Collect the photos while the customer is still on the call
Website edition prepared November 14, 2026. Practical workflow; documentation checked on the date listed below, not a product-launch announcement.
The contractor is on a ladder holding a gutter in place. The phone rings on the open tailgate of the truck. A prospective customer wants to explain a leak, but the contractor cannot safely stop the job to collect the details. Answering the call is useful. Helping the customer prepare the information for a later conversation can be more useful still.
Integrations are scoped and tested separately.
Keep the call and the next action connected
Our new short film follows a fictional version of that situation. An AI assistant identifies itself, hears that the gutter is leaking and asks whether it may send a photo-upload link by text. The caller agrees. The link is sent while the conversation is still happening, and the caller opens it to add a photo.
The film uses synthetic details and voices. It illustrates a possible workflow rather than a live customer call or an already installed ProgressorAI integration. The practical question behind it is specific: can a busy contractor return a call with the customer's description and photos already in the same request?
Start by agreeing what the assistant should collect. For a gutter enquiry, that could include the service area, the customer's description and a preferred callback route. Keep uncertainty in the record. A caller saying that a joint appears to be leaking has made an observation; the assistant should not turn that into a confirmed diagnosis or a price.
Sending the link also needs a clear place in the conversation. Ask whether the customer wants it, check the agreed destination and explain what it is for. If they decline, keep a useful callback request without forcing them through an upload. The person calling may be unable to take photos or may prefer to speak with the contractor first.
Documentation checked: Retell AI — Custom functions, November 14, 2026
A spoken promise needs a delivery result
There are documented building blocks for actions during an AI conversation. Retell's custom-function documentation explains connecting configured functions to a receiving endpoint. Twilio's Message resource documents creating and tracking messages. Those capabilities require the relevant accounts, configuration and integration; they do not establish that any particular business has a working call-to-upload process.
The implementation should distinguish requesting a message, a provider accepting it and the available delivery outcome. The assistant should describe only the result it can verify. If sending fails, acknowledge the problem and offer the agreed alternative. Repeating that a link was sent cannot repair an unsuccessful delivery.
The upload needs to reach the intended request. A practical design uses a limited upload route tied to that enquiry, with the access and expiry behaviour agreed separately. It should not expose unrelated jobs or customer records. Ask for relevant photos and explain their purpose; avoid collecting documents that the contractor does not need.
After a photo arrives, the team should see it alongside the original description and callback task. A message receipt, a successful upload and a reviewed job remain different states. If the customer opens the link but sends nothing, the record should still show what is missing rather than imply the contractor has all the evidence.
Let the contractor take it from there
In the final part of the film, the contractor is safely on the ground and personally calls back. The photos are ready for review. The assistant has not promised a repair, approved a visit or supplied a diagnosis from the image. The professional can ask the next question with more context.
A sensible first test follows the whole path using fictional details: the customer agrees, the message reaches the test phone, the upload joins the correct request and a colleague finds it before calling back. Also try a declined link, a failed message and a missing photo. These cases reveal whether the process remains useful when somebody does not follow the ideal sequence.
At ProgressorAI, we can discuss that handoff around the company's actual phone and receiving systems. Access, message delivery, upload controls and the staff workflow need checking before a particular connection is promised. Which information would make your next callback easier?