AI for Insurance Companies in Canada

When insurers call us

• Claims intake backs up after a weather event. First notices of loss arrive by phone, email, broker portal and web form, and each one is keyed in by hand.
• Underwriters re-key broker submissions. Applications, schedules and loss runs come as PDFs and spreadsheets in different layouts.
• The underwriting desk answers the same broker questions. Appetite, required documents and referral limits come up every day.
• Model risk and actuarial ask where generative AI fits. E-23 and the AMF guideline both take effect May 1, 2027, and the inventory needs to include AI tools.
• Staff use AI tools without an approved list. Compliance wants a written policy before policyholder data goes into any of them.

What we automate for insurers

Claims intake. Document processing reads first notices of loss from email and forms, extracts policy number, date and type of loss into the claims system and routes the file by line and complexity. The adjuster decides coverage, reserves and payment.

Underwriting submission intake. AI extracts applicant details, schedules and loss history from broker submissions into the underwriting workbench and lists missing documents. The underwriter decides on risk, terms and price.

Broker and policyholder correspondence. Email automation sorts incoming messages and drafts replies from approved templates and underwriting guidelines. A staff member reviews and sends.

Fraud referral support. AI flags inconsistencies across claim documents and prior claims and prepares a referral summary for the special investigations unit. Investigators decide what to pursue. The typical scenario below shows a similar triage in financial services.

Scoring models with documentation. Where a predictive model fits, such as claim complexity or triage priority, we build it with the inventory entry, risk rating proposal and monitoring plan your model risk function expects.

Service assistant. A customer service assistant answers questions about claims steps and documents from approved content, and hands policy-specific requests to staff.

Canadian rules that shape the work

OSFI Guideline E-23, Model Risk Management. Published September 11, 2025 and in effect from May 1, 2027 for federally regulated insurers, including foreign insurance company branches. OSFI's backgrounder says it covers traditional actuarial models as well as AI and machine learning. E-23 expects an inventory, a risk rating for each model, review independent from development, and monitoring. Every AI use case gets a draft inventory entry.

AMF Guideline for the Use of Artificial Intelligence (Quebec). Final text dated March 2026, in effect from May 1, 2027 for authorized insurers, financial services cooperatives, trust companies and deposit institutions. It asks for a centralized inventory of AI systems whose risk is not negligible and a risk rating for each, updated regularly. It supplements the AMF Model Risk Management Guideline.

FSRA (Ontario). FSRA's IT risk management guidance, in effect since April 1, 2024, applies to Ontario-incorporated insurers and reciprocals. A proposed FSRA framework for auto insurance rating and underwriting names machine learning models and the risk of unfair discrimination; its effective date is still open. We found no final AI-specific FSRA guidance as of September 2026.

FINTRAC. Life insurance companies, brokers and agents are reporting entities with a compliance program and suspicious transaction reporting. AI prepares files; the compliance officer decides.

Quebec Law 25, section 12.1. A decision based exclusively on automated processing of personal information must be disclosed to the person, with the main factors on request and a review by staff available. Claims and underwriting designs keep a person as the decision-maker.

The AI compliance review turns these into a checklist per use case.

Which package fits

Insurers usually start with the AI Opportunity Audit, because claims, underwriting and distribution each have candidates and model risk needs one consistent list. A carrier with one clear bottleneck, such as claims intake, can go straight to a Pilot Sprint.

Full scope of each package is on the pricing page.

How the work runs

The work follows the Remolda Cycle: Audit → Strategy → Implement → Empower → Evolve.

1. Audit. Interviews with claims, underwriting, the compliance officer, model risk or actuarial, and IT security. We review process maps, anonymized sample files and the tools in use.
2. Strategy. Ranked use cases, a 12-month roadmap and a draft inventory entry and risk rating for each, reviewed with model risk.
3. Implement. A six-week build of one workflow in your environment, with the documentation model review needs.
4. Empower. Training for adjusters, underwriters and service staff, plus a staff AI use policy.
5. Evolve. Monitoring against agreed standards and a quarterly review of the next use case.

In our experience the audit usually takes two weeks from kickoff. It depends on interview availability, claims season and your internal security review.