Classy AI Messaging.
Connecting Facebook enquiries, approved clinic knowledge and human follow-up in an AI messaging workflow built around the next useful customer action.
Explore the thinkingConversation strategy, AI workflow design & implementation
Beauty & aesthetics
Customer journeys · n8n · Meta Messaging API

Implementation diagram of the actual enquiry workflow: eligible incoming messages, approved context, an AI decision and either a response or staff review. This documents the architecture and is not a screenshot of a customer conversation.
A useful reply has to understand the conversation.
A clinic enquiry can move from a service question to pricing, a branch preference and a request to book. A generic answer can lose that intent or repeat information the person already supplied. The workflow needed to answer routine questions clearly, retain the context needed for the next step and recognize when the conversation belonged with the clinic team.
Design the conversation and the operating workflow together.
I implemented an n8n workflow connecting Facebook Messenger events, approved clinic facts and an AI response layer, with an Instagram counterpart using the same operating principles. The workflow answers in the customer’s language, including English and Persian, and uses structured context for matters such as the requested service, branch preference and booking intent. The next step can be a direct answer, an appropriate booking route or an actual staff-review request.
Different decisions.
One connected approach.
Ground answers in approved business facts.
Service information, product details, prices, promotion dates and booking links come from approved clinic knowledge. Date-aware rules keep expired offers out of current replies. A controlled update path supports authorized pricing corrections, while customer messages remain enquiries rather than a source of new clinic policy.
Preserve intent without inventing a booking.
A branch choice or follow-up price question should not erase an earlier request to book. The workflow retains relevant preferences and sends the appropriate next step without restarting the conversation. It distinguishes sharing a booking link from confirming an appointment, a location or provider availability.
Make human involvement part of the implementation.
Clinical concerns, requests for a person and uncertain details are routed to staff review. A real handoff creates a review request and pauses routine automated replies. When staff take over a conversation, the assistant stays paused until the corresponding review is closed, so automation does not compete with the person helping the customer.
Protect the event and response boundaries.
The implementation checks incoming event authenticity, message eligibility and the allowed reply window. Deduplication and echo handling prevent repeated events or the assistant’s own messages from creating another reply. Structured response decisions and minimal conversation-state metadata make the flow easier to inspect and test.
Test how the conversation continues.
Validation covered routine price questions, follow-up context, English and Persian replies, branch changes, booking requests and staff takeover. The recorded November 2026 checks include real Messenger and Instagram conversations alongside isolated synthetic tests. The assistant supports reception and consultation routing; clinical decisions remain with the clinic team.
The interface, in context.
Implementation diagram showing how approved clinic information and structured conversation context inform replies. No private messages, account identifiers or operational credentials are included.
Implementation diagram of the human handoff and takeover pause: a review request reaches the team, routine automation stays paused and the conversation can resume after review is closed.
- Implemented Facebook Messenger enquiry-response workflow
- Instagram messaging counterpart with shared operating principles
- Approved-knowledge and date-aware pricing layer
- English and Persian conversation handling
- Structured service, branch and booking-intent context
- Staff-review queue, notifications and takeover pause controls
- Inbound verification, reply-window, duplicate and echo guards
- Live conversation checks and isolated workflow validation
A working conversation-to-consultation workflow.
The implemented system connects routine clinic enquiries to an accurate answer and a useful next step, with staff involvement built into the process. Recorded live tests on Messenger and Instagram verified pricing, booking context and branch handling, while handoff and pause checks verified how the team can take over. The work demonstrates conversation design, practical AI integration and operational follow-through; it does not claim a measured conversion or appointment uplift.
Explore a related service.
See the scope and process, then share the problem you want to solve.
Notes from the work.
All insightsAI Marketing Agents: What to Verify Before You Trust a Campaign Recommendation
A practical framework for reviewing AI marketing agents: data coverage, evidence, campaign assumptions, human approval and a contained first evaluation.
Read article Digital marketing & growthConnected digital marketing: from the first click to the next useful action
How I connect paid media, websites, CRM, creative production and measurement into a practical digital marketing workflow at Airport Kia and across my projects.
Read article AI & customer journeysAI-assisted Messenger journeys: keep the enquiry connected to a human next step
Lessons from building and testing Classy Cosmetic Clinic messaging workflows: bilingual service context, qualification, lead capture, handoff and pause/resume controls.
Read articleStrategy. Technology.
A connected way of working.
I’m Farshad Ghobadvand. Let’s talk about the next problem worth solving.

