Getting stuck in a broken lift is annoying for a resident and expensive for a building owner. That is why the world’s biggest elevator companies now use software to catch faults early. This is the story of how AI predicts elevator breakdowns before a car ever grinds to a halt — and why it matters for anyone who lives, works, or invests in tall buildings.
Modern elevators are packed with sensors. AI reads that flood of data, learns what “healthy” looks like, and flags the tiny changes that signal a part is wearing out. Instead of waiting for a failure, a technician gets a warning weeks ahead and fixes the problem during quiet hours.
How does AI predict elevator breakdowns?
AI predicts elevator breakdowns by continuously analyzing sensor data — door cycle times, motor current, vibration, brake wear, and ride quality — and comparing it against millions of normal operating cycles. When a pattern drifts toward a known failure signature, the system alerts maintenance teams days or weeks before the elevator actually stops working.
The core idea is called predictive maintenance. Rather than servicing an elevator on a fixed calendar (every three months, whether it needs it or not) or repairing it only after it breaks, predictive maintenance uses live data to service each machine exactly when the data says it needs attention.
An elevator generates a surprising amount of information. Every trip involves a motor accelerating and braking, doors opening and closing, ropes or belts under tension, and a controller managing it all. Sensors capture these events, and machine learning models look for the small anomalies humans would never notice — a door that closes 40 milliseconds slower than last month, or a motor drawing slightly more current on the same load.
The typical AI elevator monitoring workflow
- Sensors collect data from the motor, doors, brakes, and car in real time.
- Data streams to the cloud over an IoT connection, often thousands of readings per day per elevator.
- Machine learning models score the data, comparing it against healthy baselines and known fault patterns.
- The system flags at-risk components and estimates how much useful life a part has left.
- A technician is dispatched with the right spare part before the elevator fails.
How is AI used in elevator predictive maintenance today?
AI elevator predictive maintenance is already live at global scale. The three largest manufacturers — Otis, KONE, and TK Elevator — each run cloud platforms that monitor connected elevators around the clock, detect faults remotely, and dispatch technicians before breakdowns occur.
This is not a lab experiment. These systems watch millions of real elevators in offices, hospitals, airports, and apartment towers every single day.
TK Elevator MAX
TK Elevator (formerly thyssenkrupp Elevator) launched MAX in 2015 with Microsoft, describing it as the elevator industry’s first real-time, cloud-based predictive maintenance service. Sensors feed data into the Microsoft Azure cloud, where algorithms calculate the remaining lifetime of key components and flag which parts will need attention and when. According to TK Elevator, MAX has since been extended to tens of thousands of customers worldwide.
KONE 24/7 Connected Services
KONE launched 24/7 Connected Services in 2017, built on IBM Watson IoT technology. The platform monitors more than 200 safety-related parameters and analyzes the data in real time to predict maintenance needs. KONE services well over a million elevators and escalators globally, giving its models an enormous training base. In 2025, KONE also began deploying a generative AI “Technician Assistant” on Amazon Bedrock to help field engineers diagnose problems faster.
Otis ONE
Otis runs Otis ONE, an IoT platform that provides remote monitoring, predictive diagnostics, and smarter service planning. When a connected unit reports abnormal behavior, Otis ONE can alert a mechanic and share diagnostic detail before they arrive on site, cutting the guesswork out of a repair visit.
What faults can AI catch before an elevator breaks?
AI is especially good at catching gradual wear — the kind that builds slowly and gives off subtle signals long before a total failure. Door systems, motors, and brakes are the most common targets because they move constantly and are the leading causes of service calls.
The single biggest culprit in elevator downtime is the door. Door operators open and close hundreds of times a day, and worn door motors, fouled tracks, or degrading locks account for a large share of entrapments. Because AI can track door cycle times so precisely, it often spots door trouble two to six weeks before it would cause a breakdown.
Common issues AI monitoring can flag early include:
- Door faults — slowing motors, misaligned tracks, and worn locks.
- Motor and drive wear — rising current draw or overheating under normal load.
- Brake degradation — changes in stopping distance or response time.
- Ride quality problems — increased vibration or jerky leveling at floors.
- Rope or belt wear — subtle tension and movement changes over time.
The goal of predictive maintenance is simple: replace a worn part on a scheduled Tuesday morning instead of during a Friday-night entrapment.
This same predictive-maintenance approach shows up across many industries. We have looked at how it works for power grids in how AI helps utilities predict transformer failures and for telecom networks in how AI predicts cell tower outages before they happen. Elevators are simply another asset where catching failure early saves money and prevents danger.
Why does AI elevator maintenance matter?
Beyond convenience, predictive maintenance affects safety, cost, and accessibility. A dead elevator in a hospital or a high-rise apartment is not a minor inconvenience — it can trap people, strand wheelchair users, and delay emergency response.
The business case is strong too. Industry reports and manufacturer case studies point to meaningful gains: fewer unplanned callouts, higher elevator uptime, and better first-time fix rates because the technician already knows what is wrong and brings the correct part. Advanced IoT analytics have helped KONE report significantly more accurate fault detection and fewer equipment issues across its connected fleet.
For building owners, higher uptime also protects reputation and tenant satisfaction. For manufacturers, connected services create a recurring digital revenue stream on top of hardware sales — which is exactly why all the big players have invested so heavily in AI.
What are the limits of AI in elevator maintenance?
AI is not magic. It predicts wear patterns well, but it cannot foresee everything — a vandalized panel, a sudden power surge, or flood damage can still cause an unexpected failure. Prediction reduces breakdowns; it does not eliminate them.
There are also practical hurdles. Older elevators may lack the sensors needed to feed the models, so buildings must retrofit hardware before AI monitoring works. Data quality matters enormously: a model trained on one elevator type may misjudge another. And a human technician is still essential — AI flags the problem, but a person does the physical repair. The most reliable setups treat AI as a co-pilot that guides skilled mechanics, not a replacement for them.
Frequently Asked Questions
Can AI completely prevent elevator breakdowns?
No. AI predictive maintenance sharply reduces unexpected breakdowns by catching gradual wear early, but it cannot prevent every failure. Sudden events like power surges, vandalism, or water damage can still stop an elevator. The goal is fewer, more predictable repairs — not zero repairs.
Do all elevators support AI predictive maintenance?
Not automatically. Newer elevators often ship with the sensors and connectivity needed for AI monitoring. Older units usually require a retrofit — adding sensors and an IoT gateway — before a platform like Otis ONE, KONE 24/7 Connected Services, or TK Elevator MAX can monitor them.
How far in advance can AI detect an elevator fault?
It depends on the component. Slow-developing issues such as door-system wear can often be flagged two to six weeks ahead. Faster-moving problems give shorter warning. The lead time is usually enough to schedule a repair during off-peak hours rather than reacting to an entrapment.
Which companies lead in AI elevator maintenance?
Otis (Otis ONE), KONE (24/7 Connected Services, built on IBM Watson IoT), and TK Elevator (MAX, built on Microsoft Azure) are the three biggest players. Each connects large fleets to cloud platforms that use machine learning to predict and prevent failures.
Is my personal data collected when AI monitors an elevator?
Generally no. Elevator monitoring systems track mechanical and performance data — door timings, motor current, vibration, and error codes — not the identities of passengers. The AI cares about how the machine behaves, not who is riding it.
Sources: TK Elevator (MAX launch) and Computer Weekly (KONE + IBM Watson).
