Agentic AI is likely to reshape high‑volume retail and small commercial lines in Canada, while large, complex risks will probably remain firmly in human hands, according to Paulo Salomao (pictured), national lead of strategy and consulting at Accenture Canada.
The technology’s impact depends on the nature of the risk and the underlying workflows, he said.
“Agentic is working best in environments where you need repetitive decisions, where you have a lot of data and where you have clear economic signals so that you can prioritize what can be done by humans, what can be done by AI,” he said in an interview.
He argued that insurance meets all three conditions in many parts of the value chain, from personal lines to small commercial.
Underwriting, claims and policy servicing are already structured as “pre-industrialized decision factories,” he said, with work broken into repeatable steps that can be assigned to agents.
In that environment, he said, it is natural for carriers to use agentic systems to take on well‑defined tasks around data collection, triage and simple decision‑making, while reserving the remaining judgment calls for people.
For small commercial, Salomao said the data footprint and repetition of risk types make it a prime candidate for wider use of agents.
“Small commercial has meaningful data and we’re starting to see some application on that, especially when it comes to adjudication or underwriting,” he said.
He contrasted that with large corporate and specialty accounts, where the risks are rare, bespoke and negotiated case by case.
“Large commercial continues to be a very human-centric task because the kind of risk that you are insuring isn’t repeated every day,” he said. “How often do you insure a nuclear reactor?”
Those kinds of risks, he said, do not lend themselves to standardised decision paths or model training at scale, which limits how far agentic systems can go beyond supporting work.
The distinction between small and large commercial mirrors what is already happening in personal lines and group benefits, where straightforward, high‑frequency items are being pushed toward straight‑through processing.
On the claims side, Salomao said, simple, low‑severity losses are increasingly handled with auto-adjudication, often with AI consolidating evidence and applying pre‑set rules.
In underwriting, he said, the first impact has been on the front end of the process rather than on the final decision.
“Underwriters used to spend a lot of time gathering information and then making a decision,” he said. “Today, the information is consolidated mostly by AI. And the underwriters just make the final decision based on the data that is presented to them.”
Salomao said the technology affects “tasks that make up the jobs” rather than wiping out roles entirely. That framing applies across retail, small commercial and parts of mid‑market business.
“The difference is AI isn’t taking over the job beginning to end,” he said. “AI takes over tasks that make up the jobs.”
Regulation remains one of the main brakes on how far carriers can go, particularly in life and other heavily supervised lines. Salomao said life insurance still requires a licensed human decision‑maker, while group benefits can support much higher rates of straight‑through claims processing where the rules allow.
The core technology enabling these shifts is not entirely new, he said. Rules engines, workflow tools and early automation have been in place for years, with initial auto adjudication rates “in the single low digits.” What has changed is the complexity of tasks that can now be delegated to software agents that handle unstructured data, orchestrate multiple steps and interact with other systems.
He said Canadian insurers are starting to scale these approaches in targeted areas, although they remain “behind the US still” in overall adoption.
As agentic AI penetrates deeper into high‑volume products and small commercial lines, Salomao said the practical question for insurers is less about whether the technology will be used and more about how they segment their portfolios and workflows to match it.
He said portfolios with repetitive, data‑rich risks and clear economics are likely to see continued experimentation with straight‑through journeys, while large corporate and specialty placements stay anchored in human judgment with AI tools supporting analysis rather than directing outcomes.
