Meta Model API

CASE STUDY

Bringing drag-and-drop simplicity to AI development with Llama

At a glance

Industry

Technology

Use case

Simplifying AI development with AI-assisted training, fine-tuning, deployment and management

Goal

Make creating custom AI agents accessible to non-technical business users

Llama versions

Llama 3.3 70B, Llama 3.2 11B, Llama 3.2 3B and Llama 3.1 405B

Deployment

IBM watsonx

Download full case study

THEIR STORY

Using AI to transform AI development

Founded in 2019, AddAI is dedicated to simplifying and accelerating the implementation of AI assistants with user-friendly tools that make AI development accessible — even for companies that don’t have their own AI teams. Customers across sectors are using the AddAI platform to create personalized AI assistants for customer experience, communication and support.

THEIR GOAL

Put the keys to AI development into business users’ hands

Building custom AI agents that understand a business’s unique brand, knowledge bases and customer needs is a massive challenge for skilled AI developers. Getting to a functional, expert AI agent can require model training and fine-tuning, custom retrieval augmented generation (RAG) workflows and complex pipelines filled with tool calls, API requests and exacting data management. The AddAI team wanted to bake all of that AI development complexity into a drag-and-drop platform where business users could create their own personalized, expert AI assistants without the help of developers.

THEIR SOLUTION

Use Llama to do the technical work of AI development

AddAI created an intuitive, no-code platform for developing and deploying expert AI assistants. They used Llama models throughout the platform to perform technical development tasks beyond the average user’s abilities.Llama models work in the background to supervise model training, connect pipeline elements and orchestrate tasks. The platform includes a Llama-powered assistant that coaches customers through prompt engineering, fine-tuning, deployment and performance analytics. Llama models also support end-user experiences by powering AI assistant conversations, coordinating intuitive multi-turn logic and generating responses in RAG-based discovery workflows.

THEIR APPROACH

Build fast with Llama and IBM watsonx

The AddAI team built its platform using IBM watsonx. With watsonx, the team was able to call different Llama models via API, fine-tune them for their specific use cases within the platform and architect complex, multi-model pipelines.Llama’s open-source licensing and open weights allowed the AddAI team to create the custom IP their platform needed to make developing AI a simple, code-free experience. Key elements AddAI built with Llama include: an AI autopilot that helps customers construct and manage assistants; conversational agents with decision-making capabilities; and a RAG workflow for intelligent, contextual discovery.

THEIR SUCCESS

Millions of expert AI interactions, thousands of hours saved each month

AddAI customers across sectors — banking, e-commerce, telecommunications and automotive — have used the platform to build and deploy custom, personalized AI assistants for customer experience (CX), support and communications. Since the platform makes it easy to train models on unique knowledge bases and includes RAG workflows, customers have achieved remarkable accuracy with unique industry vocabularies, arcane subject matter and even brand tone and assistant personality.

500,000 end-customer interactions per month

“We are preparing on-premise Llama deployments that will give us complete control over operations, data and the model customizations our clients need. The ability to adapt, optimize and deploy models in our local environment was decisive in selecting Llama.”

Tomáš Jindra, COO and Co-Founder, AddAI

“Our mission is to simplify AI assistant development so that even companies without an AI team can build and deploy their own customer agents. Llama models make that possible behind the scenes, within the AddAI platform and in the production AI assistants that serve over 500,000 queries a month.”

Jindřich Chromý, CEO and Co-Founder, AddAI

“Llama models match the performance of commercial models with significantly better pricing structures and lower operating costs, which is critical for our platform’s ability to scale. Llama models provide excellent performance, flexibility and openness. Superior cost efficiency and the ability to deploy locally give Llama models a substantial advantage over the competition.”

Matyáš Leiský, Lead AI Developer, AddAI

Previous slide

Next slide

Slide 1 of 3

Please confirm your identity