WaveAccess Newsletter - When ChatGPT does not quite hack it! How companies should use LLMs when adopting AI. Part 1

Hello Subscribers!

It's been a while since our last update, and we're excited to reconnect with you!

Over the past eighteen months, large language models (LLMs) have revolutionized how businesses interact with customers, create content, and write code. Companies are now uncovering new opportunities with these powerful tools, but they also face challenges, such as maintaining context and relevance in responses.

But with these opportunities come risks. LLMs are often seen as a "black box": they can forget context, give irrelevant answers, and ultimately disrupt processes. The challenges of LLMs only add to the complexities typical of any AI project.

Unlocking the Potential of LLM: A Case Study in AI Integration

Businesses are intrigued by the potential of Large Language Models (LLM), but monetizing these powerful tools is not straightforward. To embark on an engaging project, such as developing an AI to optimize specific processes, significant intellectual and financial investments are required.

Achieving unique results demands unique data—data that your business has accumulated. You need data science experts and prompt engineers. These professionals can prepare the data, ensure privacy, and prevent the model from generating errors. Additionally, you need people who can seamlessly integrate the new tool into your existing infrastructure.

Under these circumstances, it might seem easier to entrust an LLM with tasks like text correction rather than something with a higher business value. However, the emergence of ambitious projects that fully leverage LLM show this is entirely possible.

We were approached by a client—a Forbes 2000 manufacturing company—entering a foreign market with a new gaming equipment brand. The company needed to analyze the effectiveness of its customer support. The quality of this support was crucial for the new brand's establishment in the market. These were real dialogues with open-ended questions, requiring AI for accurate evaluation.

The case we want to share features all the “favorite” challenges of AI projects:?

• Structuring and processing real written speech to create a custom solution
• Using an LLM in an unusual way as an auxiliary tool for initial data labeling required before AI training
• Privacy expectations: customer data is GDPR protected and cannot be transferred to any foreign environment,, while the information about the company’s new devices must not leak as well?
• Extensive prompt engineering?
• Further integration of the resulting model into corporate software
• The ambitious goal of completing all this in just a month?

Spoiler alert: we solved the problem using a hybrid approach. We used the ValueXI platform to anonymize dialogues and securely access ChatGPT. The LLM was then employed for data labeling, followed by ValueXI again for verifying this annotation and developing a custom AI. As a result, the client identified weak points in their support system, updated the FAQ on their website, and plans to develop a chatbot tailored to frequently asked questions.

Coming Up Next: Understanding and Applying Large Language Models in Business

LLMs are powerful tools for content creation, SEO optimization, recommendation systems, and natural language responses. They are transforming traditional search tools and taking on complex tasks like information labeling, trend prediction, and coding. Our next edition will explore how to effectively utilize LLMs in your business.

Keep in touch and stay tuned for more insights in our next edition!