• Written by Muhammed Abdulla NC, CEO

The notion that AI cost in finance is primarily driven by the cost of the AI model itself is a misconception that can lead to significant oversights in budgeting and planning. In reality, the majority of spend accumulates in integration and data work, making it essential to reassess where resources are being allocated.

Integration and data work as the dominant line

When deploying AI in finance, the cost of integration and data work far surpasses that of the AI model. This is because AI systems require high-quality, relevant data to produce accurate results, and integrating these systems with existing infrastructure can be a complex and time-consuming process. A finance manager, for instance, may need to work with a data engineer to ensure that the AI system can access and process the necessary data from various sources, such as accounting software and customer relationship management systems.

The effort required to integrate AI systems with existing infrastructure and data sources can be substantial, involving significant resources and expertise. As a result, it is crucial to factor in these costs when planning and budgeting for AI deployments in finance. By doing so, organisations can avoid underestimating the total cost of ownership and ensure that they have the necessary resources to support the successful deployment of AI systems.

Run cost against query volume

Another critical aspect of AI cost in finance is the run cost, which is directly tied to query volume. As the number of queries increases, so does the cost of running the AI system. This is because each query requires computational resources, such as processing power and memory, to execute. A shift supervisor, for example, may use an AI-powered system to analyse maintenance logs and identify potential issues, generating a high volume of queries that can impact run costs.

Understanding the relationship between query volume and run cost is essential for managing AI expenses in finance. By monitoring query volume and adjusting resource allocation accordingly, organisations can optimise their AI deployments and minimise unnecessary costs. This may involve implementing measures to reduce query volume, such as caching frequently accessed data or optimising query algorithms.

The cost of a wrong answer nobody caught

A wrong answer that goes unnoticed can have significant consequences in finance, resulting in costly mistakes and reputational damage. The cost of a wrong answer can be substantial, and it is essential to consider this risk when evaluating AI deployments. A category manager, for instance, may rely on an AI system to analyse supplier data and identify the best options, but if the system provides incorrect information, the organisation may end up with suboptimal suppliers, leading to increased costs and reduced quality.

Most people get wrong the idea that AI systems are foolproof and that wrong answers are rare. However, AI systems are only as good as the data they are trained on, and errors can occur. It is crucial to implement robust testing and validation procedures to ensure that AI systems are providing accurate results and to detect any errors before they cause harm.

Attributing spend to a process rather than a model

When evaluating AI cost in finance, it is essential to attribute spend to the underlying process rather than the AI model itself. This is because AI systems are often used to automate or augment existing processes, and the costs associated with these processes should be factored into the overall cost of the AI deployment. For example, an organisation may use an AI system to automate accounting tasks, but the costs associated with the accounting process, such as labour and software costs, should be included in the overall cost of the AI deployment.

By attributing spend to the underlying process, organisations can gain a more accurate understanding of the total cost of ownership and make informed decisions about AI deployments. This may involve working with finance teams to identify areas where AI can be used to reduce costs or improve efficiency, such as streamlining financial reporting or automating compliance tasks.

What this does not do

The points outlined above highlight the importance of considering integration and data work, run cost, and the cost of wrong answers when evaluating AI cost in finance. However, this does not mean that AI deployments are always the right choice. In some cases, alternative solutions, such as manual processing or traditional software, may be more cost-effective or better suited to the organisation's needs. If you are unsure about the best approach for your organisation, contact us to discuss your options and determine the most effective solution.

Ultimately, understanding the true cost of AI in finance requires a nuanced approach that takes into account the various factors that contribute to the total cost of ownership. By considering these factors and evaluating the potential benefits and drawbacks of AI deployments, organisations can make informed decisions and ensure that their AI investments deliver value and drive business success.