RAG : Automation e-mail response with AI and LLMs

RAG to automate email responses in the IT industry

What does Senetic do?

Senetic is a global provider of IT solutions, supporting companies and public institutions in optimizing their daily tasks by creating intuitive digital ecosystems. Operating since 2009, the company has established itself as a leader in delivering high-end networking, server, software, and IT hardware solutions for small and medium-sized businesses worldwide. With 27 subsidiaries across the globe and sales in 151 countries, Senetic serves over 2 million customers annually. The company has been recognized with multiple awards and is a trusted Microsoft partner. To further enhance and develop its services, Senetic has partnered with Vstorm.

How does Vstorm cooperate with Senetic? RAG and automated emails

For Senetic, collaborating with AI was a new experience, so our first priority was to clearly explain the potential benefits that Large Language Models (LLMs) could bring to their operations. Our goal was to thoroughly understand Senetic’s needs, objectives, and challenges while identifying the areas where artificial intelligence could most effectively enhance its processes.

We identified a key area that was consuming significant resources and time for Senetic’s employees: email communication with clients. Since Senetic operates globally, customer inquiries arrive from all over the world, in various languages, which presents a significant challenge for email management. We proposed a solution to Senetic that would greatly streamline this process.

Previously, when a customer sent an inbound inquiry via email, the message would land in the company’s Microsoft Outlook inbox, where it awaited manual processing. An employee would then have to search through a database containing thousands of products and identify the most suitable response, taking into account the country, product, language, and any additional products that might be relevant to the customer. After selecting the appropriate products, the employee also needed to craft a well-structured email response that incorporated all these details and met the specific guidelines and expectations for communication with the client.

For this project, we aimed to leverage the full potential of LLMs, which are ideally suited for such tasks.

• Specifically, we aimed to utilize technologies like Semantic Search and Question Answering. We designed the process to adapt the LLMs to Senetic’s specific needs, ensuring they could handle the diverse and complex inquiries effectively. It’s also important to ensure the quality of training data when training LLMs to guarantee that the responses are accurate and unbiased.
• Generative AI, a transformative technology, played a crucial role in automating email responses by enabling LLMs to produce well-structured emails that follow guidelines and incorporate effective sales techniques.
• Retrieval augmented generation (RAG) further enhances the accuracy of the generated email responses by connecting multiple databases that aggregate unstructured data from various sources, allowing the model to access and generate responses based on real-time data retrieval.
• We also incorporated the LangChain framework, which supported us throughout the development of this solution.
• We used LlamaIndex to optimize document retrieval, improving the handling of unstructured data and enhancing the speed and accuracy of information extraction

This approach to Large Language Models enhances our ability to use neural networks and transformer model architectures for natural language generation, enabling better responses. The quality of training data is crucial in training the LLMs, ensuring that the responses are accurate and free from bias.

The solution we implemented works as follows:

The existing email inbox was integrated with LLMs and connected to the product catalogs, which are updated in real time. Product inquiries are automatically recognized, and the best possible options are selected from among thousands of products, based on the client’s needs. After identifying these options, the LLM generates an email according to a predefined structure, providing the client with the best available choices and solutions, along with any additional items that would complement the original selection.

Result

As a result of our solution, the entire email response process at Senetic has been fully automated with RAG. This automation not only saves significant time and money that was previously spent on crafting manual responses, but it also benefits Senetic’s clients by providing real-time answers to their inquiries. The swift response times enhance customer satisfaction and streamline their purchasing experience.

Moreover, this solution has enabled Senetic to achieve both hyper-automation and hyper-personalization in its customer interactions. By freeing up valuable time that was once devoted to handling routine inquiries, Senetic’s team can now focus on more strategic tasks, such as account management, allowing them to build stronger relationships with their clients.

Our mission is to help companies implement AI in an ethical way so that people can focus on what matters most. Through this project, we’ve empowered Senetic to harness the full potential of AI, not just to optimize their processes, but to reallocate resources toward what truly drives their business forward.