On January 29, 2026, CAFII hosted a webinar titled AI is here. Now What? An Operator’s Playbook for Unlearning and Relearning Insurance. CAFII’s Executive Director, Keith Martin, opened the session by thanking attendees and introducing CAFII’s Senior Research and Policy Analyst, Robyn Jennings, who in turn introduced the featured speaker, Tatenda Manjengwa, Executive Vice President of AI Strategy and Product Management within financial services.
Mr. Manjengwa is a recognized AI leader who transforms emerging capabilities into production-grade solutions in highly regulated environments. In his previous role as Managing Director, Head of BMO Wealth Management’s AI Innovation Lab, he drove enterprise-wide AI adoption across 5,000+ employees, launching Rovr AI, Canada’s first of its kind external-facing generative AI assistant for insurance underwriting, while establishing governance frameworks across BMO Private Wealth, Global Asset Management, Insurance, and InvestorLine. With 20+ years spanning capital markets and wealth management at BMO, Bloomberg, Citibank, and Deutsche Bank, Mr. Manjengwa operates at the intersection of AI product strategy, quantitative finance, and commercial innovation. He has led cross-functional teams of data scientists and engineers, deploying NLP- and AI-enabled workflows on secure, scalable multi-cloud platforms that improved client experience and reduced operating costs.
Mr. Manjengwa holds an MBA from Chicago Booth and an MSc in Business Analytics & AI from NYU Stern. A lifelong learner, he is currently pursuing his third master’s degree: an MS in Computer Science at Georgia Tech.
Following the introduction, R. Jennings extended a special welcome to several VIP guest attendees, including CAFII’s 14 member companies, 14 Associates, allied industry associations such as the Canadian Life and Health Insurance Association (CLHIA) and the Travel and Health Insurance Association of Canada (THIA), and representatives from various insurance and financial services regulators and policy-making authorities, including:
- Alberta Insurance Council
- The Government of Alberta
- Alberta Treasury Board and Finance
- The Autorité des marchés financiers (AMF)
- The British Columbia Financial Services Authority (BCFSA)
- The BC Ministry of Finance
- The Insurance Council of BC
- The Financial Consumer Agency of Canada (FCAC)
- The Financial Services Regulatory Authority of Ontario (FSRA)
- The Insurance Council of Manitoba
- The Insurance Council of Saskatchewan
R. Jennings opened the discussion by asking T. Manjengwa to summarize the state of AI in financial services. He framed the moment in concrete terms: AI is here, but the work now is turning hype into hard-dollar outcomes. The industry’s task is shifting from experimentation to industrialization — from one-off pilots to a disciplined, scalable AI manufacturing line that reliably ships results, not just ideas. He characterized this as the move from AI-curious to AI-capable. The deeper question, he added, is not how to use AI to do what firms already do — but what becomes possible that was previously unthinkable, and whether organizations are building themselves to ask that question seriously.
Canada, he noted, has been punching above its weight: Manulife ranks fifth among the world’s 30 largest insurers in Evident’s global AI maturity index, and is one of only three firms to place in the top ten across every dimension measured. The opportunity is tangible; the question is whether organizations have built the factory to capture it.
T. Manjengwa clarified that when he refers to AI, he means actual machine learning models, including generative AI. He encouraged intentionality and discipline as organizations scale these technologies — moving from a small team of AI engineers familiar with the tools to embedding capability across every level of the business. That structural shift, he argued, will separate firms that lead from firms that stall.
R. Jennings asked how leaders can integrate AI into deeply embedded legacy systems. T. Manjengwa replied: “We are not prisoners of our legacy systems — we are prisoners of the ideas that brought us here.” These systems were built to support the financial institutions of a different era. They have served us, but the operating logic must change. He suggested a three-phase approach:
- Preparation Phase: a complete diagnostic to map current systems, identify what is critical to core operations, and determine the cost and feasibility of modernization.
- Integration Phase: targeted deployments to test value, surface trade-offs (training, infrastructure, temporary productivity dips), and prove out the model before broader rollout.
- Continuous Optimization: monitor, retrain, and iterate. AI capability is a living system, not a finished project.
Asked what insurers must unlearn first, T. Manjengwa invoked Alvin Toffler’s line that the literate of the 21st century will be those who can learn, unlearn, and relearn — and was direct: AI is not a side conversation parallel to business — it is the business conversation. Three specific unlearnings followed: that AI is a technology problem to be solved by data scientists (it is an organizational transformation requiring workflow redesign and change management); that certainty must precede action (it must not, in environments this dynamic); and that pilot culture is sufficient (it is the trap most firms get stuck in). Digitization, he noted, fails when best-in-class tools are layered on top of analogue, redundant processes. Leaders must internalize that change management is not part of what they do — it is what they do. Deploy AI only to replicate existing workflows, he warned, and you risk scaling today’s constraints into tomorrow’s systems.
As much as we shape AI, it also shapes us.
On balancing risk-aversion with the need to experiment, T. Manjengwa was measured: being smart, methodical, and equipped with the right guardrails allows leaders to balance risk and innovation. Active engagement with regulators is essential — calibrating AI guardrails will require many participants. And firms must put the right people in the right seats; AI is complex, and capable, talented teams will make the difference.
On the qualities to hire for, he prioritized two: curiosity and humility. Curiosity drives experimentation and innovation; humility enables course correction when something is not working. Combined with a strong knowledge base, these traits enable an organization to learn, relearn, and unlearn at the pace the technology demands.
On regulatory governance, T. Manjengwa called for a shift from a posture of compliance-and-monitoring toward genuine co-creation between industry and regulators. Regulators will need deeper familiarity with business models to ensure rules promote responsible innovation rather than stifle it. He encouraged the same talent discipline on the regulatory side — the right people in the right roles.
R. Jennings raised the concern around AI replacing humans, asking where human judgment must permanently reside. T. Manjengwa responded that LLMs are probabilistic by design and require a human-in-the-loop, particularly when handling financial information and adverse decisions. Citing Satya Nadella’s recent comments at Davos, he noted that real value comes from grounding these models in firm-specific context — but that is one input, not the final answer. AI should be regarded as a tool to augment judgment, not replace it; offloading menial work — for example, summarizing a 60-page document — frees experts to focus on the complex, the contested, and the consequential. The deeper test, he suggested, is not what AI can do, but what we are willing to entrust to it — and, just as importantly, what we refuse to relinquish.
On work unbundling, T. Manjengwa explained that traditional roles are being decomposed into component tasks, with AI handling structured analysis while humans focus on judgment, relationships, and complex cases. He cited the recent McKinsey CEO framing of an organization with 40,000 humans and 25,000 agents — and the practical reality that managers will increasingly orchestrate both. The claims adjuster is illustrative: AI automates intake and damage estimation; the human is freed for disputed and complex matters. Jobs will not vanish, but they will evolve, and the skills firms recruit for must evolve with them.
The deeper shift, he suggested, is not in how insurers do the work but in what the work itself becomes. McKinsey has framed it as the move from “detect and fix” to “forecast and avert” — from a business that indemnifies losses after they occur to one that uses continuous intelligence to prevent them. Parametric products, real-time risk monitoring, prevention as a revenue line: these are not adjacencies to insurance; increasingly, they are what insurance is becoming. The unbundling of jobs is the operational symptom; the rebundling of the value proposition is the deeper event.
The corollary is a hiring redesign: the goal cannot be turning the next generation of professionals into mere exception-handlers. Roles must be designed around uniquely human skills — judgment, empathy, complex reasoning — paired with fluent AI use.
Asked who is responsible when AI makes or influences a bad decision, T. Manjengwa was clear: accountability sits with the team and the organization, not a single individual. Building interdisciplinary teams — business, regulatory, and technical expertise together — is what makes AI defensible. There will always be a chief analytics officer, but successes and failures should be read as collective. When AI succeeds, it is a team result; when it fails, the same logic applies.
On what firms owe their customers, he was unequivocal: full transparency on how data is used and where AI is involved. In financial services, disclosure is not optional.
R. Jennings asked why so many AI programs fail. T. Manjengwa pointed to MIT’s recent NANDA report finding that 95% of generative AI pilots produce no measurable P&L impact. The root cause is the execution model: teams chase flashy use cases rather than solving real business problems, and they try to do it without seasoned partners. The takeaway is twofold — solve a real business problem, and do not start from scratch. Working with vendors and consultants is a strength, not a weakness; he referenced the Rovr AI build at BMO, which leaned heavily on Microsoft, not for lack of internal talent but because the territory was new for everyone.
Closing the webinar, R. Jennings asked what would constitute evidence that an insurer has moved from AI-curious to AI-capable. T. Manjengwa pointed to several markers: a structured, rigorous implementation discipline that has evolved past use cases into a philosophy of continuous learning and experimentation; visible adoption metrics — Manulife, for instance, has reported a 75% adoption rate of AI tools across its workforce after years of systematic investment; and active regulator engagement. He flagged the regulatory horizon as well — OSFI’s Guideline E-23 on Model Risk Management takes effect May 1, 2027, and applies to all federally regulated financial institutions. The recent CAFII–Deloitte study, in which more than 60% of insurers expect AI to materially impact underwriting and claims, underscores why the move from curious to capable is now an industry-wide imperative.
T. Manjengwa concluded with a quote from John Schaar that reads “The future is not some place we are going, but one we are creating. The paths are not to be found, but made. And the activity of making them changes both the maker and the destination.” The work, he closed, is not merely to adopt tools but to align them to a clear aspiration for the kind of institutions this industry chooses to become. These are questions of identity, courage, and imagination — and the answers we choose will shape us long after today’s tools have changed.
