Collaboration on an AI Voice Agent for Saona

Saona and the Challenge of Managing Phone Calls

The client is a Saona franchisee, a restaurant where customers can dine in by booking a table in advance or simply walking in to ask if there’s room; they can also order for home delivery or order and pick it up at the restaurant. It also offers phone assistance for opening hours, birthday parties, group menus, and more. One of the main challenges the business faced in its daily operations was managing phone calls. The restaurant initially had a single phone line, and at certain times of day several calls would come in at once, meaning some of them couldn’t be answered. On top of that, a significant share of callers would hang up if the call wasn’t answered within the first few rings, so the in-store team felt compelled to answer the phone immediately, even while serving another customer in person.     This created a second problem: the in-person customer experience suffered because of the priority given to the phone. Depending on who was working at any given moment, either the in-store customer or the caller would be prioritized, leading to inconsistent service and an uneven customer experience. To solve this problem, TechsBCN developed and deployed an AI-based voice agent, integrated with the company’s telephony infrastructure and internal systems.

A Hybrid Model Between Human and AI Support

The system works on a hybrid model between human and AI support. When a customer calls, the phone first rings in the restaurant for one or two rings, giving an employee time to answer directly. If the call isn’t answered, the AI agent automatically joins the conversation, identifies the caller’s intent, and handles different types of requests. One of the main use cases is handling home delivery orders: the agent asks the customer what they’d like to order, gathers all the necessary information, and uses the caller’s phone number to automatically look up the customer database. If the number is already registered, the agent identifies the customer and only asks them to confirm their identity and the necessary details, such as the delivery address. Once the conversation ends, the order is automatically sent to the kitchen team. A second use case is answering business questions: the agent has a knowledge base to answer questions about opening hours, products, food characteristics, services, or table availability. When a question goes beyond the information it has, an automatic escalation to an employee kicks in, without the customer having to call back or being left unattended: while the AI keeps talking with the customer and gathering additional information, the system places an internal call over another line to find an available employee, briefs them on the reason for the call, and, once confirmed, connects both calls seamlessly for the customer.     A third use case is the automatic management of reservations and bookings: the agent checks availability for a specific date and time and books it directly; if there’s no availability, it automatically looks for nearby alternatives that same day or similar time slots on the following days.

Full Traceability and AI-Driven Quality Control

Every interaction handled by the agent is automatically logged in a call CRM, storing for each one:

• Customer name
• Phone number
• Date and time of the call
• Duration
• Type or reason for the call
• Actions taken during the conversation
• Call outcome
• Full transcript of the conversation
• Relevant information gathered during the interaction

This gives the business full traceability of every conversation, whether it was handled entirely by the AI or later transferred to an employee. Additionally, TechsBCN added a second AI layer responsible for automatically evaluating the quality of the voice agent’s conversations: it analyzes transcripts, scores each interaction, and allows business managers to quickly spot calls with an unsatisfactory experience, so they can proactively reach out to the customer before they even file a complaint. On the technology side, the solution is built on Microsoft .NET for business logic, integration with corporate systems, and exposing services via APIs. Automated process orchestration is handled with n8n, coordinating the voice agent, the CRM, and the customer, booking, and operations systems. The voice conversational layer is built with Vapi, integrated with a virtual switchboard based on SIP technology, and includes Speech-to-Text, Text-to-Speech, large language models, and a proprietary business knowledge base.

The Result

Thanks to this project, the phone stops being something that constantly interrupts the restaurant’s operations and becomes an intelligent service channel, capable of answering, selling, informing, booking, escalating conversations, and automatically monitoring service quality. The goal of the project is not to replace the people who serve customers, but to let them focus on the interactions where they truly add the most value. At TechsBCN we are proud to have built a solution where Artificial Intelligence doesn’t replace the in-store employee: it simply keeps them from having to look away from the customer in front of them just because a phone is ringing.