Across Canadian insurance, two pressures are colliding.
On one side is the demographic squeeze: senior experts edging towards retirement, institutional knowledge at risk of walking out the door, and intense competition for technical talent in hubs like Toronto, Montreal and Vancouver. On the other is a wave of AI and data‑driven tools promising faster decisions, sharper pricing and new ways to serve brokers and clients.
During a recent webinar on the “underwriter of 2026,” two European specialty leaders described that combination as a “perfect storm” – and, if handled well, a major opportunity.
Caroline Bedford (pictured), CEO of UK‑based innovation firm EDII, said AI has been a “wonderful catalyst” for forcing organisations to become more future‑focused. Even where a particular change isn’t directly driven by AI, the technology has sped up conversations about skills, training and operating models that might otherwise have dragged on for years.
Her co‑panellist, Marie Biggas, CUO at SCOR’s specialty business, agreed – but stressed that success will hinge less on specific tools and more on how insurers manage change and build confidence around them.
For Biggas, one point is non‑negotiable: underwriting remains, at its core, an exercise in human judgment. Technology can assist, but it cannot be allowed to dictate.
Drawing on her background in terrorism insurance – a highly “facilitised” and portfolio‑driven market – she argued that underwriters will increasingly act as pilots of AI rather than passengers. Models can surface patterns, flag outliers and suggest pricing moves, but they are “only as good as you make them” and they are not right all the time. The real value lies with underwriters who can read an output and ask: can this be right, what’s driving it, and how does it compare with my experience?
Crucially, those underwriters must keep teaching the models: feeding back where outputs missed the mark, where new exposures are emerging, and where broker or client behaviour is shifting. That feedback loop is where the technology actually improves.
For Canadian carriers experimenting with AI for submission triage, pricing support or portfolio steering, this distinction matters. Tools cannot be sold internally as black boxes that replace thinking. They need to be framed, and governed, as copilots for experienced judgment.
AI and better data are also changing how underwriters see their role.
As models mature, they will handle more of the grunt work: triaging submissions, pre‑populating fields, running standard pricing scenarios and highlighting exceptions. Underwriters, in turn, will spend more time at portfolio level – understanding how segments behave over time, how different coverages interact, and where appetites need to shift.
That shift is already visible in Canadian commercial auto, property‑cat and cyber, where the conversation has moved from “Do we like this individual risk?” to “What does this do to our overall shape and volatility?”
Done well, Biggas suggested, this should make underwriting more consistent and predictable for clients. The familiar story of a risk manager explaining to their board why premiums have suddenly doubled after years of stability should become less common. Better insight and discipline, supported by AI, ought to reduce wild swings and make capacity decisions more defensible.
The longer‑term question she raised is what will differentiate carriers once “good enough” tools and data foundations are widely available. Her answer points straight back to brokers and clients: relevance, transparency and the quality of partnership will matter more than the tools themselves.
If AI is the catalyst, change management is the hard work.
Biggas was frank about the ingredients she sees as essential. First, there has to be genuine buy‑in from the top. Leaders must view AI and data investment as “short‑term pain for long‑term gain” and be willing to absorb the disruption that comes with changing workflows, rebuilding data foundations and re‑training teams.
Second, organisations need the right champions. It is not enough to find a few enthusiastic juniors. The people driving AI‑enabled change must sit at levels where they have influence and a voice at the table. One town hall or training session will not do it; leaders have to keep repeating the message, aligning incentives and removing blockers.
In practical terms for Canadian firms, that can mean explicitly expecting senior underwriters and managers to engage with data and tech teams, not delegate everything to them, and naming a small group of respected leaders across underwriting, claims and distribution as visible sponsors of AI initiatives.
Early‑career underwriters and analysts still have a critical role, but Biggas sees them more as co‑pilots than drivers. Their openness, digital comfort and willingness to experiment are invaluable – provided the organisation gives them cover.
One of the most quotable lines from the session will resonate with many Canadian leaders who feel their organisations are behind on AI.
Biggas argued that the real value from AI will come from compounding learning, not perfect planning. Instead of agonising over the ideal roadmap while others move ahead, she urged companies to “start small, but start now”: pick a narrow, well‑defined problem, build a cross‑functional team around it, and learn as you go.
Bedford echoed that from a skills perspective. Organisations don’t need to wait for a fully fledged academy before upskilling. Simple habits – encouraging underwriters to ask one extra question about every AI‑generated recommendation, or running short sessions where underwriters and data scientists walk through case studies together – can quickly build comfort and trust.
For Canadian insurers and MGAs, the message is that AI is coming into underwriting whether the local market feels “ready” or not. The differentiator won’t be who buys the most impressive tools, but who builds the culture, skills and leadership commitment to use them wisely.
AI may not replace underwriters. But underwriters – and organisations – that learn to work confidently with AI are very likely to replace those who don’t.
