Banks use AI to optimize cash management at ATMs

Cash has not disappeared the way many predicted.

According to the U.S. Federal Reserve’s “2026 Diary of Consumer Payment Choice” report released on August 4, cash accounts for 14% of consumer payments in the United States, more than 80% of consumers used it in the past 30 days, and 90% expect to keep using it.

That persistence leaves banks locked in an old tradeoff. Either they overstock ATMs with cash and leave capital sitting idle, or they understock them and risk an outage, an extra armored car run, or frustrated customers at empty machines.

AI narrows the gap between supply and demand

Artificial intelligence is starting to narrow the tradeoff by treating it as a forecasting problem rather than a guessing game. H2O.ai, an enterprise AI software company, builds cash-demand models for individual ATMs using historical withdrawal patterns, paydays, holidays, and regional seasonal trends, achieving forecast accuracy within roughly 15% on average, according to the company’s website.

This approach is fundamentally different from traditional methods. Instead of applying the same formula to all ATMs, AI analyzes each machine as an independent entity with its own characteristics. An ATM in a commercial district will have a completely different pattern than one in a residential area or near a school.

Operational cost savings

The financial impact of optimizing cash at ATMs is significant. According to industry estimates, each armored car dispatch to replenish or collect cash costs between $500-$1,500, including fuel costs, security personnel, and machine downtime. With networks of thousands of ATMs, banks can save millions of dollars annually simply by reducing unnecessary dispatches.

Additionally, idle cash sitting in ATMs also has an opportunity cost. If a bank has 10,000 ATMs with an average of $50,000 in excess cash per machine, that’s a total of $500 million not generating returns. AI helps reduce this number to optimal levels, freeing up capital for other profitable activities.

Lessons for the Vietnam market

In Vietnam, where ATM networks remain very common with over 18,000 machines nationwide according to State Bank of Vietnam data, applying AI to optimize cash management could yield similar benefits. Major banks like Vietcombank, BIDV, or Agribank all have networks of thousands of machines, and annual operational costs are substantial.

AI cash demand forecasting models can be adjusted to fit the specific characteristics of the Vietnamese market, where factors like Tet holidays, tourist seasons, or local events have a significant impact on cash withdrawal needs. For example, before the Lunar New Year, cash demand typically surges due to lucky money traditions and shopping.

Impact on Japan

In Japan, a country with a strong cash culture and dense ATM networks, AI cash optimization technology has particular significance. Japanese banks like Mitsubishi UFJ, Sumitomo Mitsui, and Mizuho have been investing in technology to modernize ATM operations.

What’s special in Japan is the diversity of ATM types and services. Many ATMs don’t just serve cash withdrawals but also provide other complex financial services. AI can help optimize not just cash but also accompanying services, creating a better customer experience.

The future of ATMs in the digital age

Although electronic payments are growing, ATMs still play an important role in the financial system. The combination of AI and big data is transforming ATMs from passive cash dispensing machines into intelligent points capable of prediction and self-adjustment.

In the future, we may see ATMs not only forecasting cash demand but also integrating with other financial services, becoming part of a smarter financial ecosystem.

---

Source: Banks Turn to AI to Stop Overstocking ATMs — PYMNTS, August 28, 2026