How does predictive maintenance work with the telematics a fleet already runs?
AI that detects and diagnoses failures early works from two feeds a fleet already has: telematics data from providers such as Samsara, Geotab, Motive or Verizon Connect, and work orders from its maintenance system. A read-only connection brings both in. Nothing goes on the vehicle, and each breakdown alert returns as a work order in the service shop’s existing queue.
Telematics Shows the Symptom and Work Orders Show the History
Telematics logs what a vehicle is doing all day: fault codes, regeneration counts, temperatures, pressures and idle time. It does not know what was repaired last month or whether the same part has failed before. The maintenance system holds that, in work orders written by the technicians who did the job.
Read together, the two answer questions neither can answer alone. A coolant temperature that climbs a little each week means one thing on a vehicle with a new thermostat and another on a vehicle with no cooling work in two years.
Matching the two feeds across every vehicle, every week, is work for AI software built for fleet maintenance. The models learn what normal looks like for each vehicle on its own routes and loads, flag the vehicles that start to drift from it, and rank them by how soon they need attention. The work orders also keep the models honest. A vehicle whose EGR cooler was replaced three weeks ago has a different baseline than the same vehicle before the repair, so a fixed problem does not keep raising alerts. The technician still decides what to repair.
Setup Asks for Read-Only Access and No Hardware
The fleet gives read-only access to its telematics feed and its maintenance history, in whatever format the history is kept. The connections run through APIs to telematics providers such as Samsara, Geotab, Motive and Verizon Connect, and to maintenance platforms such as RTA Fleet, Whip Around and Fleetio. No personally identifiable information is needed, and the setup is done remotely, so the service shop does not stop for it. The telematics account itself stays as it is, and dispatch and safety work carry on as before.
Abel Cota, Director of Information Systems at Troiano Waste Services, put it this way: “Tensor Planet’s system connected directly with the data we already had, without requiring any manual prep work or new hardware.” Troiano runs Samsara telematics and CMMS repair history, and cut exhaust-related repairs by 41% in 10 weeks across 55 vehicles, with no export work for its team (Troiano case study).
One condition applies. The approach works best on vehicles where the fleet controls maintenance decisions and keeps its own repair history, so vehicles serviced entirely by a lessor are a weaker fit.
Breakdown Alerts Land in the Queue the Service Shop Already Uses
Each breakdown alert names the vehicle and the component, and carries a recommendation: the likely cause and the check or repair to make. It arrives as a work order in the maintenance system the technicians already work from, lined up with the PM schedule so the corrective work can be bundled into planned shop time. There is no second login and nothing new to learn. Because the work order arrives two to three weeks before the vehicle would have failed, the shop can order parts first and book a bay, which a roadside call never allows.
Brakebush Transportation, Inc., a private fleet of about 110 tractors, runs Samsara telematics and Alpine maintenance software. In its first seven weeks it acted on 9 breakdown alerts and avoided 142 hours of downtime. Examples included a faulty NOx sensor on a brand-new tractor and an exact call on an EGR cooler. Repair cost and downtime avoided came to $26,517, with parts ordered ahead and repairs done in driver home time (Brakebush case study). Curtis Cutsforth, the Fleet Maintenance Support Supervisor, said: “We did not have to install any new hardware to get Tensor Planet set up. They were able to use everything that we already had in place.”
Questions fleets ask
Is it safe to connect predictive maintenance software to our telematics data?
Tensor Planet reads a fleet’s telematics feed and maintenance history through read-only access, and needs no personally identifiable information. Nothing is installed on the vehicles. If a fleet does not continue after a pilot or evaluation, its data can be deleted on request under the agreed retention terms.
Do you need a data science team to run predictive maintenance software?
Tensor Planet builds and runs the models itself, so a fleet needs no data science staff. It connects to the telematics and maintenance systems the fleet already has, detects and diagnoses problems early, and sends each recommendation to the technician as a work order in the existing system.