Treat the missed-call event as unfinished work.
A voicemail is passive. A recovery workflow creates a lead record with the caller number, received time, dialed line, business-hours status, prior-customer match, call attempt history, and a due time for the next action. If the call comes from a known customer, route it with that context instead of pretending every missed call is a new lead.
Suppress obvious spam and repeated events carefully. Two calls in five minutes may be one urgent customer, not two leads. A second missed call should update the active record and may increase priority. Do not discard it simply because the phone number already exists.
| Signal | System action | Dispatch or staff action |
|---|---|---|
| Known active customer | Attach the missed call to the open job or account | Route to the responsible office or field owner |
| New caller during business hours | Send approved acknowledgement and create response task | Call, qualify, and schedule under normal rules |
| After-hours service request | Collect service type, location, and safe routing facts | Use the on-call path if the request fits its rules |
| Repeated call or concerning language | Pause ordinary messages and raise priority | Follow the defined escalation path |
| Invalid, blocked, or unclear contact | Keep a visible exception | Review call log or voicemail when available |
Collect only what changes routing.
An immediate text-back can ask for the service needed, job location, preferred callback time, and whether the caller is an existing customer. Avoid a long intake before a person responds. The goal is to recover enough context for the right callback, not to force a customer to estimate scope through text.
Use approved categories that match how the company dispatches work. "Emergency" should not automatically make a promise. It can trigger a defined escalation and tell the caller what response to expect. Anything outside service area, hours, trade, or safety boundaries needs a clear human or referral path.
The callback queue needs ownership and aging.
Every recovered call should show the owner, due time, attempt count, latest customer response, and next action. When a dispatcher calls back, that activity must update the same record and stop the automated sequence. Otherwise the customer receives a generic follow-up after speaking to someone.
Plan for failed delivery, landlines, shared family numbers, wrong-number replies, existing customers with multiple jobs, and calls that reach the field team's direct extensions. These are normal conditions. Put them in an exception queue with a named reviewer.
Customer test: The first message should identify the company, acknowledge the missed call, offer a simple next step, and avoid pretending the system knows the situation. A calm, factual response beats a long "AI receptionist" conversation.
Start with missed calls, then connect downstream work.
- Review several weeks of phone logs and classify known customer, new lead, spam, after-hours, and repeat calls.
- Define deduplication, assignment, response due time, escalation, and close-out rules.
- Create the lead or account task from the phone event.
- Add one approved acknowledgement and capture only routing facts.
- Make staff activity update the record and cancel pending messages.
- Track exceptions and compare recovered calls with the phone log.
- Connect estimates and scheduling only after the callback state is reliable.
Measure valid missed calls, time to first staff response, calls with no owner, recovered conversations, booked work, duplicate records, failed messages, and customers contacted after the issue was already resolved. The honest outcome is fewer good calls lost, not an inflated count of automated conversations.