Access to data

Each form of AI relies on the technology having access to the appropriate and necessary data. At the same time, companies need to store data in a way that prevents unauthorized parties, both internal and external, from gaining access to proprietary information, intellectual property or customer information.

Managing this balance between access and control can be achieved by setting a data strategy. A company’s data strategy will be determined by its specific requirements and should cover the following elements:

• How data is understood and recognized. What data does the company have, how and where (e.g., in which system) is data stored, and which solutions and systems have access to specific data? Data may need to be tagged to be accessible to various solutions.
• How data is collated. Where does the required data come from: internal systems (e.g., TMS or ERP system) or external sources (e.g., banks)? How is data captured from those various sources? Treasury needs to be confident that any data collation process is secure (i.e., third parties cannot gain access to and/or alter any data) and efficient (i.e., data is collated in a timely fashion). The use of API technology can ease the collation process.
• How data is controlled. One requirement is to ring-fence confidential data (e.g., sensitive internal data, as well as customer and supplier data) to control its use by AI-driven solutions. Any AI-driven treasury technology (e.g., a cash forecasting solution or GenAI) should only use approved and authenticated data (e.g., the data should not be subject to copyright restrictions), and it should not breach any regulatory restrictions (e.g., the timely publication of market-sensitive information).

Understanding of AI processing

It is important that treasurers understand the capabilities of AI and how it can be harnessed to improve operational efficiency in departmental processes. Mapping existing processes and finding any specific pain points will help to identify where an AI-enabled solution could fit within an operational workflow.

Any new solution should provide an output that is better — more accurate, faster and/or more efficient. During testing, and once implemented, treasurers need to be able to perform a form of variance analysis. This serves two purposes. First, it allows the new tool’s effectiveness to be analyzed. Second, identifying the reasons behind any variances will allow treasurers to continue to improve the process to achieve further accuracy and efficiency.

Ability to deploy AI

Finally, there is the perennial problem of resourcing the implementation of any new technology. In the 2025 AFP Treasury Benchmarking Survey, 60% of respondents identified high costs and a lack of resources as barriers to the adoption of new technology, including AI. Half also identified a lack of skills within treasury to use new technology and a limited availability of IT staff to implement it as a further barrier. The good news is that only 7% of senior managers and 10% of treasury staff are unwilling to adopt new technology. The incentive for companies is simple: “Company valuations are higher for those organizations that have an AI strategy,” said James Kelly, Co-founder, Your Treasury.