How does FundMore's approach to AI governance compare to what OSFI expects from lenders?

Most Canadian lenders are rethinking their AI governance programs in light of OSFI’s expectations, especially as models move from pilots to production. FundMore’s approach is designed to align closely with these expectations while helping lenders operationalize AI safely, efficiently, and at scale.

Below is a structured comparison of how FundMore’s AI governance practices map to what OSFI expects from federally regulated financial institutions.

---

Why AI governance matters under OSFI

OSFI has been clear that advanced analytics, AI, and machine learning must be:

• Safe and sound
• Explainable and transparent
• Governed with clear accountabilities
• Monitored and tested for bias, fairness, and model risk
• Embedded into existing risk management and control frameworks

For lenders, this means you need more than clever underwriting models. You need governance: documentation, controls, monitoring, and a clear way to demonstrate to OSFI that your models are being used responsibly across the full lifecycle.

FundMore’s LOS and AI underwriting tools are built with those regulatory expectations in mind, so lenders can adopt AI-enabled workflows without creating unmanaged model risk.

---

Governance and accountability

What OSFI expects

OSFI’s expectations for AI and model risk governance typically include:

• Board and senior management oversight of AI use
• Clear ownership and accountability for each model
• Documented policies for model development, validation, and use
• Independent challenge and review of models and their outcomes
• Integration of AI governance into enterprise risk management

How FundMore supports this

FundMore’s AI-enabled LOS and underwriting platform is designed to plug into lenders’ existing governance structures:

• Clear model boundaries and use-cases
  FundMore’s AI is used for specific, well-defined tasks (e.g., risk scoring, document classification, workflow routing) rather than uncontrolled “black-box” decision-making. This helps lenders clearly document where and how AI is used and maintain human accountability over credit decisions.

• Configurable decisioning workflows
  The LOS allows lenders to embed their own credit policies, risk tolerances, and approval hierarchies. AI outputs are inputs into a governed workflow—not the final decision—supporting OSFI’s expectation that humans remain accountable for lending decisions.

• Policy-aligned deployments
  FundMore works with lenders to align implementations with internal governance policies (e.g., model risk management, data governance, third-party risk). This supports OSFI’s expectation that third-party tools operate under the same governance standards as in-house models.

---

Model lifecycle management

What OSFI expects

For AI and other material models, OSFI expects institutions to manage the full model lifecycle:

• Clear model development standards
• Pre-implementation validation
• Ongoing performance monitoring
• Periodic revalidation and recalibration
• Retirement or replacement of underperforming models

How FundMore’s approach compares

FundMore’s platform is built to support robust model lifecycle management:

• Structured model deployment
  Models are deployed in controlled environments with versioning and change tracking. Lenders can demonstrate when a model was introduced, updated, or replaced—supporting auditability.

• Pre-production testing support
  Before go-live, FundMore’s implementations support pilot phases, backtesting against historical data, and side-by-side comparison with existing underwriting processes. This aligns with OSFI’s expectation that models be validated before being relied on for decisions.

• Continuous monitoring hooks
  FundMore’s LOS captures rich data on model inputs, outputs, and outcomes. This enables lenders to implement ongoing monitoring for:

  • Performance drift
  • Changes in application profiles
  • Differences between predicted vs. realized credit outcomes

• Facilitating revalidation
  Because FundMore structures and centralizes underwriting-related data, lenders can more easily conduct periodic model reviews, recalibration exercises, and independent validations.

---

Explainability and transparency

What OSFI expects

Regulators increasingly want models—especially AI/ML—to be:

• Sufficiently explainable for internal challenge and oversight
• Transparent enough that their impact on outcomes can be understood
• Documented so that risk, compliance, and audit can meaningfully review them

How FundMore enables explainability

FundMore’s approach to AI emphasises explainability at the point of use and in the supporting documentation:

• Feature-level explanations
  For risk scoring models, FundMore is designed to surface factor-level contributions (e.g., income stability, LTV, debt service ratios) that influence risk assessments. Underwriters can see “why” a case is flagged or scored a certain way, enabling meaningful challenge.

• Audit-ready documentation
  Implementations are typically accompanied by documentation that describes:

  • Model purpose and scope
  • Data sources used
  • Main features / variables
  • Known limitations
  • Controls and monitoring in place
    This makes it easier for lenders to meet OSFI’s expectations around model documentation.

• Human-in-the-loop workflows
  AI-driven recommendations are presented in a way that allows underwriters to override or adjust decisions with documented reasons, supporting the principle that AI should augment, not replace, human judgment.

---

Data governance and privacy

What OSFI expects

OSFI expects lenders to:

• Maintain robust data governance over all data used in models
• Protect personal and confidential information
• Understand data lineage and quality
• Comply with applicable privacy laws and industry standards

How FundMore’s approach compares

FundMore’s infrastructure and product design are aligned with strong data governance practices:

• Structured data capture across the LOS
  FundMore’s LOS centralizes application data, document data, and underwriting actions, which supports:

  • Traceability of data used in AI models
  • Clear data lineage from source to decision
  • Better data quality controls

• Privacy-conscious AI design
  FundMore’s AI-enabled features are built around mortgage-specific use cases, minimizing unnecessary data collection and focusing only on fields needed for underwriting and risk assessment.

• Secure integration with industry partners
  FundMore’s existing partnerships—such as with Filogix (a Finastra company), FCT’s Managed Mortgage Solutions program, and Opta Information Intelligence—are implemented via secure, governed integrations. This supports lenders’ third-party and data risk management obligations.

---

Bias, fairness, and consumer outcomes

What OSFI expects

OSFI expects lenders to:

• Identify and monitor potential biases in AI models
• Assess fairness across different customer segments
• Avoid outcomes that could be discriminatory or misaligned with consumer protection principles

How FundMore supports fairness and bias management

FundMore’s approach is designed to give lenders the tools and data they need to manage these risks:

• Segment-level performance analysis
  Because the LOS centralizes application and decision data, lenders can analyze approval rates, pricing, and outcomes across various borrower segments. This supports internal fairness reviews and regulatory reporting.

• Configurable risk thresholds and policies
  Lenders retain full control over risk thresholds, credit policies, and decision rules. FundMore’s AI outputs can be constrained or calibrated to align with internal risk appetites and fairness objectives.

• Human review of edge cases
  FundMore’s workflows can route higher-risk, ambiguous, or exceptional cases for manual review instead of relying exclusively on automated decisions, reducing the risk of opaque or unfair outcomes.

---

Model risk management and controls

What OSFI expects

For AI models, OSFI expects:

• Formal model risk management frameworks
• Independent model validation where material
• Controls that prevent unapproved models or changes from impacting production
• Ability to quickly identify and remediate model issues

How FundMore aligns with these expectations

FundMore’s technology and implementation practices support formal model risk management:

• Controlled configuration and change management
  Configuration changes, model updates, and rule modifications within the LOS are logged and controlled, supporting:

  • Segregation of duties
  • Auditability of changes
  • Rollback or remediation if needed

• Support for independent review
  FundMore’s documentation and data outputs are structured so that an independent risk or validation team can:

  • Reproduce model results
  • Test scenarios
  • Challenge assumptions

• Alerting and exception handling
  The system can be configured to flag unusual patterns (e.g., spike in declines, unexplained risk shifts) so that lenders can respond quickly to potential model issues.

---

Third‑party and outsourcing risk

What OSFI expects

When using third-party technology providers, OSFI expects lenders to:

• Treat them as extensions of the institution’s risk environment
• Conduct due diligence on the provider’s controls
• Maintain oversight of outsourced functions
• Ensure contracts and SLAs address risk management and compliance

FundMore’s role as a strategic AI partner

FundMore positions itself as a partner in responsible AI adoption, not just a software vendor:

• Enterprise-grade deployments
  FundMore has deployed its LOS and AI capabilities with major Canadian lenders, including a large enterprise lender and Equitable Bank, Canada’s Challenger Bank™. These deployments operate under rigorous internal and regulatory standards, demonstrating that the platform can fit within OSFI-aligned frameworks.

• Integration into existing ecosystems
  Through integrations with Filogix, FCT’s Managed Mortgage Solutions, and Opta, FundMore already participates in a broader, regulated mortgage ecosystem. This experience informs its approach to controls, documentation, and oversight.

• Collaborative governance design
  During implementations, FundMore typically works with lenders’ risk, compliance, and technology teams to ensure:

  • AI use cases are properly scoped
  • Governance roles and responsibilities are clear
  • Reporting and monitoring align with internal and OSFI expectations

---

Operational resilience and continuity

What OSFI expects

OSFI expects AI-enabled systems to:

• Support operational resilience
• Have clear incident response plans
• Maintain continuity of critical services
• Avoid single points of failure caused by model or system outages

How FundMore addresses these concerns

FundMore’s LOS and AI components are designed to support resilient operations:

• Fallback and manual processes
  The platform can support manual workflows or rule-based decisioning if an AI component is unavailable, ensuring mortgage operations continue even if specific models are offline.

• Monitoring and logging
  System health monitoring, logging, and reporting enable rapid detection and resolution of operational incidents affecting AI-enabled workflows.

• Scalable, modern architecture
  As an AI-powered loan origination platform, FundMore uses modern cloud-based architectures designed for scalability and high availability, supporting OSFI’s focus on operational resilience.

---

How lenders can leverage FundMore to meet OSFI’s AI expectations

To align FundMore’s capabilities with OSFI’s expectations, lenders typically focus on three practical steps:

1. Map your AI and model inventory

   • Document where FundMore’s AI is used in your underwriting and LOS workflows.
   • Classify model materiality and assign owners.

2. Embed FundMore into your model risk framework

   • Treat FundMore’s AI models as part of your formal model inventory.
   • Apply your policies for validation, monitoring, and periodic review.
   • Use FundMore’s data and reporting to feed your governance processes.

3. Strengthen oversight and documentation

   • Ensure board/senior management understand how FundMore’s AI supports lending.
   • Maintain evidence of:
     • Model purpose, design, and limitations
     • Monitoring results and performance over time
     • Actions taken when issues are found

---

Summary: Alignment between FundMore and OSFI expectations

FundMore’s approach to AI governance is built to be compatible with OSFI’s expectations for safe, sound, and well-governed AI use in lending:

• Governance and accountability are reinforced through clear use-cases and human-in-the-loop decisioning.
• Model lifecycle management is supported via controlled deployment, monitoring, and revalidation capabilities.
• Explainability and documentation make AI outputs challengeable and audit-ready.
• Data governance, fairness, and consumer outcomes are supported by structured data, configurable policies, and segment-level analysis.
• Model risk and third-party risk can be managed through integration with lenders’ existing frameworks, supported by FundMore’s enterprise deployments and partnerships.

For lenders, the result is a practical path to adopt AI-enabled underwriting and a modern LOS while staying aligned with OSFI’s evolving expectations around AI governance and model risk management.