Perfect Information, Suboptimal Decisions: Agentic AI Reaches the Bank Treasury Desk
Forecasting tells a treasury team where cash will be. It does not decide where it should go.
On October 12, FIS launched the Agentic Insights Suite, an AI layer for the office of the CFO. It pulls fragmented data from ERP, CRM, and treasury systems, joins it with outside sources such as credit bureaus, and produces cash flow forecasts and risk scores that a finance team can query in plain language.
It is a serious product, and it signals where treasury is heading. It also raises a question every bank and credit union treasurer should be ready to answer: once the forecast is perfect, who decides what to do with the cash, and how do they prove it?
The numbers behind the headline
The launch rests on an adoption gap. FIS cites a Deloitte survey in which 46% of global finance leaders name working capital optimization as a primary use case for agentic AI, yet only 14% have integrated agents into their operations.
What the suite does is worth stating plainly:
Multi-horizon forecasting. Instead of one estimate, machine learning runs many simulations, testing how faster outflows or delayed inflows change liquidity.
Account scoring. Teams see where risk sits instead of reconciling conflicting numbers by hand.
Fast deployment. A self-serve setup that FIS says is typically live within weeks, with natural-language queries and Power BI dashboards on top of its Integrity and Quantum treasury products.
Next on the roadmap. Payment anomaly detection and dispute forecasting.
Notice the verbs: unify, forecast, score, flag. Every one of them is about knowing. None of them is about placing a dollar.
What a forecast solves, and what it leaves open
A good forecast tells a treasurer that surplus liquidity will arrive in three weeks, or that a shortfall is likely in the second half of the quarter. That is real value. But it leaves the next set of questions untouched:
Which counterparties will pay the most for that cash today, not six months ago?
For what term, and within which concentration and credit limits?
Who approved it, and where is the record?
Those questions carry more weight at a bank or credit union than almost anywhere else, for three reasons.
Margins leave little room for a stale rate. The FDIC's second-quarter profile put industry net interest margin at 3.32%, up just one basis point on the quarter. S&P Global Market Intelligence reports that 87.2% of certificates of deposit mature within 12 months and 42.4% within three, which means funding costs reprice quickly while the placement side of the balance sheet often does not.
Examiners ask for the decision, not the forecast. NCUA's 2026 supervisory priorities keep liquidity risk and interest rate risk at the top of the list and tell examiners to test whether earnings can support capital under interest rate, credit, and liquidity stress scenarios. A forecast shows foresight. A documented placement shows control.
Manual placement is still the norm. Rate checks by phone and email, a spreadsheet of limits, a calendar reminder for maturities. Legacy systems give perfect information about balances. The decision that follows is still made by hand, which is how perfect information produces suboptimal decisions.
The governable-AI test
J.P. Morgan's payments team made a related point in its June whitepaper, "Proof of Movement," covered by Treasury Today in August. Risk, liquidity and exposure data scattered across ERPs, treasury systems and bank portals limits what agents can responsibly do. Treasury, it argues, is becoming a control system that senses, predicts, decides, executes and audits, and the winners will be the institutions with the most governable AI, not simply the most AI.
That is a useful test for any treasury technology. Four questions:
Can it see every account? Forecasting tools increasingly can.
Can it compare every counterparty? Not by phone call, and not once a quarter.
Can it place the cash inside policy limits? Approvals, concentration caps and credit limits enforced in the workflow, not in a spreadsheet.
Can it prove what it did? A full audit trail, ready before the examiner asks.
Most stacks answer yes to the first and stop there.
The takeaway
The agentic wave will make forecasting cheaper, faster and better. That raises the value of the step that follows it. If every institution can see next month's surplus, the advantage moves to the one that places it at the best risk-adjusted rate, within policy, with evidence.
That is the gap Likwidity automates: See it, Compare it, Place it, Prove it, in a single workflow. Compare rates across 10+ banks in one click, place deposits within limits, and keep the audit trail as a by-product. No brokers, zero commission, pure SaaS, and it sits alongside the ERP, TMS and forecasting tools you already run.
The bottom line
Agentic AI is solving the knowing problem. Banks and credit unions still have a doing problem, and examiners will ask about the doing. The treasury teams that pair better forecasts with an automated, governed placement process will not just see the market sooner. They will act on it sooner, and be able to show how.
Ready to see where your institution's cash sits against today's market? Schedule a demo with Likwidity
Likwidity. Intelligent Liquidity, Automated.
Comments