Agentic AI could automate 80% of routine treasury work, leave strategic decisions to humans: EY

New Delhi [India], September 4 (ANI): Agentic artificial intelligence could handle around 80 per cent of routine coordination and processing in treasury operations, allowing professionals to focus on risk management, strategic funding and other decisions requiring human judgement, according to a report by EY.
The report said agentic AI could significantly change treasury operations, but companies must first establish reliable data systems and automated workflows before adopting the technology at scale.
EY said a mature treasury operating model could see about 80 per cent of routine coordination and processing handled by automated workflows and AI agents, while the remaining 20 per cent involving judgement, risk assessment and strategic decisions would continue to be managed by treasury professionals.
The objective is not to reduce headcount but to move treasury teams away from low-value administrative tasks and towards activities that create greater business value, the report said.
According to the report, 60-70 per cent of treasury bandwidth at mid-levels is currently spent on low-value activities, including collecting data from different systems, manual validation and preparing reports.
More than 50 per cent of companies globally also continue to rely on manual batch reconciliation, highlighting the scope for automation, it said.
Agentic AI could also improve liquidity forecasting, with forecast accuracy of around 90 per cent potentially achievable for 30-, 60- and 90-day liquidity periods.
Higher forecasting accuracy could help companies reduce excess liquidity buffers and free up capital, the report said.
However, EY cautioned that companies should address fragmented data systems before deploying agentic AI. A treasury data lake that integrates information from enterprise resource planning systems, treasury platforms, banks and other sources is essential for reliable AI-based decision-making, it said.
“The path forward runs through the datalake, not around it,” EY said.
The report recommended that companies begin with smaller and lower-risk use cases, including cash forecasting, cash reconciliation and know your customer (KYC) and anti-money laundering (AML) exception handling.
Reconciliation and reporting could be among the most effective starting points because they involve significant manual effort and offer measurable outcomes, it said.
EY also recommended an “advisory before autonomous” approach, under which AI provides recommendations that are reviewed and approved by humans. It suggested that companies avoid autonomous payment execution during the pilot phase.
For treasury leaders, the immediate priorities should include assessing data architecture, selecting a suitable pilot workflow and establishing governance frameworks before scaling AI deployment, the report said.
“The future of intelligent treasury operations is built one governed workflow at a time,” EY said. (ANI)