Intact sharpens AI strategy across specialty lines
AI is out of the lab and in production at Intact Insurance - especially in specialty lines where risk is complex and every account is different. The mandate: speed up decisions and improve precision without sidelining underwriters.
"Being able to identify how innovation comes in and makes us better, stronger, and more competitive is something that we've spent a lot of time on," said Lynn A. O'Leary, global COO of Intact's specialty lines group.
A decade of lab investment
Intact built its foundation early. The Intact Lab, now over 600 people across Montreal, Toronto, and Hong Kong, focuses on automation and AI. What started in Canadian personal and commercial has expanded into the UK and specialty segments.
In specialty, the lab's work touches underwriting, pricing sophistication, operations, and claims. The goal isn't flashy tools - it's better decisions and faster service where it matters.
Underwriting: faster intake, firm judgment
"It is still very much a human-driven process, but by leveraging AI we can ingest submissions, assess a risk quickly, and get a quote out the door," O'Leary said. Speed matters - and so does conviction.
"Not surprisingly, in insurance, the quicker you can ingest a submission, assess that risk, and get a quote out, the better the service to our broker partners," she said. "And it's not just about speed; it's about getting a quote that we feel strongly about."
Human oversight isn't optional
Guardrails are clear. "Our process currently benefits from an underwriter's involvement, and we're focused on leveraging human oversight where it adds the most value," O'Leary said. "You need that human interaction, oversight, and engagement."
AI's job is to get high-quality information into the hands of experts. "AI's role was to improve the flow of information into underwriters' hands to make us a more sophisticated carrier and business partner," she said.
Claims: faster ingestion, clearer answers
Claims is another win. "It accelerates our ability to digest information, make decisions, and get back to our policyholders and customers," O'Leary said. The point is simple: faster responses, stronger relationships.
Lab and business at the same table
Intact doesn't build in isolation. "They work hand in hand," O'Leary said of the lab and underwriting teams. "If the business is not at the table to talk about the output they need… we miss the mark."
It's a co-creation model. "The data lab experts sit at the table and say, 'We could do this, this, and this. What really makes sense for you?' Then we can build a model off that," she said. Solutions built with end users earn adoption. Those built in a vacuum don't.
Build the core, buy the boost
Intact prefers to own its core capabilities. "Our reflex is to build it and own it internally," O'Leary said. Control of data and pace matters.
Vendors still have a place. "We are not agnostic to the value of external vendors," she said. Partners act as accelerators - not substitutes. "At the end of the day, we would never heavily rely on external expertise, but we see it as an accelerator for things already underway internally."
Adoption is the real hurdle
The technology is rarely the blocker. "The biggest challenge is the speed of adaptation and adoption," O'Leary said. People need to see, touch, and use the tools.
Bring teams in early. "If you just put something in front of them and ask them to adapt to it, that's very difficult," she said. Early involvement flips resistance into pull. "Once people are using it, they see what it can do and what it can deliver, and it's pretty amazing."
What this means for carriers and brokers
- Keep underwriters in control. Automate intake and enrichment; let humans own pricing intent and terms.
- Co-build with the front line. Sit data scientists with underwriters and claims leaders from day one.
- Measure service, not just model lift. Track submission-to-quote time, broker response time, and quote confidence.
- Use vendors to accelerate, not outsource. Own your data pipelines, governance, and model roadmaps.
- Invest in enablement. Short, hands-on sessions and co-pilot tools beat long decks every time.
How Intact is deploying AI now
- Underwriting: submission ingestion, risk triage, and pricing support with underwriter sign-off.
- Operations: workflow automation to reduce manual handling and rekeying.
- Claims: faster document intake and analysis to speed decisions and communication.
- Pricing: added sophistication where specialty data is sparse and unstructured.
Practical metrics to run on every AI initiative
- Cycle time: submission-to-quote and claim FNOL-to-decision.
- Quality: hit ratio, quote-to-bind, and loss ratio deltas in controlled pilots.
- Capacity: underwriter and adjuster time returned to judgment work.
- Experience: broker and policyholder satisfaction after AI-enabled touchpoints.
- Governance: model explainability, exception rates, and human-in-the-loop outcomes.
Governance and control
Human oversight, auditable decisions, and data control are the non-negotiables in specialty AI. For teams formalizing their approach, the NIST AI Risk Management Framework offers a useful baseline for controls and lifecycle discipline.
NIST AI Risk Management Framework
Where to skill up next
If you're building similar capabilities, focus on practical workflows: intake, enrichment, triage, and decision support. For curated playbooks and training built for carriers and brokers, see AI for Insurance.
As O'Leary put it: "They may not be experts in AI, but they're experts in their business." The winning strategy pairs that expertise with the right data and tools - so the front line can move faster with more certainty.
Your membership also unlocks:
