CodeixLab built an AI receptionist SaaS for appointment-heavy local businesses. The idea is not a chatbot bolted onto a website. It is AI-assisted front desk operations, with real tenant-owned records behind every conversation.
Project overview
- Industry: AI SaaS for local service businesses
- Services: AI automation and integrations, SaaS product development
- Platforms: Website widget, QR pages, kiosk mode, voice and phone channels, business dashboard, SaaS admin
- Tech stack: Ruby on Rails, PostgreSQL with pgvector, OpenAI, Twilio, Stripe
The challenge
Small service businesses answer the same questions every day: opening hours, services, prices, appointment requests, wait times and follow-up details. Staff also need to know which conversations became leads, which customers need a human, and which requests are safe to act on automatically.
Generic chatbots answer questions but cannot be trusted to take actions. A front desk assistant has to create real leads and appointment requests, and it must never do something the business has not allowed.
Our approach
Rails as the source of truth
Every public channel, whether a website widget, QR page, kiosk, voice session or phone call, creates a tenant-scoped conversation in the core application.
AI answers, the backend acts
The assistant answers from each business’s approved knowledge using retrieval (RAG) on pgvector. When it wants to do something, it requests a tool call. Rails validates every action before creating leads, handoffs, check-ins or appointment requests. Retrieval and memory are used for answers, never as permission to change records.
Clear handoff to staff
Conversations that need a person are flagged and routed to staff with full context, so customers are not asked to repeat themselves.
Usage-aware billing
Usage events are recorded as part of the workflow, and billing checks run before expensive actions, so the SaaS stays profitable per tenant.
Key features
- Multi-channel entry: website chat, QR codes, kiosk, voice and phone via Twilio
- Answers grounded in each business’s own approved knowledge
- Validated tool calls for leads, check-ins and appointment requests
- Staff handoff with conversation history
- Business-owner dashboard and SaaS-admin boundaries
- Stripe subscriptions with usage tracking and audit logs
Engineering decisions that mattered
- Tool calls as backend actions: see our guide to designing AI tool calls safely.
- Tenant boundaries everywhere: knowledge, conversations and billing are isolated per business.
- Booking rules before AI: the assistant only books against reliable availability data.
What this means for your project
Clinics, salons, restaurants and service businesses in Dubai and across the UAE can turn repetitive front desk work into structured, automated workflows. Explore our AI automation services or book a free consultation.
Outcome
AI chat turned into operational software: tenant boundaries, validated tool calls, audit logs, staff handoff paths and usage-aware billing.