Pre-trip and post-trip inspections usually live in three places at once: photos in a messenger, defects in somebody's head, and the DVIR in a folder nobody opens until a roadside stop. The two questions that matter, is anything out of service right now and who still owes me a signed inspection, take a phone call to answer.
Inspections puts the loop in one module: the driver's report and its photo evidence, an AI first pass, the reviewer's decision, the shop's defect list, and the fleet-wide compliance picture.
The loop, five stages
- Driver evidence - filled on a phone from a texted link. No app, no login.
- AI first pass - reads the photos against the fields and flags what it doubts.
- Human decision - a reviewer approves. The AI never does.
- Defect resolution - findings become a prioritized work list for the shop.
- Fleet compliance - what is blocked, what is overdue, what is waiting on you.
Dashboard
Four headline figures, each with the sentence that explains it: out of service, overdue, needs review, and a compliance rate, all recalculated by the range switch for today, 7 days, 30 days or all. Under them, how the review queue splits and defects by severity, then the two queues that need a person: inspections awaiting a reviewer, and unresolved defects with how long they have been open. Every figure is an entry point, so a card opens the list with that filter already applied.
The review queue
A two-pane queue: the list on the left, the full report on the right. No page reload between them.
The list on the left, the full report on the right, no page reload between them. Status chips carry live counts, and each count is its own query, so the number on the chip is the number you get when you click it. Filters for template, reviewer, performer and date run in the database, across the whole fleet rather than the page you happen to be on. Every row shows how long the inspection has been waiting, so nothing quietly ages out.
AI first pass, never the verdict
A verdict on the inspection, then a line for every page it assessed, each with the reason spelled out in words a reviewer can act on.
The AI reads the report the way a reviewer would, photos against the fields, and writes down what it doubts: a verdict on the inspection, then one per page it assessed, each with the reason in words you can act on. An odometer photo that disagrees with the typed reading gets flagged; a checklist item marked fail with a note and a photo behind it passes. It scores the inspection, so a weak report reads as weak before anyone opens the photos. It runs on submission and can be re-run from the report. Approve stays with a human.
Photo evidence, grouped the way it was captured
The defect tire photo on top, then the labeled walkaround. Two of the four truck slots came back dark, visible in one second instead of after a roadside inspection.
Photos sit under the template page they belong to, truck and trailer, with a count per group, so a reviewer can go straight to the section the AI complained about. Every shot is labeled by its capture slot, driver side, passenger side, rear of the rig, truck front, which is how a trailer photo sitting in the truck section gets caught. Re-shoots collapse into their slot instead of stacking up as duplicates, and driver identity shots sit in their own block.
Defects, the shop's work list
One prioritized list for the shop, with the driver's own words and the evidence photo attached to each finding.
Defects raised inside an inspection get their own workspace instead of dying inside a PDF: status chips with counts, filters for severity and system section, and keyboard-first triage where a waiver asks for its reason before it goes through. Real actions, not tags: resolve, waive, reopen, mark duplicate. The detail pane carries the evidence photos, the driver's note in their own words, and a link back to the inspection. When the same defect has been raised before, the pane shows the chain, so a repair that did not hold reads as a pattern. The out-of-service switch says exactly what it is: a tag on the defect, not a change to the vehicle's status.
Map
Dispatch can see whether the rig that needs attention is two hours away or two states away before deciding what to do about it.
Pins come from the coordinates captured on the driver's device, with an honesty counter next to the legend, so you always know how many pins are real GPS rather than an assumption. The side panel puts the worst first: out of service, then rejected or awaiting the driver, then everything compliant.
Templates you control
Three ship with the module; each row shows its shape, pages, triggers and signatures, and whether it is active.
Every template shows its shape at a glance: pages, triggers, signatures required, and whether it is active. Three ship with the module, Driver Pre-Trip, Driver Post-Trip and Pre-Trip, and you can create your own.
Raising an inspection
Request inspection sits top-right on every tab. Pick the rig and the driver, and the driver gets a text with a no-login link that opens the inspection on their phone. When a reviewer approves, the driver is notified.
Why it matters
- One loop, one place. Driver evidence, AI first pass, human decision, defect resolution and fleet compliance, without leaving the module.
- Evidence you can defend. Photos labeled per capture slot, the coordinates the inspection was captured at, and a signature block on the report.
- AI that assists, not decides. It flags missing walkaround views, mislabeled shots and odometer mismatches. The approval stays yours.
- Nothing gets lost. Waiting time on every row, worst-first ordering on the map, and repeat defects shown as a chain.
How to use it
Inspections is in Beta. Open Maintenance → Inspections, run a few DVIRs through it, and tell us what breaks: [email protected]. Recurring inspection plans, who inspects what, how often, and when it turns overdue, are next.
