A customer is standing in front of a machine that did not dispense a product. They should not need to install an app or understand the operator's internal structure. A phone call is a natural entry point. For operations, however, the value lies in what follows: a complete and traceable report.
1. Capture the issue in the caller's own words
The voice agent starts with an open question. The customer can explain the situation before being forced into a category. Targeted questions then fill gaps. A public vending fault form asks for details such as machine number, date and time, useful orientation for a voice process as well.
2. Complete the required data
Depending on the case, the workflow needs a location, machine or equipment area, time, observed behaviour, payment or product context and a contact option. OUTAG3 structures the responses into configured fields while retaining the customer's wording for review.
Unknown values should remain visible. If a caller does not know the machine number, the system must not invent one. It can ask for location markers or pass the case to the team with a verification flag.
3. Apply triage as a suggestion
Priority follows operator rules. A full location outage may need immediate attention, while a cosmetic issue can enter planned work. Impact, alternatives, repeated reports and support coverage matter.
AI suggests category, priority and routing. The operator team makes the decision about money, on-site work or closure. This prevents a plausible suggestion from turning into an unauthorised business commitment.
4. Deliver the case to the working channel
The structured case moves to the agreed mailbox or ticket system. Defined urgent categories can trigger faster notification. Other cases remain visible for scheduled review. The exact integration and escalation path is configured for the pilot rather than assumed for every operator.
5. Measure intake quality
Useful measures include complete-case rate, follow-up questions per category, time to the first team decision, operator corrections and recurring patterns by location. These measures improve questions and rules. They do not replace the professional assessment of a case.
See the full flow in the product overview, definitions in the glossary and setup answers in the FAQ. A limited set of real categories can be tested through a pilot request.