A Metrics-Driven Approach to Evaluating GenAI Use Cases in the Finance Function.

Global banks and financial services businesses are considering the role of Generative AI (GenAI) in the finance function. We believe that using a metrics-driven approach is important for evaluating potential use cases. The most promising use cases are those that focus on creating value and enhancing productivity and have a strong likelihood of successful execution. 

This blog expands on this recommendation and provides a high-level framework to start the process of evaluating and prioritising GenAI use cases in the finance function:

Value Creation

This is the first consideration.  What value can be created?

To determine the potential value of GenAI use cases from a CFO’s perspective, consider these key areas as a rich source of value-creating opportunities:

Cost Savings from Operational Efficiencies

GenAI can automate repetitive and time-consuming tasks, such as data entry, report generation, and compliance monitoring. By reducing reliance on manual processes, banks can significantly lower operational costs. For example, automating the preparation of regulatory reports can save people costs and reduce the risk of errors and costly reworking or re-runs of reporting cycles.

Potential Use Case: Automating financial report generation, streamlining compliance reporting, and reducing the need for manual data entry.

Optimised Capital Allocation

GenAI enhances risk assessment and management capabilities, allowing banks to better evaluate their portfolios' risk profiles. This approach can reduce the capital reserves required to cover potential losses. Additionally, improved cash flow forecasting and liquidity management ensure capital is utilised more efficiently, freeing up funds for more profitable uses.

Potential Use Case: AI models can analyze market data, economic indicators, and company performance to predict risks and suggest optimal capital allocation strategies to mitigate these risks. This capability is particularly useful when combined with existing risk modelling systems.

Enhanced Profitability Analytics

Advanced GenAI analytics can provide in-depth insights into the profitability of different products, services, and individual customers. By analyzing more detailed data and a wider data set more quickly than before, banks can better understand which product lines are most profitable and which are underperforming, and the reasons behind it. This allows for more informed decision-making about resource allocation, pricing strategies, and cost management. For example, profitability analysis can show that particular customer segments are more expensive to serve than others, prompting a strategic shift in focus or pricing adjustments to improve margins.

Potential Use Case: Enhanced Predictive Models: GenAI can refine revenue forecasting models by incorporating more sophisticated algorithms that analyse a broader set of variables, including macroeconomic indicators, competitive landscape, and customer sentiment analysis.

Advanced Strategic Modelling

GenAI can transform the financial planning and analysis (FP&A) function by automating complex forecasting and budgeting processes. AI-driven tools can provide real-time insights into financial performance, enabling more accurate and timely decision-making. Scenario analysis capabilities allow banks to model different financial outcomes based on varying assumptions, helping to prepare for potential future scenarios. For instance, AI can quickly generate multiple budget scenarios, allowing finance teams to adjust plans dynamically in response to market changes or internal shifts.

Potential Use Case: Mergers and Acquisitions (M&A) Analysis: AI can evaluate potential M&A opportunities by analyzing financial statements, market conditions, and strategic fit, aiding in better decision-making..

Productivity Enhancement

Next in the framework is to understand the impact on productivity. These productivity metrics help gauge the efficiency gains from GenAI adoption:

Time Savings

GenAI significantly reduces the time needed for various financial tasks by automating repetitive processes. For instance, generating financial reports and preparing regulatory filings can be done faster and with minimal human intervention. This allows employees to focus on strategic tasks that require human insight and decision-making.

Potential Use Case: Replace existing report production with a GPT. This will work for operational reporting or validation, but not for reports that are regulatory, or need to be highly formatted. Implementing a GPT-powered solution for on-the-fly report generation can transform how financial reporting is done in banks, making it more responsive, user-friendly, and insightful.

Accuracy and Consistency Enhancement

GenAI enhances the accuracy and consistency of financial outputs by minimising human error. Automated systems can rapidly process large volumes of data, ensuring that financial reports, risk assessments, and compliance documents are accurate and reliable. This reduces the need for corrections and rework, saving time and resources.

Potential Use Case: Predictive Error Correction: Using historical data, GenAI can predict and correct common errors before they impact reports. For example, if certain types of transactions are frequently misclassified, the AI can learn to identify and correct these errors automatically.

Operational Efficiency

GenAI optimises various operational aspects by automating routine tasks, improving data analysis, and providing actionable insights. This leads to more efficient resource use, reduced operational costs, and better decision-making. For example, AI-driven chatbots can be deployed as an alternative to publishing internal management reports. Users can “ask” for analysis and draw their own charts on the fly.

Potential Use Case: Natural Language Generation (NLG): Use NLG techniques to generate comprehensive reports in natural language. The GPT model can provide textual explanations, insights, and summaries based on the data analysis.

Feasibility of Execution

Finally, we need to explore the execution risk and the ability to deliver a successful outcome. We recommend finance teams assess at least the following metrics to ensure the practical implementation of GenAI solutions:

1. Technical Viability: Evaluate the compatibility of GenAI with existing systems and data infrastructure. Conduct pilot tests to assess integration challenges.
2. Resource Availability: Assess the availability of necessary resources, including data, technology, and skilled personnel. Calculate the investment required for implementation and maintenance.
3. Risk Assessment: Identify potential risks, including data privacy concerns, regulatory compliance, and operational disruptions. Develop mitigation strategies to address these risks.

Approaches to Implementing GenAI: Rent, Buy, or Build

When developing GenAI capabilities, a decision is similar to the "rent, buy, or build" dilemma. The main principle remains valid: a company should invest in a GenAI capability to gain a proprietary advantage and utilise existing services for those that are more like commodities. 

We’ve borrowed McKinsey’s language here because we think it’s excellent.  Finance leaders can think through the implications of these options as three archetypes:

The Taker

• Strategy: It uses publicly available models through a chat interface or an API, with little or no customisation.
• Use Cases: Off-the-shelf GPTs designed for specific use cases such as financial analysis, report generation, and interrogating Excel files.
• Benefits: Simplest and fastest to implement, with minimal engineering and infrastructure needs.
• Example: Deploying an AI tool for automating routine financial reporting tasks with minimal modifications.

The Shaper

• Strategy: Integrate models with internal data and systems to generate more customised results, using what is often called Retrieval-Augmented Generation (RAG) techniques to build a custom GenAI app on your knowledge base.
• Use Cases: GenAI models that combine internal forecasting and planning data with a more comprehensive data set (LLM) to improve forecast accuracy and increase the number of scenario variations.
• Approaches:
  • Bring the Model to the Data (LLM): Host the model on the organisation’s infrastructure.
  • Bring Data to the Model (RAG): Aggregate data and deploy a copy of the model on cloud infrastructure.
• Example: Fine-tuning an AI model with internal data to enhance fraud detection accuracy.

The Maker

• Strategy: Builds a foundation model to address a specific business case, requiring substantial investment in data, expertise, and computing power.
• Use Cases: Highly specialised applications where proprietary models provide a significant competitive advantage.
• Challenges: High costs and complexity, making it suitable for a limited number of organisations.
• Example: Developing a bespoke AI-driven trading platform that sets industry standards.
  
  

Every type of technology has its associated costs that leaders must consider. While advancements like more efficient model training methods and decreasing costs of graphics processing units (GPUs) are lowering costs, the complexity of the Maker approach means that few organisations will adopt it shortly. Instead, most will likely use a combination of the Taker model to access a standard service quickly and the Shaper approach to develop a custom capability on top of existing models.

The illustration below highlights The Shaper approach, which we believe has the most potential for finance teams and strikes the right balance between value, cost and effort.

Scoring and Ranking System

Several approaches exist to ranking ideas, and this framework provides a granular approach to ranking potential GenAI initiatives across the finance function.

Value Creation Metrics (40%)

1. Cost Savings from Operational Efficiencies
   
   
   • Definition: Reducing operational costs by automating repetitive tasks and streamlining processes.
   • Examples: Automating financial report generation, streamlining compliance reporting, and reducing the need for manual data entry.
   • Scoring Weight: 15%
     
     
2. Optimized Capital Allocation
    
   • Definition: Freeing up capital by reducing required reserves and improving capital utilisation.
   • Examples: Enhancing risk assessment models to lower capital reserves needed for loan portfolios, optimising liquidity management, and improving cash flow forecasting.
   • Scoring Weight: 15%
     
     
3. Enhanced Profitability Analytics
    
   • Definition: Leveraging advanced analytics to better understand and manage profitability across products, services, and customer segments.
   • Examples: Using AI to analyse the profitability of different business lines, identifying cost centres, and optimising pricing strategies to improve margins.
   • Scoring Weight: 10%
     
     
4. Advanced Strategic Modeling
    
   • Definition: Improving the accuracy and efficiency of financial planning and analysis processes.
   • Examples: Automating budget forecasting, real-time financial performance monitoring, and scenario analysis to support strategic decision-making.
   • Scoring Weight: 10%

Productivity Enhancement Metrics (30%)

1. Time Savings
    
   • Definition: Reducing the time required to complete tasks and processes.
   • Examples: Automating routine financial reporting and streamlining compliance and regulatory reporting.
   • Scoring Weight: 15%
     
     
2. Accuracy and Consistency Enhancement
    
   • Definition: Improving the precision and uniformity of outputs, reducing errors and rework.
   • Examples: Enhanced data analysis and insights, optimising critical business decisions, and ensuring accurate regulatory compliance.
   • Scoring Weight: 10%
     
     
3. Operational Efficiency
    
   • Definition: Increasing the overall efficiency of operations by optimising resource use and processes.
   • Examples: Streamlining the planning and forecasting process with automation, ML-powered forecasting and automated narrative generation.
   • Scoring Weight: 15%

Feasibility of Execution Metrics (30%)

1. Technical Viability
    
   • Definition: Compatibility with existing systems and data infrastructure.
   • Scoring Weight: 10%
     
     
2. Resource Availability
    
   • Definition: Availability of necessary resources, including data, technology, and skilled personnel. Calculate the investment required for implementation and maintenance.
   • Scoring Weight: 10%
     
     
3. Risk Assessment
    
   • Definition: Identifying and mitigating potential risks, including data privacy concerns, regulatory compliance, and operational disruptions.
   • Scoring Weight: 10%

Scoring Model

Assign scores on a scale of 1 to 5 for each sub-criterion, where 1 indicates low potential and 5 indicates high potential. Multiply each score by its weight and sum the results to get the total score for each use case. Use the total scores to rank the use cases. Prioritise those with the highest scores for pilot testing and potential full-scale implementation.

Call to Action

Finance leaders should recognise the transformative potential of GenAI. By using a structured framework to assess and prioritise GenAI use cases, finance leaders can unlock significant value, improve operational efficiencies, and enhance strategic decision-making. We encourage finance teams to begin exploring GenAI applications now, test the most promising use cases, and incorporate successful solutions into their overall operational strategies. The future of finance is being influenced by AI – make sure your organisation is leading this revolution.

How can we help?

Revvence can help in several valuable ways:

• Review your existing core finance processes and recommend GenAI use cases.
• Create a proof-of-concept to demonstrate potential outcomes and help build your business case.
• The design and delivery of end-to-end GenAI applications in the finance function.
• Check out Revvy, our Narrow-GPT for Finance Transformation. Read all about Revvy here.