LLAMA 3.1 405B

Brain4Data

banner image
logo
logo
logo
CASE STUDY

Enhancing guest experiences with an AI customer service chatbot

At a glance

Industry: Technology

Use case: Automating micro-hotel customer service

Goal: Improve the guest experience and streamline operations

Llama versions: Llama 3.1 405B

Deployment: Oracle Cloud Infrastructure (OCI)

80%
of customer service requests are handled by Llama-powered AI
100%
multilingual — the agent converses in the customers’ language

*All results are self-reported and not identifiably repeatable. Generally, expected individual results will differ.

THEIR STORY

Empowering businesses to put AI to work

Brain4Data specializes in helping businesses use advanced AI technology and retrieval augmented generation (RAG) to prepare data for generative AI applications. Their digital AI assistant, FRED, transforms distributed company information into high-value, AI-ready knowledge bases and generates precise, context-specific answers. Currently, FRED generates 40,000 responses a month for clients in 30 countries.

logo

Roatel Holding AG is the leading provider of smart micro-hotels in Germany. Self-service Roatel facilities provide travelers with convenient, affordable lodging with all the modern conveniences. Based in Germany, Roatel plans to establish a network of hotels across Europe.

THEIR GOAL

Automating guest service to unlock business growth

Brain4Data’s client, Roatel, operated self-service micro-hotels in Germany. Hotel guests relied on a remote customer service team to answer their questions, but as Roatel began to expand, the service team struggled to keep up with increased demand. Brain4Data set out to build a scalable AI solution that could automate customer service in multiple languages and provide accurate, fast responses 24/7. The goal was to maintain guest satisfaction and reduce reliance on live customer service agents.

THEIR SOLUTION

A Llama-powered customer-service chatbot

Brain4Data used its digital assistant platform FRED, Llama and OCI Generative AI to develop a chatbot for Roatel. The chatbot provides fast, reliable and up-to-date answers to complicated questions — no matter the time of day.

Llama 3.1 405B is at the heart of the solution. Brain4Data chose the Llama model for its multilingual capabilities and strong performance in nuanced, semantic understanding — all delivered without fine-tuning. The team also knew the chatbot would benefit from an open-source model with broad community support.

their solution image

The chatbot responds to guest requests in 11 languages at any time of day.

THEIR APPROACH

Streamlining development with OCI tools

The chatbot is built on OCI using OCI Generative AI and a RAG workflow. When a guest asks a question, the system gathers relevant context from Roatel’s document repository and injects it into a custom-engineered system prompt that is passed to the Llama model, which generates a response in the guest’s preferred language. Behind the scenes, Oracle’s Database 23ai Vector Search identifies the most relevant document chunks from large volumes of unstructured data, and Cohere’s embedding model supports semantic understanding.

their solution image

The chatbot solution uses Llama to craft responses to guest queries in natural language.

THEIR SUCCESS

Streamlined operations and elevated guest experiences

With OCI, Llama and their FRED assistant platform, Brain4Data developed a fully operational chatbot in just two months. Today, the chatbot handles the majority of customer queries for Roatel. The result is reduced workloads for service staff, operational efficiencies for the business and enhanced experiences for customers. With its multilingual AI chatbot, Roatel is now set to scale its micro-hotel network across Europe.

    • 80% of customer service requests are handled by Llama-powered AI

    • 100% multilingual — the agent converses in the customers’ language

*All results are self-reported and not identifiably repeatable. Generally, expected individual results will differ.
With no fine-tuning, Llama provides excellent responses to challenging — not always well formulated — questions, in multiple languages.
"With no fine-tuning, Llama provides excellent responses to challenging — not always well formulated — questions, in multiple languages."

Jan-Peter Meyer - CTO, Brain4Data

Start building

Create generative AI applications for business with open-source large language models that bring unmatched control, customization and flexibility.
Download Llama
Horizon banner image

Stay up-to-date

Our latest updates delivered to your inbox

Subscribe to our newsletter to keep up with the latest Llama updates, releases and more.

Explore more
Dec 10, 2025
Stoque
Unifying internal knowledge for faster insights and efficiency
Tech
Read story
Nov 14, 2025
Contextual AI
Reducing hallucinations with a better context layer for AI agents
Tech
Read story
Nov 13, 2025
AITEM
Equipping veterinarians to make faster, better clinical decisions
Tech
Read story