Classify and auto-close customer SMS replies for a car dealer group
Stellantis &You UK's digital process team used Make, Knack and an OpenAI model to read customer replies to booking texts, judge intent and sentiment, and either alert aftersales or close the message. Over 12 months it analyzed 47,000+ messages, closed 18,000+ automatically and saved about 151 hours.
- Difficulty
- Intermediate
- Setup time
- 2 hr
- Monthly cost
- $20/mo
- What it saves
- about 151 hours a year
How to do it
- Connect your SMS provider to a Make scenario that triggers on each inbound reply.
- Send the reply text to an AI module that returns intent and sentiment.
- If the reply is a simple confirmation or thanks, mark it closed in your database.
- Otherwise notify the right team with the message and the AI's reasoning.
- Review closed messages weekly to catch anything that should have been escalated.
The prompt
A customer replied to our service-booking text message. Classify the reply as one of: CONFIRM, CANCEL, RESCHEDULE, QUESTION, COMPLAINT, THANKS, OTHER, and rate sentiment as positive, neutral or negative. Return JSON with intent, sentiment and needs_human (true unless intent is CONFIRM or THANKS with non-negative sentiment). Reply: [text]
Where this comes from
One person reported it. Promising, not yet confirmed. Original report: Make success story (vendor).
Tried this? Tell us how it went and help move its proof level up.