United StatesTechnology and telecom2022 – ongoing

Diebold Nixdorf

Machine learning to predict and diagnose ATM failures

Diebold Nixdorf said predictive ATM servicing raised uptime by 1.3% to 3.5%

WorkingCompany's own claim

Problem

ATMs typically run at about 96% or 97% availability, and each failure can mean a technician visit.

Approach

Diebold Nixdorf built wireless sensors, cloud computing and machine learning into its DN Series ATMs. The system diagnoses problems, tracks patterns and each device's history to anticipate failures, and lets the company repair or upgrade remotely. When a technician is needed, it helps estimate the labor and time. Repairs can be scheduled when customer use is low.

Results

1.3% to 3.5%increase in ATM uptime on the earliest deployments, per Diebold NixdorfCompany's own claim[1]

about 200,000machines that had the technology for at least 18 monthsCompany's own claim[1]

Takeaway

Diebold reported the uptime gain from the roughly 200,000 earliest machines, which had the technology for at least 18 months. The article did not describe an independent measurement.

Sources

Is your ATM running? AI can catch it (if it's not).
American Banker17 Nov 2023News article

Source summaries

[1]American Banker17 Nov 2023

American Banker describes how Diebold Nixdorf uses sensors, cloud and machine learning to predict ATM failures and repair remotely. It reports that about 200,000 early machines saw uptime rise 1.3% to 3.5%, against typical availability of 96-97%. The figures are Diebold's.

Full details are in the original sources.

Added 2 Nov 2026 · Updated 2 Nov 2026