Generative AI (GenAI) offers the financial services industry unprecedented opportunities to drive innovation, efficiency, and customer value – but implementing a GenAI-powered solution requires careful planning and execution to ensure safety, compliance, and responsible usage. 

Scaling GenAI while navigating dynamic risk and regulatory landscapes can be challenging, and there is an opportunity for organizations to take a leadership role in driving AI governance in their sector and set a clear example for the effective and prudent use of GenAI.  

The EU AI Act and the UK National AI Strategy are two significant regulatory frameworks that aim to guide the development and use of AI within their respective jurisdictions. While these frameworks share common goals in respect of the ethical development and use of AI, they differ significantly in their approach, scope, and enforcement mechanisms.

Capco’s AI Governance Framework ensures organizations are aligned with current global regulations and responsible AI principles while also adhering to a minimal viable governance approach that prevents innovation from being stifled. 

Drawing on insights from Capco's internal GenAI pilot, we outline our six-point framework for safely embedding GenAI solutions in financial services.

1. Establish AI Enhanced Governance

Putting the power of GenAI into the hands of employees with little to no experience of the technology necessitates the creation of tailored GenAI governance controls. Aligning GenAI governance frameworks with existing corporate and IT governance structures increases the likelihood of adoption and success. 

This involves creating interlocks and synergies to enhance overall governance effectiveness, risk management and strategic alignment:

• Engage key stakeholders, including business leaders, risk managers, and compliance teams, to ensure alignment and secure necessary approvals.
• Establish a robust governance structure with clear roles, responsibilities, and accountability. 
• Implement controls and safeguards to mitigate risks such as data privacy breaches, algorithmic bias, or unintended consequences. 

Capco has identified over 90 controls across 11 business areas – underscoring the importance of a rigorous governance framework.

2. Select High Value Opportunities

Organizations should identify a high-priority use case that balances business value and technical complexity. Consider factors such as strategic alignment, data availability, regulatory considerations, and potential risks. "Does the use case require strong management of unstructured data?" and "What level of governance and controls does the use case need?" are key questions to ask. Once you have selected your use case, thoroughly define the problem scope and success metrics. 

3. Prepare Your Data  

Data must be the priority for organizations to obtain the greatest degree of impact and usage from GenAI. 

Prepare your data by ensuring its quality, relevance, and security. Banks’ efforts to date have been focused on structured data, but they will need to review how current controls will scale to also cover all their unstructured data. Unstructured data needs more care and attention, so additional investment in data governance, cataloguing, and curation processes will be required. Subject matter experts should also be engaged to validate assumptions and refine your design.

Many organizations have already begun this journey by enhancing their data governance, management, and architecture initiatives to obtain a better understanding of what data they have, as well as the quality of that data.

4. Conduct Rigorous Testing and Risk Assessment 

Traditional testing frameworks and approaches will not be sufficient for GenAI. New test types and strategies need to be implemented, and continuous testing and monitoring is required to account for GenAI's evolving nature. 

Capco have developed a GenAI testing framework that builds on traditional testing methodologies to accommodate the unique characteristics of GenAI applications. Key tenets include:  

• Conduct thorough testing to validate the performance, accuracy, and safety of your solution. 
• Leverage techniques such as adversarial testing, fairness assessments, and sensitivity analysis to identify and mitigate potential risks.
• Engage risk management and compliance teams to assess the solution against applicable regulations, ethical standards, and company policies. 
• Document risks and mitigation strategies, and establish clear protocols for monitoring and responding to issues.

5. Train and Engage Your Users   

The success of your GenAI solution hinges on the ability of users to effectively interact with – and derive value – from it. Tailored training and support is therefore key, covering both the technical aspects of the solution and the broader context of responsible AI usage.

Capco realized additional efficiencies of up to 50% in certain tasks where individuals had the right training – compared to a decline in efficiency of up to 10% where users did not have the training or knowledge to unlock value. 

By leveraging formats such as ‘promptathons’ – a Capco workforce upskilling program on prompt engineering and GenAI best practices, you ensure safe and effective use of tooling. This fosters a culture of learning and collaboration, encouraging users to share feedback, ideas, and lessons learned.

6. Monitor, Measure, and Iterate

Once your GenAI solution is live, its performance, usage, and impact should be continuously monitored. 

• Track key metrics related to your success criteria, such as productivity gains, quality improvements, or user satisfaction. 
• Leverage automated monitoring tools to detect anomalies, model drift, or potential issues in real-time.
• Adopt Machine Learning Operations (MLOps), an extension of DevOps, to manage the end-to-end lifecycle of your GenAI model. 

In addition, ModelOps – an evolution of Devops and MLOps – provides a framework that defines clear roles and responsibilities for how AI models are built, tested, deployed and monitored across environments. ModelOps enables standardization, scaling, and augmenting AI initiatives to accelerate models transitioning into production. This provides greater monitoring and governance of ML models and operational readiness for AI, while mitigating concerns around drift, bias, explainability and integrity. 

The future is built, not just imagined 

As the financial services industry continues to evolve, the ability to safely and effectively deploy GenAI solutions will be a key differentiator. Those firms that can balance innovation with responsibility and empower their employees to harness the power of GenAI will be best positioned to drive sustainable value and shape the future of the industry. By approaching GenAI pilots with diligence, care, and a spirit of responsible experimentation, financial services firms can begin to build a future that leverages the best of human and machine intelligence to drive transformative outcomes.