GenAI and the future of finance: ensuring good governance and maximizing returns

Apr 26, 2024

IT
finance

share on

Driven by AI, finance departments are pivoting from traditional analysis to a more streamlined, automated approach that focuses on planning, risk management, and strategic decision-making. But that’s not the only way in which AI and genAI are changing work for finance teams. The CFO, together with the CIO, also plays a key role in AI selection, adoption, and governance.

The main task of the finance department is to ensure that the organization’s financial resources are used efficiently and in a way that aligns with short-term requirements and the long-term business strategy. This makes finance a de facto gatekeeper when it comes to introducing new technologies, like AI and genAI. Generally speaking, there are two main considerations: return on investment (ROI) and risk.

The price of AI

The finance team is charged with the task of making a comprehensive cost-benefit analysis of AI technology and AI use cases. This requires a basic understanding of AI models and functionalities. “For example, the price of large-language models (LLMs) can vary greatly,” says Wouter Labeeuw, AI expert at delaware. “Choosing the most suitable one for each use case is paramount.”

“To give you an idea, GPT-4 costs 10 to 15 times more than GTP-3.5, so it makes sense to test the latter first and see how far it gets you. Moreover, as all major models seem to converge in terms of intelligence level and functionality, we’re witnessing a rise in smaller and cheaper models aimed at specific use cases. Figuring out which one works best for what can be very cost-effective.”

GPT-4 costs 10 to 15 times more than GTP-3.5, so it pays to test the latter first and see how far it gets you.

Wouter Labeeuw, AI expert at delaware

Risky business

On top of cost-benefit analysis, finance is often charged with risk management as well. AI certainly adds another level to that – not limited to the financial risks of using overpowered and overpriced AI models for relatively simple use cases. These include:

Data leakage or exposure of sensitive information when using external or personal AI tools or models.
Legal or ethical issues when using AI-generated images or texts without proper attribution or consent.
Quality or reliability issues when using AI models or tools that are not properly trained, tested or monitored.

How AI is revolutionizing finance

Businesses all over the world are already using genAI across their full supply chain, including finance operations. “CFOs and their teams have to sift through heaps and heaps of data and reporting every day,” says Stijn Robberechts, solution lead SAP Sell, Procure & Deliver at delaware. “GenAI empowers them to find what they’re looking for faster, while also offering valuable insights and even very specific next-step recommendations.”

Some of the key advantages of genAI in finance include:

Saving time by automating time-consuming manual processes like data entry, invoice processing and inventory management.
Reducing costs by identifying inefficient supply chain processes and risks and reducing unnecessary spending.
Improving accuracy by automating tasks like forecasting and planning (and taking out the risk of human error in the process).
...

Mitigating risks with good governance

One way to mitigate these risks is with good governance: the CFO and CIO should work together and build a comprehensive framework to ensure the responsible deployment and use of AI technology. The concept of ‘people, process, and technology’ is fundamental here.

People: Thoroughly understand users’ needs, create room to experiment, and offer training where needed. This keeps users from looking for shadow-IT solutions instead.
Processes: Make sure responsibilities, ownership, and when and where a certain tool can be used are clear from the start. Assigning an ‘AI champion’ for each business department can help with this.
Technologies : In terms of technology, access to high-quality data, the right AI-models for each use case, and monitoring and auditing tools are key. Without these, effective AI governance – and thus risk mitigation – is nearly impossible.

Mitigating the risks of AI requires solid governance that takes into account people, processes, and technologies.

Stijn Robberechts, solution lead SAP Sell, Procure & Deliver

From experiment to prototype

Deploying AI in an organization – and thus also in your finance team - follows this three-step process.

Experiment: Provide your team with AI tools they can actually use in their daily work and present these as good alternatives for their own solutions or apps. Let users experiment with different models in a so-called ‘playground’ environment to figure out what works best with each use case. Here, the model doesn’t have access to sensitive or critical company data.
Develop: Set up a development environment with the right tools and a specific scope of production data, providing a certain level of ‘controlled freedom’. Key here is that the department itself gets ownership over development (and related costs) – of course while still getting the necessary guidelines and support from IT.
Industrialize: Before an AI use case can be scaled up, it needs to be robust, safe and repeatable. This requires safety controls, quality tollgates, successful prototyping, and more.

Putting AI into practice... with delaware

“At delaware, we have over a decade of experience with AI,” says Stijn Robberechts, solution lead SAP Sell, Procure & Deliver at delaware. “In the beginning, this included mainly machine learning, visual quality inspections, and other data-driven projects. However, with genAI, there’s a whole new world of opportunities opening up.”

“Over the past few years, we have invested heavily in building a team of dedicated experts to help our customers leverage the full power of AI. The focus here lies on delivering value, which is why we immediately bring an industry focus to our inspiration sessions. The next step, then, are our ideation sessions, in which we consider the feasibility and business value of AI use cases. Some of these will be turned into pilot projects that can be scaled up – or not, depending on how much value they generate.”