By Louis Ross, Vice President, Consulting Services
BOTTOM LINE
Start with the data. An AI plan built on clear data ownership, shared definitions and quality checks is one the business can rely on. Settle who owns the data and who will deliver the work.
How confident are you in the data behind the decision?
Before an AI tool influences a claims, underwriting or customer decision, someone should be able to answer a simple question: how confident are we in the data it relies on? For many insurers that is harder to answer than the AI plan itself, because the data sits across policy, claims, billing and customer systems built at different times.
Interest is high on both sides of the border. NAIC surveys found that 92% of U.S. health insurers, 88% of private passenger auto insurers, 70% of homeowners insurers and 58% of life insurers reported using, planning to use or exploring AI [1]. In Canada, the most recent OSFI and FCAC survey data available, a voluntary 2023 survey of federally regulated financial institutions that includes insurers, found about 50% using AI, up from about 30% in 2019 [2]. These findings illustrate the growing interest in AI, although they measure different stages of adoption and do not establish business outcomes.
HOW EXPECTATIONS DIFFER
United States: the NAIC’s model bulletin on insurers’ use of AI (December 2023) is guidance, not a model law or regulation, and states decide whether to adopt it. It expects a written AI program and reminds insurers that existing insurance laws still apply [3].Canada: the OSFI and FCAC report is a risk report, not guidance or a requirement. It found data privacy, governance and quality among institutions’ top concerns [2].
Quebec: the AMF’s guideline on the use of AI, published in final form on April 7, 2026 and effective May 1, 2027, applies to insurers authorized in Quebec, including those outside OSFI’s federal remit. It expects board-level oversight, an accountable senior manager and a central inventory of AI systems [4].
Why the data has to come first
A model learns from the data it is given. Where the policy system and the claims system define a customer or a loss date differently, a model can absorb that difference. How much it matters depends on the use. None of that means waiting on AI. It means getting the order right:
- Name an owner for each major data set: policy, claims, customer and billing.
- Agree on common definitions, so a term means the same thing in every report and every model.
- Measure data quality before building on it, and fix the worst gaps first.
Who owns the data, and who delivers the work
Before launching an AI pilot, establish who owns the data, whether its definitions are consistent and who is responsible for preparing it. Cleaning, mapping and governing data takes delivery capacity as well as decisions, and the result needs an owner.
QUESTIONS TO TAKE BACK TO YOUR TEAM
- Who owns each of our major data sets, and do we agree on what the key terms mean?
- How is data quality checked before it reaches an AI tool?
- Who will deliver the data work, and who answers for the result?
- Which parts of the data work do we keep in-house, and which could a partner own for an agreed scope?
Next steps
Building a stronger data foundation takes clear ownership and the right technical expertise. Connect with Louis Ross to discuss how Procom can support your data modernization priorities.

About the author
Louis Ross, Vice President, Consulting Services
Louis has spent over 20 years helping organizations build agility through the right people, processes and technology. He leads Procom’s Consulting Services business in Quebec.
SOURCES
[1] NAIC, “Artificial Intelligence” (survey reports issued December 2022, August 2023, December 2023 and May 2025). https://content.naic.org/index.php/cipr-topics/artificial-intelligence
[2] OSFI and FCAC, “OSFI-FCAC Risk Report: AI Uses and Risks at Federally Regulated Financial Institutions,” September 24, 2024. https://www.osfi-bsif.gc.ca/en/about-osfi/reports-publications/osfi-fcac-risk-report-ai-uses-risks-federally-regulated-financial-institutions
[3] NAIC, “Model Bulletin on the Use of Artificial Intelligence Systems by Insurers” (adoption announcement), December 4, 2023. https://content.naic.org/es/node/10624
[4] Blakes, “Quebec’s AMF Issues Guideline on the Use of AI for Financial Institutions” (AMF final guideline published April 7, 2026; effective May 1, 2027). https://www.blakes.com/insights/quebec-s-amf-issues-guideline-on-the-use-of-ai-for-financial-institutions/

