Luna organises Instagram enquiries for El Mago del Color, a hair colour salon in Bogotá. The project connects direct messages, story replies and enquiries from comments with a working inbox. The team keeps the context needed to prepare an assessment, coordinate an appointment and continue the client relationship.
Real recording: from a comment to a direct message
This June 2026 recording follows a balayage enquiry through a comment, a public reply and a direct message requesting photos and chemical history to prepare a professional assessment. It also shows the public response to a compliment.
The edit includes reading pauses and protects personal information with crops and permanently encoded masks. Messages retain their original content. The interface and walkthrough captions remain in Spanish, with the original mixed-language messages. Watch the complete Luna Instagram recording (45 seconds).
From content to a conversation with context
A story reaction and a question about a service express different intentions. Luna identifies the source of incoming events and adapts the conversation to that context. History, tags and private notes help the team understand what the person needs before suggesting the next step.
For comments, the workflow distinguishes service enquiries, social messages and ambiguous messages. Compliments and brief reactions receive a public reply; a service enquiry can continue through a private reply associated with the comment, when the channel allows it. Further conversation depends on the person replying and on Instagram’s conditions.
Photos and references to prepare an assessment
Image analysis with Gemini distinguishes a photo of the client’s current hair, a reference for the desired result and other image types. That information helps request missing details and prepare the consultation. The salon professional remains responsible for the technical assessment, feasibility of the change and service decision.
The panel keeps files alongside the conversation. Receipt recognition records that a receipt was submitted; the team must check that the funds reached the account. Sharing a QR code makes deposit instructions easier to follow, but does not turn Instagram into a payment gateway or verify a bank transfer.
A connected calendar and team-led coordination
From the panel, a team member can schedule, reschedule or cancel an appointment. The record syncs with the operational calendar in Firebase Realtime Database and is linked to the Instagram conversation. Enquiry context and visit coordination become part of the same journey.
If the client shares a phone number, the team can associate it with the contact to recognise the relationship with WhatsApp. Conversations retain their channel and its messaging conditions. Linking a number provides context; it is not blanket permission to initiate messages through another channel.
The inbox supports pausing the assistant per conversation, quick replies, files, tags and private notes. The workflow includes alerts and a cooldown after a human reply to help coordinate automated handling with the team’s intervention.
Follow-up with explicit messaging conditions
The reviewed implementation checks a 24-hour window from the person’s last message for ordinary automated sending. Reminders and follow-up use a delivery router that attempts sending when the channel permits it and records pending work for team review when it cannot complete the send.
Recording consent and holding a technical permission from Meta are separate conditions. Additional functions outside the ordinary window require a review of availability, approval and configuration. Human support outside that window also has its own conditions; it is not presented as an extension of automation.
Meta’s official Instagram Login documentation describes professional-account and permission requirements. Scope is defined around those requirements and the salon’s process before enabling additional functions.
What can be measured and how results are interpreted
The panel brings together interest signals, enquiry sources, photo and receipt submissions, pending work and survey responses. Interest classification uses rules based on messages and context to help prioritise attention. These states need interpretation: a submitted photo or recorded receipt does not, on its own, represent a sale or a completed visit.
The authorised El Mago del Color case reports 91% completed appointments and a 2% no-show rate for its first operational quarter, 14 March to 14 June 2026. The indicators cover the salon’s complete system and the team’s work. They do not isolate Instagram’s contribution or establish an improvement caused solely by this agent.
The integrated El Mago del Color case study explains the authorised figures and measurement period. This case focuses on the Instagram channel implementation and its connection with the operation.
Architecture and scope of this implementation
The project combines Next.js, TypeScript, Instagram API, Gemini, Cloud Firestore, Firebase Realtime Database and Cloud Storage, with scheduled tasks on Vercel. Gemini generates conversational replies and analyses images. Firestore stores context; Realtime Database connects the calendar, and Storage keeps files used by the panel.
Features are described from the reviewed local code. Their actual availability depends on configuration and channel permissions. The diagrams are illustrative: they explain the process without exposing client conversations, profiles or internal metrics.
To see how the parts fit together, explore the Luna WhatsApp enquiry and appointment case, the website with booking and loyalty or our AI agent service for beauty salons.
