Meta Model API

CASE STUDY

Llama-powered generative AI agents automate ticketing and help technical support deliver excellent customer service

At a glance

Industry

Advertising

Use case

Automating customer support ticketing and communications

Goal

Enable better and faster customer support for a growing customer base

Llama version

Llama 3.1 8B

Deployment

On-premises Kubernetes

Download full case study

THEIR STORY

The AI-powered advertising platform that produces at scale

Smartly is an AI-powered advertising technology company that manages nearly five billion dollars in ad spending for more than 700 brands worldwide. Its platform seamlessly integrates the capabilities of media, creative and intelligence to reach and engage consumers and learn what performs best. Providing excellent customer service and technical support is a core Smartly value and competitive advantage in the marketplace.

THEIR GOAL

Free up technical support agents for improved customer service

The Smartly Support Operations team works behind the scenes to provide world-class technical and customer support. In addition to serving customers, the team managed support processes and communications, including manually creating tickets and composing resolution messages. As the Smartly platform grew, support processes became unmanageable, driving the team to look for ways to reduce workloads and save time with AI.

THEIR SOLUTION

Llama-powered AI agents for automated ticketing and communications

After testing several competing models for the application, Smartly chose open-source Llama, which showed superior performance, flexibility and ease of integration.Using Llama 3.1 8B, the team created an agent capable of populating multiple ticketing platforms, retrieving engineering and support notes and then summarizing complex technical issues. Thanks to Llama’s nuanced natural language understanding, the application could also generate customer resolution messages that maintained a professional, on-brand tone while clearly explaining technical topics.

THEIR APPROACH

Customized agents deployed on-premises for maximum data privacy

Llama 3.1 8B required no training or fine-tuning. The team delivered the quality responses they needed using basic prompt engineering and few-shot learning. The prompt template and context provided clear instructions, including formatting examples (such as ticket titles and descriptions), subject matter context and writing style.Smartly deployed the solution on their private, on-premises Kubernetes environment. Deploying locally kept them in complete control of their data and their customers’ information. The model and prompt template integrated seamlessly with Smartly’s existing platform, so other applications could easily access AI services.

THEIR SUCCESS

Support staff spend more time helping customers, less time chasing tickets

By taking on repetitive, meticulous tasks, the Llama-powered AI agents have fully automated support ticket creation, giving the support team time back to help customers one-on-one. Customer satisfaction scores for Smartly technical support are up, and support agents are happier.

Projected results:

80% less time to create support tickets
50% less time to draft customer resolution messages

Please confirm your identity