AI inspections for Samsara

Overview

A DVIR is a signed statement about a vehicle at one moment, not a record of how that vehicle is ageing. Samsara captures it properly: the driver walks the truck or trailer in the Samsara Driver App, marks each area pass or fail, picks the defects, writes a description, attaches a photo, sets the vehicle safe to drive or unsafe, and signs. What the report cannot do is compare the photo the driver just took with the photo of the same wheel end taken in March, or notice the thing in frame that the driver did not name. RapidEye reads those photos, every submission, and returns a finding per problem with the vehicle, the driver and the two photos attached to it.

How it works with Samsara

The inspection already has a shape, so the checks follow it rather than asking your drivers to do anything new.

• Pre-trip defect photos. A driver picks a defect, types or dictates a description and attaches a photo before the vehicle moves. RapidEye reads the photo against the same area on that asset's previous reports and comes back with whether this is new, whether it is worse, and how bad it looks next to the rest of your fleet.
• Post-trip defect photos. The end-of-day report is where damage from the shift shows up, and it is the report most likely to be filed fast in a dark yard. A bent mud flap bracket or a fresh gouge along a trailer skirt comes back as a finding with the shift bracketed by the two photos.
• Unresolved previous defects. The next driver has to look at what the last one reported. RapidEye keeps the photo from the original report beside the photo from the report that certified it fixed, which is the pair a warranty claim or a roadside dispute actually needs.
• The walkaround photos on a passing inspection. A clean DVIR with photos on it is still evidence. RapidEye reads the front, side and rear frames of an inspection where nothing was written up, and says when something in the frame deserves a work order.
• Trailers and yard equipment. Trailers get walked at hook and drop and then not again for weeks. Their DVIR photos, read in sequence, are the closest thing most fleets have to a condition history on a trailer.

Each finding names the asset, the area, the kind of problem, how serious it looks, the two photos that show it, and a suggested next step. It lands in the RapidEye dashboard and, if you want, a Slack channel or an inbox the shop watches. What becomes a work order in Samsara is your call, always. If your team documents condition some other way, a video walkaround at the gate, a phone walkthrough of a reefer, a scan of a spec vehicle, that is a conversation we have rather than a limit of the product.

How to connect Samsara to RapidEye

1. A Samsara account with DVIRs running. The vehicles and trailers your drivers inspect, on the DVIR templates you already use. Assets you have left out of the DVIR workflow stay out.
2. Defect photos on the inspection. They come from the Samsara Driver App, where a driver reporting a defect can attach a photo with the description. If a class of asset in your fleet is inspected without photos, RapidEye sends a capture link that opens on any phone, no app to install, a fixed set of angles taken at the walkaround.
3. Invite RapidEye as a user. Samsara's setup documentation has an administrator invite the new user by email, and the user accepts the invitation before logging in. The modal has the name and address to use. Nothing changes for your drivers.

What a Samsara DVIR carries

Checked against Samsara's help centre, September 2026.

According to Samsara's help centre documentation (samsara.com), a driver creates a DVIR from the DVIR tile on the Samsara Driver App home screen: pick the vehicle or trailer, go on duty if the clock says otherwise, fill in the asset details including location, choose pre-trip or post-trip, work through the areas marking each pass or fail, then set the vehicle safe to drive or unsafe and sign. In DVIR 2.0 the whole inspection can be completed offline and syncs when the phone is back on a network, and the driver has to be physically close to the asset to submit. Samsara calls the same report a Walkaround Inspection in the UK.

Photographs enter at one place: reporting a defect. The rules around that step are tighter than most fleets realise.

Samsara's DVIR product page adds that its AI quality checks confirm the driver was near the vehicle and that a submitted photo is of an actual vehicle component rather than an irrelevant image. So by the time a photo reaches your dashboard it is fresh, it is of the right part, and it was taken at the truck. The question the fleet still has to answer is who looks at it.

The three signatures, and what each one certifies

A completed report carries the current status of any defects, everything the driver entered, and three named signatures. Read as a chain, they are the whole accountability story of a defect in Samsara.

That last line is the one worth sitting with. It is a documented and perfectly legitimate path, and on a small crew it is the only practical one. It also means a defect can be opened and closed by the drivers of the vehicle, on the strength of two photos, with no third party ever comparing them. Nobody designed that as a hole; it is what a workflow built for compliance does at the edges, and it is exactly where an extra reader earns its place.

Does Samsara review DVIR photos?

It checks them, which is not the same as reading them. Samsara's documentation describes real inspection of the photo: DVIR 2.0 rejects a photo older than an hour or one that does not match the requested angle, and its product page describes AI quality checks confirming a photo shows an actual vehicle component and that the driver was at the vehicle. Those are quality controls on capture, and they are better than what most platforms do. Separately, Samsara's Media Library documentation notes that when no asset photo has been taken, the system falls back to the DVIR walkaround photo as the asset's profile picture, front for a vehicle, back for a trailer, which tells you these images are held as a real part of the asset record. What none of the documentation describes is anything reading a defect photo for content: whether the crack is longer than last month, whether the tread in frame is legal, whether the thing beside the defect is also a defect. That reading is a person's job, and RapidEye is a second one who never has a queue.

Samsara integration FAQ