As Canadian insurers prepare to adopt a new generation of agentic artificial intelligence (AI), they are determined to maintain strict human guardrails to protect both their firms and clients.
“How far and how fast we go will depend on the task,” says Jamie Campbell, head of artificial intelligence and data in Canada for Manulife in Toronto. “Autonomy will increase where the risk is low and the results are proven, and people will stay accountable for the outcome.”
Global management consulting firm McKinsey & Company says agentic AI has, over the past year, moved beyond the experimental stage to become a major focus of enterprise investment in technology.
“Agentic AI refers to AI systems that can plan and execute multistep workflows with limited human direction. They reason, interact with enterprise systems, coordinate with other agents, and autonomously adjust their actions as context evolves,” says McKinsey’s Technology Trends Outlook 2026.
The insurance industry today is “being rightfully extremely cautious, and we haven't seen institutions really implement any agentic [AI] yet,” says Cassandra Tontini, director of risk modeling services for PwC Canada in Waterloo, Ontario.
“But they're looking at it and really analyzing all of the risks. One of the reasons they're being so cautious is [because] the current risk management framework of an insurance company was set up assuming that what you're protecting [and setting] controls against, is human."
“So, if it is now an agentic agent, you need to rethink potentially your whole risk management framework,” she elaborates.
AI integral to business
Campbell says AI is used across Manulife, from underwriting and claims to sales, advice, customer service and investment research.
For example, he explains, the recently launched Manulife Advisor AI Assistant provides advisors with 24/7 “plain-language access to underwriting guidance drawn from our underwriting advisor handbooks, so they can prepare for client conversations and set clearer expectations before an application is submitted.”
AI also helps customers get coverage sooner, which is advantageous because “when qualified applicants can be approved in as little as two minutes, and advisors have clearer expectations before they apply, it’s easier for Canadian families to get protected,” says Campbell.
Adopting AI has forced Manulife to rethink its processes by redesigning work around it so routine work takes less time, and processes are managed by exception, allowing people to focus more on judgment and relationships. It has also required a rethinking of skills.
“Leaders are expected to use AI themselves and to challenge their teams to rethink how work gets done. We’re focused on creating value through how work evolves, not on shrinking the workforce. AI is designed to strengthen human judgment, creativity and decision-making, and AI fluency is increasingly part of many roles,” Campbell explains.
AI has assisted Manulife with productivity, says Campbell. Tools like Intelligent Document Processing, a shared AI service that reads documents and extracts key information, take manual data entry out of the process.
“Our underwriters spend more time on complex cases because straightforward ones are handled automatically. And advisors can get underwriting answers at any hour instead of waiting for a call back,” says Campbell.
“We use AI across the insurance value chain, including pricing and risk selection, customer service, broker support, claims operations, fraud detection, software engineering and routine-work automation,” says Yoram Perez, the Montreal-based senior vice president and chief digital and data officer at Intact Insurance Company.
“We’ve been investing in machine learning and AI for a decade. We currently run almost 600 AI and machine learning models that support our business,” he adds.
Pricing and risk selection
Perez says that AI contributes to the firm’s capabilities that support growth, including more sophisticated pricing and risk selection, faster speed-to-quote, and improved customer experiences. These investments, in turn, strengthen the firm’s ability to “earn the confidence of new customers and build lasting relationships with those who already choose us.”
The biggest gains come from redesigning work end-to-end, not just adding tools. That frees our people to focus on judgment, customers and brokers. “AI supports our employees; it does not replace their accountability,” he explains.
AI also impacts internal controls. Intact has put a comprehensive advanced AI policy, along with guardrails in place. This covers approved use cases, human oversight and validation, strong data and security protections, and clear accountability and safeguards around how AI is developed and used.
“Our goal is to move quickly and capture the benefits of AI, but to do so responsibly and in a way that maintains trust,” adds Perez.
“Responsible AI is part of how we [at Manulife] design, build and use AI, not a final check,” says Campbell.
“Our six Responsible AI Principles cover safety, fairness, reliability, sustainability, human involvement and continuous learning. Our Responsible AI approach is built on four pillars: principles and governance, AI risk management, responsible solution development, and AI assurance,” he says.
The benefits of AI also hit the financial bottom line at Manulife. “We achieved $300 million in enterprise AI value by the end of 2025, and we expect to generate more than $1 billion in cumulative value by 2027, from run-rate expense reductions, revenue growth from AI-powered workflows, fraud reduction and growth absorption,” Campbell says.
Scaling-up AI use
Manulife intends to scale up the use of AI. Becoming an AI-powered organization is one of the company’s strategic priorities, but “our approach is clear - AI supports decisions, and people remain accountable for them,” Campbell explains.
For example, “our new Advisor AI Assistant offers guidance only. It doesn’t make underwriting decisions or give a binding indication of insurability, premium or coverage, and it complements our underwriters and wholesalers rather than replacing them. Advisors retain decision-making authority and are expected to use their professional judgment, because AI-generated responses can contain errors,” he elaborates.
Perez says that Intact intends to continue scaling up AI in four priority areas: pricing and risk selection, customer and broker experiences, software engineering, and efficiency. This will provide better risk insight, faster delivery, more responsive service, and greater capacity for employees to focus on complex, higher-value work, he explains.
Intact is not prepared to hand accountability for all decisions to AI.
“AI can improve speed, consistency and insight, but people remain responsible for judgment, empathy and final decisions, especially in complex or sensitive customer situations. Human oversight also helps protect privacy, fairness and trust,” stresses Perez.
Perez says Intact is exploring how advanced AI, including agentic capabilities, can safely carry out defined tasks within its four priority areas: improving pricing and risk selection; making service faster and easier for customers and brokers; accelerating software engineering; and increasing efficiency so its people can focus on complex work where human expertise matters most.
“The level of autonomy will depend on the use case, risk and controls, with clear accountability, monitoring, and the ability for people to intervene, particularly when customers could be materially affected,” he explains.
Aviva Canada believes there are opportunities across the company to apply agentic AI to specific tasks and workflows, including within customer service, underwriting, operations and claims, says a company spokesperson.
Those opportunities include increasing customer engagement through real-time personalization via conversational AI and large language model-driven next best actions; and radically simplifying end-to-end back-office processes via agentic orchestration, enabling colleagues to focus on exception handling and faster resolution.
There are also opportunities around deploying automated adjudication of claims against dynamic rules and clinical criteria with agentic algorithms; and deploying AI-enabled sourcing, spend analytics and contract intelligence to improve procurement effectiveness, says the company spokesperson.
However, it is also Aviva’s belief that success will rely on combining AI capabilities with human judgment. “As a result, our approach to implementing AI tools is focused on supporting human expertise rather than replacing it,” says the spokesperson.
In some areas at Manulife, says Campbell, AI already completes steps on its own, such as automatically approving qualified life insurance applications through Manulife’s proprietary underwriting engine MAUDE (Manulife Automated Underwriting Decision Engine), or reading and extracting information from documents.
The next step will involve agents capable of carrying out more of a multi-step process, such as gathering information, preparing a case, and routing it to the right person, Campbell adds.
“The Advisor AI Assistant is an early example of where this is heading. It starts with underwriting guidance, and over time we plan to expand it with additional knowledge areas and capabilities that support advisors across more aspects of doing business with Manulife,” he elaborates.
Client protection paramount
Much of the client data for an insurance firm contains personal and sensitive information. Furthermore, clients often reach out for insurance when they are in a vulnerable state, “so when we think about what insurance companies need to manage from an AI governance perspective that [must be kept] top of mind,” says Brenda Vethanayagam, AI trust leader for PwC Canada in Toronto.
Clients need to receive the same level of care for protection of their data today, whether that is provided by humans or by AI, as they did prior to the existence of AI, she stresses.
Another key need is to protect against the potential for bias.
The potential for bias existed long before AI, but AI has amplified that risk, and protection for both the client and insurer must be incorporated into the insurer’s risk management framework, says Tontini.
“Before we had processes and procedures in place in the insurance industry to try to limit bias, especially in pricing insurance products,” she adds. “Now if you are using AI to aid with that, you need to make sure you have the correct guardrails in place to prevent that bias.”
The insurance firm has a duty to determine how it will advise its clients about how AI is and can potentially be used with respect to their account. In addition to transparency, the firm needs to assess and provide the client options to decide the degree to which they would prefer to interact with a human or deal with technology such as a chatbot, says Vethanayagam.
If AI is not governed properly, the trust developed with clients over a period of years can be broken. “And so, making sure you're governing ethically is non-negotiable. This is not just an IT project; it is a leadership responsibility,” Vethanayagam stresses.
