What problem does it solve?
Fleet operators have long relied on a driver or an agent walking around a vehicle. Inspections are done in varying light and weather and miss small defects such as a sidewall tear or a nail in a tread. Rental companies also need a record of the condition before and after each rental to decide who pays for damage.
Amazon says its fleet managers previously had to rely solely on the human eye and manual inspections for daily safety rounds, and Hertz says rental assessments have always relied heavily on manual inspections in varying conditions.
How does it work?
- Drive through the gate. The vehicle rolls through an archway, over sensor plates at low speed (about 5 mph in Amazon's archways), at pickup, return or the end of a work shift.
- Capture images. Cameras and sensors capture the body, glass, tires and undercarriage, often in 3D.
- Detect and classify. A vision model finds dents, scratches, tire wear and mechanical defects and classifies them by severity.
- Compare with the baseline. For rentals the scan at return is compared with the scan at pickup, so only new damage is flagged.
- Route the result. Results go to a fleet portal or the rental system with photos and repair suggestions. A person decides on repairs and on any charge to a customer.
- Audience
- Employee facing
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Internal tools
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Accuracy | Too few to pool | 96% | 1 | 1 vendor |
Value drivers: Lower cost to serve, Risk and loss reduction, Speed and cycle time.
Who already uses it?
3 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Hertz
United States · Logistics and transportation · 2025
Hertz announced on 16 April 2025 that it is adding UVeye's AI inspection systems, which scan a vehicle's body, glass, tires and undercarriage in seconds, to its US operations, starting at Atlanta Hartsfield-Jackson airport and rolling out to major airports. The release gives no result figures. The rollout drew customer disputes over damage charges, reported by The Drive.
No outcome disclosed.
Sixt
Germany · Logistics and transportation · 2025
Sixt scans rental cars with its Car Gate system when a customer drives away and again on return, and says it issues damage claims only after a staff member has reviewed what the computer flagged. The Drive reported in July 2025 that two US customers were billed for damage that was already present at pickup, and Sixt apologized and said discrepancies are reviewed and resolved. No result figures are published.
No outcome disclosed.
Amazon
United States · Logistics and transportation · 2023
Amazon and UVeye built an Automated Vehicle Inspection archway that delivery drivers pass through at the end of each workday, replacing manual driver vehicle inspection reports for vans operated by Delivery Service Partners in the US, Canada, Germany and the UK. The scan takes seconds, sorts issues by severity and shows photos and repair suggestions in Amazon's Fleet Portal. UVeye reports 96% defect detection against 24% for manual inspection.
- Accuracy: 96%
"AI-driven accuracy: Identifies defects with 96% accuracy vs. 24% manually, spotting even the smallest anomalies in Amazon delivery vans."
Claimed by: vendor
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Vehicle identification and contract or driver assignment at scan time
- Baseline scans from the previous checkout or inspection
- Repair standards that define what counts as damage or a defect
- Photos and scan history kept for the retention period
Systems to integrate
- Scanner vendor platform and cloud storage
- Fleet management or rental system
- Maintenance and work order system
- Customer billing and dispute handling
Complexity: Medium
The scanner is physical equipment that has to be installed at each site and tuned to the fleet, for example Amazon trained the models on its own roadworthy standards and van types. The scan has to be linked to the vehicle record and the rental contract, and the process around disputes has to be ready before any charge is made.
- 1
Start at a single site
Install at one airport branch or delivery station and run it beside the manual inspection until you have compared both on the same vehicles.
- 2
Define damage rules before billing
Decide thresholds for chargeable damage and the wear that is normal, and write them down for customers and staff.
- 3
Put a person before any charge
Have a trained staff member review every flagged item before a customer is charged. Sixt says it does this, yet The Drive reported two 2025 Sixt invoices that used the pickup scan photo as proof of return damage. The review must check the date and scan of every photo, not only the flag.
- 4
Feed repairs from the results
Link detected defects such as tire wear to the work order system so the scan becomes maintenance, as Amazon does through its fleet portal.
- 5
Handle disputes with the evidence
Give customers the dated photos from pickup and return on request and track dispute outcomes.
Guardrails
- Human review of every flagged item before any charge to a customer
- Dated before and after images available to the customer on request
- Baseline scan must exist before a damage claim is raised
- Retention and access rules for vehicle images and any visible personal data
KPIs to instrument
- Defects found per scan against manual inspection on the same vehicles
- Share of flagged items overturned on review or dispute
- Inspection time per vehicle
- Repairs triggered before a breakdown
Human in the loop
Staff review flagged damage, decide on repairs and approve customer charges. Operations leads review dispute outcomes and adjust the rules and thresholds.
Common failure modes
- Wrong photo as evidence
- A damage claim uses a photo from the wrong date or scan. Match each claim to the pickup and return scans by timestamp.
- Automatic charging
- Customers are billed for flagged damage without review. Require staff review before a charge.
- Site variance
- Weather and light at the gate change detection quality. Test at each site and track false positive rates.
- Customer backlash
- Charges that look automated damage trust and reach the press. Publish the process and handle disputes quickly.
What are the risks and rules?
EU AI Act
Depends on design
Scanning vehicles for damage and defects is not an Annex III use, and Article 50 does not apply because no person interacts with the system, so it is normally minimal risk. If scan results are used to attribute damage to individual drivers and to monitor, evaluate or discipline them, as could happen in a delivery fleet, Annex III point 4(b) can make it high risk, and under Article 26(7) employers must inform workers' representatives and the affected workers first. Charges to rental customers remain a consumer law matter.
Controls to put in place
- Human review before every customer charge
- Dated baseline and return images retained and shared on request
- Written damage thresholds and a dispute process
- Image retention and access limits
When it went wrong elsewhere
- Sixt customers charged for damage present before their rental. The Drive reported in July 2025 that two Sixt customers were invoiced for damage that predated their rental, in both cases with photos from the pickup scan. Sixt apologized in one case, said discrepancies are reviewed, and dropped both claims.
Frequently asked questions
- How accurate is AI vehicle inspection?
UVeye says its system identifies defects in Amazon delivery vans with 96% accuracy, against 24% manually. This is a vendor claim without a stated method, so run both on the same vehicles before trusting it.
- Do rental companies charge customers automatically?
Practice differs. Sixt says a staff member reviews flagged damage before a claim, which The Drive contrasted with Hertz, and The Drive still reported two Sixt invoices for damage that predated the rental. Enterprise told The Drive it does not use digital damage scanners and Avis said its damage assessment remains human led.
- What do fleets learn beyond damage?
Amazon reports that 35% of all issues found by its scanners involve tires, and Delivery Service Partners are notified to replace tires before a blowout or flat.
How to cite this page
Blits.ai AI Use Case Library, "AI drive through vehicle inspection for rental and delivery fleets", last verified 11 November 2026, https://www.blits.ai/ai-use-cases/drive-through-vehicle-inspection. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 11 November 2026: First published