Transformation Solutions

AI/ML Capability Framework Implementation

Challenge

A large APAC-based universal bank had previously engaged Harmonic Strategy to design a Data Capability Framework for AI/ML (see Data Case #1). Following the success of that strategic blueprint, the bank invited Harmonic back to lead the implementation and transformation phase — building the project management models, PMO governance, and delivery frameworks required to operationalise AI/ML capabilities enterprise-wide.

The challenge was to translate the strategic vision into executable delivery structures, balancing speed, compliance, and scalability. With multiple concurrent use cases spanning credit risk, customer analytics, and operational efficiency, the bank required a disciplined yet adaptive approach that integrated Agile execution, regulatory oversight, and MLOps maturity under a single delivery umbrella.

The initiative needed to:

• Define a fit-for-purpose hybrid project management model for regulated AI/ML deployment.

• Build a PMO and governance layer aligned with APRA and internal model risk frameworks.

• Establish delivery processes that ensure traceability, explainability, and compliance across all AI initiatives.

Solution

Harmonic Strategy acted as Lead Transformation Consultant, designing and deploying a practical project management and delivery framework tailored for enterprise-grade AI/ML rollout within a regulated banking environment. The engagement integrated hybrid Agile-Waterfall models, CRISP-DM methodology, MLOps lifecycle orchestration, and governance-aligned delivery checkpoints.

1. Hybrid Agile–Waterfall Governance (Agile@Scale)

• Combined Waterfall for strategic planning, approvals, and compliance checkpoints with Agile for use case execution and sprint delivery.

• Established governance controls including Model Review Boards (MRBs), risk sign-offs, and executive reporting.

• Implemented Jira and Confluence for sprint tracking, documentation, and visibility across squads (e.g., Lending AI, Risk Analytics, FM Data).

• Introduced Scaled Agile (SAFe) principles for multi-squad coordination and dependency management.

2. CRISP-DM for AI/ML Use Case Development

• Adopted the Cross-Industry Standard Process for Data Mining as the foundation for AI lifecycle management.

• Aligned phases — Business Understanding, Data Preparation, Modelling, Evaluation, Deployment — to internal risk and compliance checkpoints.

• Embedded explainability frameworks (SHAP, LIME) and post-hoc audit trails to ensure transparency and model accountability.

3. MLOps Lifecycle Framework

• Designed a bank-wide MLOps model lifecycle integrating model development, deployment, and monitoring.

• Implemented CI/CD pipelines for model versioning and retraining using tools such as MLflow, Azure ML, and Databricks.

• Introduced Prometheus and Grafana dashboards for live monitoring of model drift, bias, and accuracy decay.

• Enabled blue/green deployment practices to support continuous delivery in production environments.

4. Governance-Integrated Delivery Lifecycle

• Embedded regulatory oversight across every phase — from use case intake and ethics review to post-deployment validation.

• Integrated model risk validation in alignment with APRA CPG 229 and SR11-7 standards.

• Implemented delivery checkpoints for privacy, consent, and ethical screening, supported by Collibra, Alation, and RSA Archer tools.

5. Program Oversight & PMO Setup

• Established a central AI Transformation PMO responsible for roadmap management, dependency mapping, and reporting.

• Built executive dashboards tracking sprint health, model release velocity, and risk closure rates.

• Developed templates for benefits realisation, issue escalation, and stakeholder communications.

Key Deliverables

• Enterprise AI/ML Project Management Framework (Hybrid Agile–Waterfall)

• CRISP-DM-based Use Case Delivery Templates

• Bank-Wide MLOps Lifecycle Architecture

• Governance and Ethics Review Framework (aligned to APRA and global standards)

• Centralised AI PMO Operating Model & Reporting Dashboard

• Delivery Playbook with Stage-Gate Checkpoints and KPI Metrics

Client Benefits

• Accelerated AI delivery through standardised project execution models.

• Enhanced model governance ensuring regulatory compliance and ethical AI adoption.

• Improved collaboration between data scientists, risk managers, and business leads.

• Continuous improvement capability via automated retraining and real-time model performance insights.

• Institutionalised PMO oversight, providing full transparency from ideation to production.