Verify proof-of-delivery photos before closing an order
When drivers upload a proof-of-delivery photo, a decision model checks that it shows a package at a door or agreed spot, and flags missing, blurry, or doubtful photos so the dispatcher can follow up the same day.
Uses image input, which TypeSafe AI labels a preview feature for Jev. Test it on your own sample images before relying on it.
An example pattern written by Agents Lookup. We have not yet found a public build of this exact job.
The use case
Local delivery businesses ask drivers to take a photo at each drop, but nobody checks them until a customer says the parcel never came. Jev looks at each photo and answers whether a package is visible, where it appears to be left, and whether the photo is clear enough to use. Photos where every answer clears a 0.85 confidence and show a package at a door, porch, or reception are marked verified. Photos showing no package, an unexpected location, or low confidence are flagged for the dispatcher. Jev does not decide whether a customer gets a refund or redelivery. The dispatcher reviews flagged drops, contacts the driver or customer, and decides what to do.
Example workflow
- Drivers upload a delivery photo from the driver app or a simple form at each drop.
- Make sends the photo and the delivery note to Jev with the image preview header.
- Jev answers a Noul question, a package is visible in the photo, a Choice question, where the package appears to be, with the allowed answers front door, porch or step, reception or mailroom, other location, and not visible, and a Noul question, the photo is clear enough to identify the location.
- Verified drops are logged against the order in a Google Sheet with the confidence scores.
- Flagged drops are posted to the dispatcher in Slack before the end of the shift.
- The dispatcher follows up with the driver or customer and records the outcome on the order.
Tools
- Jev (TypeSafe AI)
- Shopify
- Onfleet
- Make
- Google Sheets
- Slack
Risks and guardrails
- Image input is a preview feature in Jev 1.14 and TypeSafe says it is not production-ready yet — guardrail: test it on a few hundred of your own photos against a person's labels before relying on it, and keep a person checking every answer until the results hold up.
- A clear photo can still be at the wrong address — guardrail: compare the photo's location data with the delivery address in code, and flag mismatches separately.
- Night or rain photos may be judged unclear too often — guardrail: review the flag rate weekly and ask drivers to use the flash rather than lowering the threshold.
- Verification could be used to deny customer claims — guardrail: a verified photo is one input for the dispatcher, who still decides every claim.
Related recipe
Shopify Order-Exception Triage Agent: An agent reviews new Shopify orders for address problems, risk flags, stock gaps, split shipments, and customer notes, then sorts them, routes them, and drafts customer messages for a person to approve.
Want it set up for you? Request a quote.
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Shopify Dev MCP server (AI Toolkit) (official page): shopify.dev/docs/apps/build/ai-toolkit (checked 2026-11-09)
- Make MCP server (official page): developers.make.com/mcp-server (checked 2026-11-09)
- Slack MCP server (official page): docs.slack.dev/ai/slack-mcp-server (checked 2026-11-09)