Kyriba has made a number of product and service announcements this month. Perhaps the most interesting is the introduction of new AI features within its platform designed to improve cash forecasting, bank connectivity-as-a-service and custom report generation.
The company says it is able to increase predictability of cash, risk and liquidity decisions and drive new levels of liquidity performance because its platform is designed to leverage large data sets without interruption.
So how exactly does Cash Flow AI enable treasurers to incorporate more advanced data into their forecasting processes? Bob Stark, Global Head of Market Strategy at Kyriba explains that the embedded model is trained in historical cash and bank transactions to make predictions, taking into account a number of variables such as seasonality and business cycles.
“Once an alternative AI forecast is generated, this can be compared against other forecast models to measure which projections are more accurate and therefore when the company should be using AI and when, for example, it should rely on data from the ERP,” he says. “This reconciliation and measurement is critical to understanding the right roles for AI to increase forecast confidence.”
One of the major pain points for ERP-to-bank connectivity projects is support for payment formats.
“When a new payment is required, we will develop that new format need and make it available for any customer to use to connect their treasury and ERP-initiated payments to their banks,” says Stark. “With AI, we have an additional option for new payment formats where an individual customer can customise an existing format with ChatGPT in real time and apply that to their payments on the fly.”
When using AI to make business-critical decisions, treasurers must ensure the data the technology uses to make these decisions is accurate and has not been compromised in any way. It is also crucial that provisions are in place to ensure automated processes remain compliant with regulatory requirements.
Stark notes that in this instance customers are only using their own data. “Furthermore, when we are embedding these AI models within our platform, the customer isn’t sharing their data with a large language model,” he adds. “Our customers were clear that data privacy continues to be a critical requirement and we have invested in and received certifications such as the EU-US Data Privacy Framework.”
Kyriba’s AI strategy consists of developing embedded AI models within its platform and customer workflows as well as opening the platform to leverage and integrate third party GenAI solutions. The embedded AI models are closed models that are built into processes such as cash forecasting, payment fraud detection and cash optimisation.
“Opening our platform via our OpenAPIs enables data to securely and automatically be presented to a third-party tool such as ChatGPT for enrichment or update, with the results automatically ingested into the platform,” says Stark. “An example of this is payment formats, where customers can use ChatGPT to customise bank payment formats and integrate those scripts into their own payment format library.”
The company has been conducting formalised pilot programmes for new functionality such as the Open Reports Studio, which uses ChatGPT and Microsoft Copilot. These pilot programmes allow customers to test new capabilities with their data to evaluate not only how the functionality works but also how it fits alongside existing processes.
“This is especially important when AI is used to complement rather than replace existing processes and data,” says Stark.
