extremely regulated industry.” He suggests that brokers use digital tools to increase the volume of clients they can serve. In an industry where labor is scarce, making the best use of human resources is a valuable solution, according to Bélanger-Lamoureux.
“In insurance, we talk about technology, but in my opinion, we should be talking about connectivity first and foremost. Cemiar made its name by creating integration technology. We certainly have one of the best connectivity solutions on the market. When we create robust routes, we can use those routes for agents to use,” he continues.
Cybersecurity remains a constant concern. “Everything generated by your agents must be validated, auditable, and traceable. We call that logs—the requests must have been made, and we must know where. Without all of that, an agent isn’t an innovation, they’re a risk,” underlines Samuel Bélanger-Lamoureux.
The workshop, L'IA agentique : la puissance opérationnelle des grands, à la portée des PME, is available in French on the Portail de l’assurance.
