Calendar
Wednesday
- Free
- Leak · sink
- Descaling
After 7 p.m.
At 11:47 p.m., someone picks up.
A leak under the sink. The office is closed. The answering service listens, offers initial guidance and arranges what comes next during the call.
23:47 · Understanding the call
The service listens for what matters.
CustomerHello, I have a leak under the sink. I closed the shutoff valve, the water has stopped. Can you come by tomorrow morning?
AgentUnderstood. Keep the shutoff valve closed. I can offer you tomorrow at 8:30 a.m. and I’m passing the request to the technician.
23:48 · What follows
It books, confirms and passes it on.
AI agentcall ended · 23:48
CRM
New customer · file created
Leak under the sink, shutoff valve closed.
Advice given: keep the valve closed.
Call recorded · 1 min 47 sec
Text to the customer
AquaService: appointment confirmed for Wednesday at 8:30 a.m. Please keep the shutoff valve closed until then.
Team
Wednesday’s route
Contained leak under the sink, visit at 8:30 a.m.
At 8:30 a.m., everyone knows what to do.
The customer knows the arrival time. The technician finds the reason and the advice already given. The office finds a complete file in the CRM.
Reconstructed call · client’s name changed
Behind the scenes
Before automating anything, we wrote down the rules.
During the day, the office answered. After 7 p.m., calls went to voicemail. In the morning, the team had to call every customer back, piece their request together and tell the leak that could wait from the one that needed someone on the road right away.
The company described, case by case, what a good answer had to produce: give the first useful advice, find or create the file in the CRM, offer an appointment, then alert the right person. An emergency calls the on-call technician immediately; a visit planned for 10 a.m. joins the technician’s route before 9 a.m. The workflow came afterward, to apply those rules to every call.
The client’s name has been changed.
The workflow, replayed
Three calls. Three different outcomes.
Each scenario follows the same workflow, then branches depending on the language, the customer and the urgency. Pick a call: the page replays the information read and the actions triggered.
Tuesday · 23:47
“I closed the valve and the leak under the sink stopped.”
RetellPick up the call06 12 •• •• 48 · answered
ChatGPTUnderstand the requestvalve closed · leak contained
CRMIdentify the customerNew customer · file created
CalendarBook the slotWednesday 08:30 · booked
SMSSend confirmationConfirmation sent
TeamAlert the team08:30 technician alerted
CRMLog the callSummary and recording attached
Replay leak
- Retell06 12 •• •• 48 · answered
- ChatGPTvalve closed · leak contained
- CRMNew customer · file created
- CalendarWednesday 08:30 · booked
- SMSConfirmation sent
- Team08:30 technician alerted
- CRMSummary and recording attached
Appointment confirmed in 1 min 47 sScheduled follow-up: technician alerted
What the call leaves behind
A reply for the customer. A clear next step for the team.
01The customer
They receive initial guidance, a time slot and written confirmation.
02The technician
They know the reason, the urgency and the appointment before hitting the road.
03The office
They find the file, the recording and the actions already taken in the CRM.
The tools
What works during the call.
The conversation
Picks up, answers in the caller’s language and checks the calendar during the call.
Analyzing the request
Identifies the reason and the urgency, translates, reads photos and writes the summary.
The actions
Applies the rules set with the company: calendar, CRM, SMS and team alerts.
CalendarReads availability and books during the call
CRMFinds or creates the customer record, then logs the exchange
SMSConfirms the appointment and sends initial guidance
TeamAlerts the right person at the right time
Retell, OpenAI, ChatGPT and n8n are trademarks of their respective owners. The delivered workflow has twenty-eight executable steps; the page shows seven to keep the journey readable.
Beyond the phone
One day, three workflows.
Once the rules are written, the same principle applies to other tasks: monitoring a temperature, processing invoices or preparing a quote. Each workflow reads a piece of information, decides and leaves a record.
03:15
Food service
The cold room calls the chef.
A door left ajar after service, €3,000 of stock warming up. The probe sees it, n8n wakes the right person.
01 / 06Every 5 minutes, the probe in cold room no. 2 sends its reading.
02 / 06Three readings in a row above the threshold: the deviation has lasted since 03:05.
03 / 06The chef gets a text with an “On my way” button.
04 / 06Ten minutes to reply. At 03:25, n8n places a call.
05 / 06Retell calls, announces the temperature and notes the chef’s answer.
06 / 06The deviation, the call and the corrective action go into the temperature log.
Cold room no. 23.2 °C
SMS · head chef · 03:15
Cold room CF2: 4.8 °C since 03:05 (threshold 4 °C).
On my wayWaiting for a replyVoice reminder at 03:25
Retell call · chef · 00:38
Agent Cold room no. 2 has been at 5.1 °C since 3:05 a.m.
Chef The door was left ajar, I’m on my way.
Temperature log
Oct 3 · 03:25 · CF2 · 5.1 °C max
Action: door closed at 03:31
LoRaWAN probe
Read the temperature
CF2 · 4.8 °C
Code
Read the trend
3 readings above 4 °C
Twilio
Alert the chef
SMS sent · 03:15
Wait
Wait for acknowledgment
10 min elapsed · reminder
Call the chef
Answered · 00:38
Google Sheets
Keep the log
Deviation and action logged
The cold room is back to 3.2 °C. The stock is saved, and the log is up to date for the next inspection.
n8n workflow · 12 nodes · .jsonThese three workflows are examples created for this page. Each workflow can be imported into n8n; the companies and amounts are fictional. The brands mentioned belong to their owners; icons from Simple Icons.