The treasury world is teeming with interest in AI, a technology that promises to redefine how companies manage cash and liquidity, mitigate risks, and automate processes. Yet, for many treasurers, the transition from interest to implementation presents a challenge, with the biggest hurdle being how to unlock AI’s full potential in a practical and impactful way.
To tackle this challenge and bring clarity to the decision-making process, it is important to decide the right AI use cases for their functions. As a starting point, it can be helpful to use a structured framework that examines AI adoption through two critical lenses: the potential for the greatest impact and the ease of implementation. Using such a framework will allow organisations to identify and prioritise relevant use cases that will deliver the most value for them.
Looking at the first lens, the impact of AI on treasury operations can be assessed effectively by focusing on two key areas for potential improvement: the effort required for current operations and the accuracy of data outputs. For example, if a treasury team spends excessive time on preparing management reports, integrating AI can enhance and automate these tasks, freeing the team’s time to focus on more strategic initiatives.
Similarly, organisations struggling with data accuracy, such as in cash-flow forecasting, can leverage AI tools to enhance precision by analysing historical patterns and continuously refining predictions. By identifying these concerns, treasurers can pinpoint which AI applications will deliver the most significant benefits to their operations.
High-quality data is essential, requiring consistency, structure and controls to ensure accuracy and reliability
The second lens helps treasurers assess how easily AI can be integrated into their operations. This requires not just a deep understanding of their business, but also a basic knowledge of AI’s application – i.e., how it functions and how to deploy it effectively. AI deployment becomes much easier when key enablers are in place, specifically, high-quality data and the necessary infrastructure and tools to train, integrate and operate AI effectively.
Building on this, treasurers must evaluate their organisation’s maturity and readiness for AI adoption. High-quality data is essential, requiring consistency, structure and controls to ensure accuracy and reliability. A treasury management system can support this by automating processes and maintaining structured data. Beyond data, factors such as AI expertise, proprietary models and infrastructure play a key role. However, many of these can be outsourced or developed progressively as part of the AI deployment process.
Treasurers can also look to proven AI use cases from other organisations to guide their own implementation decisions. High-feasibility applications include cash-flow forecasting and fraud detection, both of which have shown success in treasury. Learning from these established cases helps anticipate challenges and avoid roadblocks, whereas implementing something entirely new may be more complex and difficult to execute.
The matrix below offers a helpful starting point for treasurers to identify and prioritise relevant use cases, providing an illustrative perspective to support their assessment. The use cases and their placement may vary depending on the organisation’s needs, treasury processes and existing IT/data infrastructure.
Ultimately, treasurers must decide whether to introduce AI as a small-scale innovation or a full-scale transformation. Whatever the approach, it must strike the right balance between impact and feasibility while aligning with the organisation’s short- and long-term goals.
Treasurers understand their operations best, but bringing in experts who bridge treasury and technology can help ensure AI is implemented in a way that is feasible, effective and aligned with their organisation’s needs.
After exploring the framework and selecting the right AI solution, the next steps involve preparing to deploy AI. Achieving AI readiness involves many facets, the most important of which is ensuring availability of appropriate data for the selected use cases. An AI model is only as good as the data on which it is trained. High-quality, relevant data ensures the model can identify meaningful patterns, generate accurate insights and provide the desired outcome.
Inaccurate data leads to flawed outputs, undermining the model’s effectiveness and business value. For example, when AI models perform cash-flow predictions, they rely on accurate invoice data and customer databases, typically sourced from the finance department, accounts receivable and customer relationship management systems.
But data selection is just the start, as strong data governance is also crucial. This includes defining data sources, ownership and controls, to maintain accuracy, consistency and continuous reliability.
The data-preparation process starts with collecting and cleansing data to make it suitable for training the AI model. This step is a collaborative effort, requiring the expertise of IT and data experts to prepare and transform the data, while the business and treasury teams ensure the objectives of AI enablement can be met from the transformed dataset.
Given the complexity of AI implementation and the early stage of adoption in treasury operations, a gradual approach is key
Choosing the right AI models tailored to the specific use case is equally important. For instance, machine-learning models typically excel at tasks involving structured and data-intensive activities, such as prediction, classification and trend analysis. Conversely, generative AI and large language models are better suited for tasks that require human-like reasoning and creativity.
Beyond ensuring the availability of the correct data and AI models, assembling the right team and having the right infrastructure is essential. A cross-functional team consisting of treasury, business users, IT and data experts, along with reliable technology, scalable platforms and seamless data access, help ensure AI initiatives are properly planned and aligned with business goals.
Given the complexity of AI implementation and the early stage of adoption in treasury operations, a gradual approach is key. Starting with a proof of concept using a limited dataset or data from a limited number of entities allows for controlled testing, refinement and a clearer path to scaling AI effectively. This measured approach lowers risk by limiting exposure and reducing the impact of potential setbacks. A successful proof of concept builds confidence, strengthens the business case and paves the way for broader adoption.
Treasurers can also look beyond their function and leverage insights from other departments already using AI. For instance, if another division within finance has implemented a similar use case, it will be easier to adapt their existing models to a new dataset rather than starting from scratch. Engaging with teams within the organisation that have undertaken similar initiatives can provide valuable guidance and accelerate the implementation process.
For treasurers deploying AI for the first time, the journey is full of unknowns. Defining success and setting clear stopgaps is essential to knowing when to pivot. A proven approach is to break success into trackable milestones aligned with objectives, budgets and efforts spent. If the desired outcomes are not met, it is crucial to recognise the limitations, take the lessons learned, and move on to a use case with greater potential value.
The aforementioned steps are effective for initial AI deployments on a smaller scale, but long-term success requires a strategic approach. To maximise AI’s impact, treasurers should look beyond individual use cases and define a broader strategy that aligns with the organisation’s long-term goals. This ensures AI adoption remains sustainable and scalable, and delivers lasting value.
Nish Nagpal is a corporate treasury adviser at PwC; Nandini Soondram is a manager in the corporate treasury team at PwC
This article first appeared in The Treasurer Issue 2, 2025