A return or rental photo tells the story,
an agent reads it the same way every time
Returns and rental damage claims come in as photos. Deciding whether damage is normal wear, pre-existing, or something to charge for currently depends on whoever is reviewing that day. We build an agent that reads the submitted photo and compares it against the item's condition on record, where available. It classifies severity and flags anything genuinely ambiguous for a human to decide.
Why two reviewers reach two different verdicts
A customer returns a product or ends a rental and submits photos of its condition. Someone has to look at each one and judge how much damage is really there. Then decide whether it is normal wear, something to charge for, or grounds to reject the return. Without a baseline photo to compare against, this is a subjective call. It depends heavily on the individual reviewer’s standards and mood that day.
Inconsistency compounds across a team. One reviewer is lenient, another is strict, and customers who get different outcomes for similar damage notice and complain. That creates its own dispute volume on top of the original claim.
Turnaround is the third cost, especially when claim volume spikes: a rental business during a busy season, a return surge after a holiday. The backlog of unreviewed photos grows, and customers wait longer for a refund or resolution than the decision should actually take.
None of this shows up as one dramatic failure. It shows up as a steady drag. Work that should take minutes stretches into a backlog item. A quality bar holds on a quiet week and slips on a busy one. A team knows the fix is mechanical but never has a free afternoon to build it.
What the agent scores
The agent reads submitted damage photos and classifies severity against criteria your team sets for each category: scratch versus crack versus structural damage, cosmetic versus functional. Where a baseline exists, intake photos from a rental check-out, or original listing photos, it compares directly against that baseline instead of scoring in isolation. That catches both genuine new damage and pre-existing wear being wrongly disputed.
Every assessment gets the same criteria applied no matter which reviewer would have looked at it. That removes the day-to-day and person-to-person inconsistency that otherwise creeps into manual review. A case gets flagged for a human to make the final call whenever the photo is unclear, the damage is borderline, or the claim is contested. The agent’s assessment and reasoning come attached, so the reviewer is not starting from scratch.
Typical setup: your returns platform, rental management system or claims tool, wherever photos are submitted. Disputed cases land in your CRM or support desk for human review.
Where the charge decision still sits with your team
The decision to charge a customer, waive a fee, or escalate a dispute stays with your team. The agent provides a consistent severity score and comparison, not a final financial decision. Anything the agent flags as ambiguous, an unclear photo, a borderline severity call, a contested claim, goes to a person rather than being resolved automatically in either direction.
How the score stays defensible
Every assessment is logged with the photos, the baseline comparison if one existed, and the severity score, giving your team a defensible record for any customer dispute. Scoring thresholds start conservative, routing more borderline cases to human review than will eventually be needed. They loosen only once the criteria are proven against your own claim history.
Before it runs unattended, we run a side-by-side dry run against a sample of your own material. Your team sees exactly what it would have done. Every build ships with a short written runbook, so your team can pause it, adjust a threshold, or roll it back without waiting on us. The running-cost estimate below is a starting budget you set, with an alert built in before it is crossed.
Price and timeline
| Option | Price | What it covers | Timeline |
|---|---|---|---|
| Single automation | from $700 | Damage detection and severity classification from photos | 6 to 12 days |
| Department package | from $2,500 | damage assessment, returns handling and complaint intake across your operations team | 2 to 4 weeks |
Running cost is usually $10 to $80 a month in model usage depending on volume, with a budget cap set before launch.
Related
Pair this with refund and return handling so the assessment feeds directly into the resolution workflow. Warranty and complaint intake covers the front-end side of the same process. For marketplace listings that need photo-based condition checks before a sale, see product photo quality check. The full package breakdown is on the AI agents service page and the development service page. For a real build involving structured condition data, see the factory ERP recovery case study and the Bali real estate CRM case study.
Ready to stop relying on whoever is reviewing that day? Get in touch and we will build scoring criteria from a sample of your real claim photos.
Tired of doing this by hand? We can take the whole routine off your team, not only this step: Routine takeover, from $400 →
FAQ
How much does damage assessment automation cost?
from $700 to set up scoring criteria for one product or asset category, live in 6 to 12 days. Additional categories are quoted after a sample batch of real claim photos.
Does the agent decide whether to charge a customer?
It scores severity and flags disputes. The decision to charge, waive or escalate a claim stays with your team, since that often involves judgment beyond what a photo alone can settle.
How does it compare damage to the item's original condition?
Where intake or baseline photos exist, from a rental check-out or a product's original listing, the agent compares against them directly. Without a baseline it scores against a general condition standard you set.
What happens with a customer who disputes the assessment?
Every assessment is logged with the photos and the reasoning. Your team gets a clear record to review the dispute against, rather than relying on the original reviewer's memory.
Can it work across different product or asset types?
Yes. Scoring criteria are built per category since damage looks different on electronics, furniture, vehicles or clothing. Each category gets its own criteria set.