Markel International launches AI Centre of Enablement to scale responsible AI
Markel International has appointed Maureen Tomlinson as Head of AI and launched an AI Centre of Enablement (CoE) to accelerate AI adoption across its five businesses. The CoE sits within International Portfolio Analytics, led by Managing Director Simon Cooper-Williams, and delivers education, advisory support, governance, and engineering capabilities.
Tomlinson will continue as Senior Vice President, Operations for Markel in Canada while leading the new function from Toronto. She will partner with International Portfolio Analytics, IT, and Change Delivery to identify high-value use cases and drive safe, scalable delivery.
Why this move matters for IT and engineering
Markel is treating AI as a business capability, not a set of experiments. As Chief Operations Officer Carys Lawton-Bryce put it, the CoE exists to make adoption safe, scalable, and business-led, while staying ahead of regulatory expectations.
For technical teams, this means clear standards, shared infrastructure, and a defined path from proof-of-concept to production. Less rework, fewer one-off builds, and stronger alignment with underwriting, claims, and portfolio goals.
What the AI Centre of Enablement will likely provide
- Reference architectures: Patterns for retrieval-augmented generation, decision support, and predictive models across underwriting, claims, and finance.
- MLOps + LLMOps foundations: Model registry, feature store/vector store, prompt library, evaluation harnesses, red-teaming, and automated testing.
- Governance-by-design: Risk classification, documentation, approvals, and audit trails tied to policies and regulatory expectations.
- Secure data access: Role-based controls, data minimization, PII handling, and lineage from source systems through to model outputs.
- Delivery playbooks: Use case intake and triage, build-vs-buy guidance, vendor assessment, and production readiness criteria.
- Enablement: Training for engineers, analysts, and product owners; coding standards; prompt patterns; reusable components.
Initial priorities to watch
- Use case pipeline: A lightweight intake that scores impact, feasibility, data readiness, and risk level-so teams ship value fast without skipping controls.
- Data readiness: Clear contracts and quality checks for policy, claims, pricing, and document stores; synthetic data options for safe experimentation.
- Evaluation and monitoring: Objective metrics for accuracy, bias, latency, cost, and drift, plus human-in-the-loop guardrails for higher-risk workflows.
- Security and privacy: Isolation for model workloads, secrets management, content filtering, and prompt injection defenses for LLM apps.
- Compliance alignment: Mapped controls to internal policy and external frameworks such as the NIST AI Risk Management Framework.
Operating model signals
- Central platform, federated delivery: The CoE provides shared tools and guardrails; business-aligned squads ship features with embedded data/ML talent.
- Product ownership: Each AI capability (underwriting assist, claims triage, document intelligence) treated as a product with a roadmap and SLAs.
- KPIs that matter: Quote-to-bind speed, loss ratio impact, leakage reduction, cycle time, model uptime, and unit cost (tokens/transaction).
- Change management: Clear playbooks for user training, feedback loops, and phased rollouts so adoption matches real-world workflows.
Leadership and experience
Tomlinson brings deep operations, technology, and analytics experience. Since 2023, she has led business technology strategy and delivery for Markel Canada across data, pricing, and operational processes supporting policy and claims services.
Before Markel, she spent 11 years at Verisk (formerly Opta Information Intelligence) leading analytics solutions for actuarial and underwriting teams across Canada, including roles as SVP of Sales and Professional Services, VP of Analytics and R&D, and VP of Production Solutions and Operations. She also held senior technology leadership positions at Economical Insurance Group.
What to expect next
Cooper-Williams noted that the CoE strengthens operational foundations for AI at scale. With Tomlinson at the helm in Toronto, collaboration across all five businesses should tighten-bringing consistency to how AI is developed and deployed, and giving teams the tools and knowledge to deliver stronger outcomes.
For IT and development teams, this is the cue to align backlogs to the CoE's standards, consolidate tooling where possible, and turn pilots into production services with measurable business impact.
Further resources
Your membership also unlocks:
