Banks have a surprisingly expensive problem. They need enough cash inside ATMs without leaving too much money sitting idle.
Artificial intelligence is making that calculation far more precise.
Banks and ATM operators are now using AI to forecast cash demand at individual machines. These systems study withdrawal patterns, paydays, holidays, local events, and seasonal changes. The goal is simple. Put the right amount of cash in the right ATM at the right time.
The potential savings are substantial. Brink’s says better forecasting can reduce total cash demand across an ATM network by 30% to 40%. AI is also being used for maintenance. Some systems can spot early signs of hardware trouble before an ATM fails.
Cash Still Matters More Than You Might Think
Digital payments keep growing, but cash has not disappeared. It still represents a meaningful share of consumer transactions in the United States. Many people continue to use cash regularly, and most expect to keep using it. That leaves banks with a difficult balancing act. Too little cash means empty ATMs and frustrated customers. Too much means valuable capital sits inside machines without being used elsewhere.
AI Turns ATM Cash Management Into a Forecasting Problem
Banks have traditionally relied on broad cash buffers to avoid shortages. AI allows them to be much more specific. Forecasting systems can study the transaction history of each ATM and estimate future withdrawals. They can also account for paydays, holidays, and seasonal changes. This gives banks a clearer idea of how much cash each machine actually needs. The result is less guesswork and fewer oversized cash loads.
Brink’s Says Smarter Forecasting Can Cut Cash Demand by Up to 40%
The savings become more noticeable across a large ATM network. Brink’s uses forecasting alongside cash supply planning, branch inventory, and order optimization. The company says this approach can reduce total cash demand by 30% to 40%. Its system can also forecast requirements several days in advance. That gives operators more time to prepare for shifts in demand. Instead of reacting to shortages, they can plan for them.
An ATM Near a Stadium Shouldn’t Be Treated Like Every Other Machine
Cash demand can change dramatically depending on location. An ATM near a stadium may see a rush before a major game. A machine in a quiet residential area may need far fewer refills. AI can detect those differences. It can then adjust cash levels based on expected local demand. This helps banks avoid using the same refill schedule for every machine. It can also reduce unnecessary cash deliveries.
Less Idle Cash Means More Capital Stays Productive
Cash inside an ATM is still part of a bank’s capital. However, that money is not available for other uses while it sits inside the machine. More accurate forecasting helps banks reduce those idle reserves. They can keep enough cash for customers without overloading every location. Across hundreds or thousands of ATMs, the savings can become significant. AI helps banks balance availability with better capital efficiency.
AI Is Moving Inside the ATM Too
AI is also being used at the machine level. Hyosung Americas uses technology that studies transaction history, nearby events, and seasonal patterns. The system can estimate the ideal cash balance for each location. This helps reduce excess cash and unnecessary refill trips. The approach can be especially useful for cash-recycling ATMs. Those machines reuse money deposited by customers, which makes cash flows more complex to predict.
AI Can Spot ATM Problems Before the Machine Breaks
AI is not only helping with cash levels. It is also being used for predictive maintenance. Hyosung uses sensor data to look for signs of mechanical trouble. Small changes in machine behavior can reveal a problem before a full breakdown happens. Technicians can then respond earlier. They can also arrive with a clearer idea of what needs to be repaired. That can reduce downtime and repeat service visits.
NCR Atleos Is Also Using AI Across ATM Operations
NCR Atleos is taking a similar approach. The company uses AI and automation across cash management and ATM operations. Its systems can model cash movement from central vaults to individual machines. NCR Atleos also supports predictive diagnostics and intelligent dispatch tools. These technologies can help improve machine availability. They can also make maintenance and cash planning more efficient.
ATM Cash Management Is Becoming an Optimization Problem
Research is pushing the idea even further. A 2026 study examined ATM replenishment through a QUBO-based optimization model. Researchers tested the framework using data from hundreds of ATMs in Italy. The model reduced costs while maintaining a very high service level. That suggests ATM replenishment could become even more automated. Future systems may combine AI forecasting with advanced optimization models.
AI Doesn’t Need to Reinvent Banking to Save Banks Money
This is not the most glamorous use of artificial intelligence in finance. There is no chatbot replacing a banker. There is no futuristic financial adviser making investment decisions.
Instead, AI is solving a narrow operational problem. How much cash should go inside a specific ATM tomorrow?
That question sounds simple, but it matters at scale. Better forecasts can reduce idle capital, cut refill trips, and prevent empty machines. Predictive maintenance can also reduce service costs and downtime.
The bigger lesson is familiar across enterprise AI. Some of the most valuable systems are not replacing entire workflows. They are simply turning an expensive guessing game into a better forecast.
Sources
- PYMNTS — Banks Turn to AI to Stop Overstocking ATMs
https://www.pymnts.com/news/artificial-intelligence/2026/banks-use-ai-stop-overstocking-atms/ - NCR Atleos — Cash Management and Optimization
https://www.ncratleos.com/banking/atm-itm/software/management/cash-management
