Fleet AI Is Spreading Across Daily Operations

APTA released a public-transit AI primer and four guidance briefs drawing on agency surveys and staff interviews. The material spans operations, maintenance, planning, safety, customer support, and ticketing.

The primer records a spread of live applications, including MTA bus predictive maintenance, AC Transit bus-lane image recognition, Riverside Transit disruption updates, and CapMetro’s paratransit virtual agent. APTA says customer support and analytics are the most common current deployments, while operations and back-office uses are prominent among planned applications.

The publication is an implementation guide and cross-agency snapshot, not an independent audit of every reported outcome. Its practical value is the combination of real examples with procurement, governance, infrastructure, and staff-readiness guidance.

Linxup announced an AI chatbot for managers using its GPS tracking and safety products, with small and midsize businesses as the target market. The company’s examples span HVAC, plumbing, landscaping, construction, and towing fleets.

Managers can ask questions about which vehicles started after a given time or which driver-coaching metrics need attention. Linxup says the assistant can flag idle-time and safety-score anomalies and calculate savings scenarios from a user’s fuel-price and idling assumptions.

The release includes a customer testimonial from Accurate Towing but no quantified independent performance study. The launch moves report queries into a conversational interface, while the quality of its recommendations depends on the underlying telematics history and the user’s assumptions.

CARNET-UPC announced the European-funded CAISE project to connect shared-mobility fleet information with public-transport conditions. The 12-month effort is planned for Turin, Berlin, and Munich.

The proposed agentic platform combines operator fleet data with service disruptions, weather, and local events. It is intended to recommend service-area plans before deployment and fleet rebalancing or disruption responses during operations.

CAISE is at project-launch stage, so the source gives targets and workflows rather than measured results. CARNET-UPC leads coordination and pilot evaluation across the three cities.

The 2026 School Bus Fleet district survey gathered responses from more than 80 U.S. public school transportation officials during July through September. Its findings put artificial intelligence alongside persistent operating pressures such as driver availability and vehicle purchasing.

Among respondents, 20% reported using predictive AI, 21% generative AI or large language models, and 14% both; another 28% were considering AI, while 17% said they were not interested. The survey is a snapshot of reported adoption and interest, not an independently measured record of system performance.

For district leaders, the result suggests that AI planning is beginning to compete for attention with staffing and fleet renewal. The respondent group leaned toward larger districts, so smaller operators should not assume the same readiness or investment capacity.

At the American Trucking Associations’ Management Conference & Exhibition, Samsara described growth in usage of Agent Studio, a tool for creating AI agents through plain-language instructions. The company said the most recent 500,000 prompts arrived in 16 days, compared with 112 days for the first 500,000, and about one-third of participating customers had deployed at least two agents.

The examples extend across safety, hours-of-service, driver briefings, reporting, and incident response. Jordan Carriers, which operates about 1,000 drivers, uses an agent to coach lower-risk behaviors and HOS agents to help identify drivers nearing their limits and locate safe parking; Samsara reported that the carrier saw a 30% crash reduction.

These are company-reported adoption and customer results, not a controlled study. The new customer examples and usage figures add evidence beyond the earlier product launch, while leaving the comparison periods and independent validation unspecified.

Innovate UK Business Connect profiled OpenCapacity’s work with Transport for London on public-transport capacity management. The project addresses demand swings that manual surveys and spot checks can miss.

OpenCapacity combines vehicle sensors, passenger-count information, CCTV, weather, events, and seasonal patterns to estimate occupancy and forecast where capacity will be needed. The system can also identify wheelchair and buggy occupancy and share wheelchair-space availability with passengers and operators.

The case study describes a 14-week development effort and collaboration with TfL, not a measured network-wide outcome. Its planning contribution is to make passenger loads and accessibility constraints visible at a finer operational level.

Waste Dive reported that Waste Connections is pursuing seven AI projects through 2027, with roughly $100 million of planned investment. One workstream targets routing across about 14,000 trucks.

The route initiative is described as combining the company’s software with Google, Waze, weather, traffic, and street-closure information. Management framed the portfolio as an investment intended to improve operating margin, with an initial margin target around $100 million in 2028–29.

The figures are management targets, not realized savings. The fleet decision is how to stage large-scale routing changes while measuring service, mileage, labor, and route-completion effects across different operating areas.

Floating Fleet AI introduced an agentic scheduling workflow for business aviation operators. Its system generates and compares schedule options while applying operational constraints, with the operator retaining the decision and visibility into why schedules differ.

In a representative three-day scenario involving 40 aircraft and 116 trip demands, the company said its workflow changed 51 scheduling decisions and projected approximately $86,000, or 5.78%, in cost savings. Those figures are scenario projections, not measured savings from an identified customer deployment.

The demonstration frames schedule planning as a capacity-allocation problem: dispatchers can compare alternatives across aircraft availability and trip requirements before committing. The value will depend on how well the optimization reflects real crew, maintenance, airport, and customer constraints.

Port of Felixstowe confirmed a third order from Westwell, expanding its autonomous truck fleet to 100 units in live container-terminal operations. The latest vehicles will use 128-line LiDAR and upgraded camera systems.

The port links the trucks to a private 5G network and terminal control systems. Its automated battery-swap station changes a depleted pack in five to six minutes, and the new investment includes a second station.

The release documents an expansion decision, not a new measured safety or throughput result. The order illustrates how an autonomous fleet acquisition depends on perception hardware, connectivity, charging logistics, and integration with conventional terminal equipment.

SANY said it shipped the first batch of SKT110Ei pure-electric autonomous mining trucks to a customer in South America on August 12. The project includes trucks, roadside infrastructure, cloud dispatch, and local operations and maintenance support.

SANY describes a system combining multi-sensor perception, AI-based decisions, and cloud-based dispatch. The trucks are intended for mine-site haulage, with remote takeover and autonomous operation available alongside on-site control modes.

The shipment marks the company’s first autonomous mining deployment in Latin America; the release does not name the customer or report operating outcomes for this site. SANY separately reported more than 300 autonomous mining trucks deployed globally by July 2026.

The San Mateo County Transit District board approved a contract with Samsara to equip 235 buses with dual-facing AI safety cameras and cloud services. The decision extends prior agency experience with AI cameras on contracted bus services.

Road-facing and driver-facing cameras identify risky conditions, issue alerts, and preserve video for incident review and operator coaching. SamTrans says the system is intended to improve pedestrian awareness as well as support collision prevention.

The board action is an approved deployment, not a reported collision-reduction result. Installation and training will determine how alerts enter safety investigations and how operators can review or contest events.

Zurich Airport says two automated shuttle buses have begun operating on airport grounds without a safety driver onboard. The employee shuttles run on a defined route that does not cross aircraft taxiways.

WeRide supplies the automated-driving technology, while Swissport and Krummen Kerzers employees monitor the vehicles from a remote cockpit and can intervene. The former onboard safety drivers were trained for remote supervision, and the shuttles had completed more than 15,000 kilometers during the project.

The airport plans to operate the pilot through the end of 2026 while it reviews whether to continue. Its phased rollout began with empty-vehicle tests before staff were permitted to ride.

MOIA America and Beep began invited passenger rides in self-driving ID. Buzz vehicles in Lake Nona, Florida, marking MOIA America’s first U.S. passenger operations. Riders request trips in the Beep app and board at designated pickup locations on three planned routes.

The turnkey package combines Volkswagen ID. Buzz vehicles, Mobileye’s driving system, mobility-service software, and operator training. Trained human operators remain onboard during this phase while passenger feedback informs testing.

This is a passenger test stage rather than an open public service or unsupervised deployment. The operation makes onboarding and role definition part of the fleet system, including rider support and an onboard safety function.

Jeppesen and ForeFlight announced Airflow Agents, an aviation-focused agent platform intended to support commercial airline operations. Initial workflows include trip trading, crew pairing and rosters, and disruption recovery.

The companies describe a context layer built around aviation data and safety-management principles, with customers choosing how much automation to apply and human accountability retained. They are seeking co-development partners ahead of broader rollout; the announcement does not document a production airline deployment or measured crew outcome.

Crew planning and recovery decisions affect staffing coverage, schedule resilience, and the workload imposed on operations teams. The announcement therefore marks a product-development step, while the evidence needed to judge adoption and operational effect remains ahead.

Beacon AI announced that its Murdock pilot-assistant system flew aboard three operational U.S. Air Force KC-46A aircraft over five flights and about 20 flight hours during August and September. The flights included Emerald Flag at Eglin Air Force Base and two other airfields under an AFWERX Tactical Funding Increase with Air Mobility Command.

Murdock reads certified avionics-bus data through a modular cockpit interface and sends information to Beacon’s Lighthouse fleet platform over satellite communications. Beacon characterizes the system as advisory: it does not modify certified flight-critical systems, and the crew retains authority.

The flights represent operational-aircraft use rather than a dedicated testbed, but the release does not quantify safety or workload outcomes. The next stated step is progressive assistance across more of the flight lifecycle, informed by operational experience and crew feedback.

Karsan says its autonomous electric e-JEST began carrying passengers in Atlanta’s ATL Spoke pilot, a first-of-its-kind public-transit service for the city. The route links MARTA’s West End Station with the Atlanta Beltline Southwest Trail.

The vehicle operates with Karsan, autonomy technology partner ADASTEC, distributor Damera, and service orchestration partner Beep. The service is described as free passenger transport; the release does not publish ridership, on-time performance, or remote-intervention rates.

For transit planners, this is an operating passenger connection rather than a vehicle-only demonstration. The pilot gives partners a live setting to assess first- and last-mile access and autonomous service under actual urban conditions.

Changi Airport put two fully driverless tractors into live baggage-transfer work between Terminal 1 and Terminal 4. The deployment followed nearly a year of trials that covered more than 5,000 test trips.

Each tractor uses more than ten sensors and cameras to navigate airside conditions, with operations monitored from a control center where a remote operator can intervene. The airport has marked autonomous-vehicle zones and says six more tractors are planned for another baggage route, with a longer-term fleet target of 24 by 2027.

The system moves baggage tractors from trial operation into a live airport workflow, while staff shift toward last-mile handling and higher-skill roles. The airport’s release describes deployment and expansion plans but gives no quantified impact on turnaround times or incidents.

Logan Bus announced a Cisco-built AI phone system for families of students riding its New York City school routes. The operator transports about 30,000 general-education students each day.

The phone assistant links a student’s route to a parent’s caller ID and answers common bus-location, delay, and arrival questions using route information and GPS tracking. Callers can reach a live agent, and the company said a summer pilot handled about 60% of calls.

The rollout was planned for September, and the company said a later phase would add a direct Via GPS API connection for more precise answers. The disclosed pilot result concerns call handling, not route punctuality or student safety.

San José expanded an AI-enabled transit signal-priority system to all city bus routes in partnership with LYT. The mayor’s office reported that citywide bus speeds had increased 20%.

The system uses real-time traffic data to prioritize green signals as buses approach intersections. The expansion followed a 2023 two-route pilot that the city said cut red-light waiting by 50% and improved on-time performance.

The public release does not provide route-level speed comparisons or disclose how the 20% measure was calculated. Signal priority is an operating intervention whose effects can vary with corridor congestion, bus frequency, and intersection controls.

PlusAI announced a Level 4 autonomous trucking trial using a factory-integrated International truck in Ryder’s freight operation. The truck is scheduled to run a daily 600-mile route between Ryder locations in Laredo and Temple, Texas.

PlusAI’s SuperDrive virtual driver controls the truck during the dedicated corridor trial, while Ryder and International assess real-world performance, reliability, and operating requirements. The partners describe the effort as a customer freight operation rather than a closed test track.

The trial tests how autonomy fits into an existing carrier route and terminal model; it does not establish broad commercial readiness. The announcement is a deployment start, without route-level safety or cost outcomes yet.

The California DMV adopted updated autonomous-vehicle rules that allow manufacturers to apply for permits to test and deploy heavy-duty autonomous vehicles on state roadways. The rules also address oversight and enforcement for light- and heavy-duty systems.

Applicants must operate within permit, safety, and enforcement requirements; the regulatory framework establishes how truck and transit deployments can move from testing toward road use. The state release describes an authorization pathway, not approval for any specific fleet.

For carriers and transit authorities, a public-road trial now requires a documented permit pathway tied to the vehicle configuration and operating geography. That changes which compliance work must be complete before a deployment schedule is credible.

Brussels Airport began testing an autonomous Charlatte electric tow tractor at Brucargo with WFS Cargo, Charlatte Manutention, and Navya Mobility. The trial studies cargo-trailer movements between warehouses and aircraft aprons.

The vehicle uses satellite navigation and laser sensors on four predefined routes, capped at 12 km/h. A trained operator remains onboard to supervise the tractor and intervene in the airport’s mixed-traffic environment.

The project is a constrained test, not an unattended operating service. Route limits and onboard supervision keep the pilot within a manageable safety envelope while the airport studies cargo efficiency and emissions.

UK glassmaker Encirc selected Powerfleet’s Unity platform to support safety across its manufacturing enterprise. The announced scope includes forklift operations and on-site interactions among operators, vehicles, pedestrians, and production areas.

Powerfleet describes combining vehicle, operator, and safety data with AI-enabled video and predictive safety capabilities. The announcement identifies the intended workflow but gives no quantified safety outcome or detailed deployment timetable.

For a glass plant, the operational question is whether the system can improve visibility of vehicle-pedestrian risks without disrupting material movement. Encirc’s selection is a customer decision; the resulting change in incidents or near misses has not yet been reported.

Fullbay announced its acquisition of Pitstop, an AI predictive-maintenance platform, with plans to integrate the technology into its heavy-duty repair and fleet software. Fullbay says its shop platform serves more than 5,000 repair facilities.

The planned combined system will use repair histories, diagnostics, fault codes, inspections, and service orders to flag emerging issues, create service requests, and forecast parts demand. Fullbay cites more than ten years of repair data and $6.5 billion in annual service orders and parts across its network.

The transaction and integration plan are announced; the source does not document a completed fleet deployment of the combined capability. The proposed workflow links vehicle warnings to shop capacity and parts availability before a truck reaches a breakdown state.

Jurni Leasing is integrating Targa Telematics’ OEM vehicle data into the Jurni 360 fleet platform. The arrangement draws on Targa’s network of more than 45 OEM data partnerships and is intended to support service, maintenance, repair, warranty, and mileage workflows for Jurni’s fleet customers.

The planned AI analysis is described as finding patterns and prioritizing maintenance needs from vehicle data, helping teams plan servicing and monitor warranty compliance. The announcement does not specify a model design, customer-level deployment stage, or measured reduction in downtime or maintenance cost.

For leasing operators, OEM-originated data may reduce reliance on separate aftermarket devices when it is available and consistent across vehicle makes. Mixed fleets and differing data permissions will affect how uniformly the workflow can be applied.

Freeway Fleet Systems announced that it would demonstrate “Ask Freeway,” an AI maintenance assistant, at Euro Bus Expo on November 3–5. The planned tool lets users query connected fleet records in plain language.

The described records include engineering information, defects, labor, parts, inspections, servicing, and compliance. The assistant is intended to draw on information captured at the point of work; the announcement does not establish that named bus operators are using this specific assistant or provide measured results.

A natural-language interface could help workshop and engineering staff locate relevant records without knowing where each field is stored. The quality of answers will depend on the completeness and consistency of the underlying maintenance entries.

Caterpillar reported that Luck Stone’s Bull Run Plant in Virginia had moved two million tons using Cat 777 autonomous trucks during the system’s first year. The quarry deployment, integrated into site operations in November 2024, is the first Cat autonomous-hauling application in a quarry.

The MineStar Command system coordinates autonomous haul trucks with quarry production and site-control processes. Caterpillar said the operation reached one million tons in July 2025 and recorded no injuries during the reported period.

The milestone is a vendor-reported production result, not an independently audited safety study. It demonstrates how a fixed quarry haul cycle can support repeatable autonomous utilization while keeping site-specific risk controls central.

Motive published four customer examples describing fleet technology transitions and results. One oilfield services operator, EnerCorp, compared the same seven trucks on two platforms over sequential three-month periods before its broader rollout; the company later said it had 25% fewer cellphone events, 10% less distracted driving, and 19% fewer stop-sign violations within two months.

Other examples include a 125-vehicle line-haul provider reporting fewer DOT accidents and preventable incidents after adopting AI-tagged video and coaching, and Reliable Carriers reporting 99.9% damage-free and 100% on-time delivery over a year. These figures come from Motive’s vendor case material, with differing baselines and methods across customers.

The examples show operators using camera-based detection and coaching to compare driver behavior and operating measures after a platform change. They do not provide a common controlled methodology, so the reported percentages should be treated as customer-reported results rather than an independent comparison of vendors.

Keolis reported that SOH Fleet is deployed on its Blois AZALYS network and tested in Barcelona with TMB. Six Blois buses and ten Barcelona buses have been equipped for the pilot.

Onboard sensors collect battery operating data that ACTIA’s AI analyzes to estimate state of health and detect degradation. Threshold alerts can prompt charging changes, vehicle rotation, targeted maintenance, or adjustments to the transport plan.

Keolis reported early pilot indicators of 20% longer battery life and 20% lower vehicle maintenance cost, but the source does not provide the comparison design or period behind those figures. Keeping a bus within its full-day range is the immediate maintenance and service constraint.

EIT Urban Mobility announced ten new strategic innovation projects, including BatterySense, which involves Transdev and partners in France. Transdev has about 16,000 buses and coaches in France, with more than 60,000 auxiliary starter batteries in its worldwide fleet.

BatterySense combines used-battery diagnostics and industrial regeneration with AI prognostic models that estimate remaining useful life from sensor and operating data. The project plans pilots in Le Havre, Rouen, and Valence, then a documented go/no-go decision for broader use.

The project is funded and moving into pilot work; its targeted savings and life extension are not yet measured outcomes. The fleet decision is whether measured battery condition can replace a fixed annual swap cycle without increasing non-start failures or service disruption.