Recent announcements from OpenAI, Meta, and Mistral could make generative AI more accessible to contact centers. OpenAI’s release of GPT-4o mini brings a significant cost reduction compared to its previous ChatGPT 3.5 Turbo model (a 60% decrease!), making it more affordable for developers to integrate generative AI into their applications.  

Meanwhile, Meta’s Llama 3.1 releases offer several fully open-source language models that rival the capabilities of proprietary frontier models, providing another option for businesses looking to adopt generative AI without high costs. Not to be outdone, Mistral also released a new more powerful open source model with its Mistral Large 2. It’s important to note that while both Llama 3.1 and Mistral Large 2 are open source models, there are some restrictions on their commercial use. 

Lower Costs Gives Rise to More Flexible Use Cases 

The significant cost reduction of generative AI models like GPT-4o mini and Llama 3.1 are a potential game-changer on two fronts. They are less expensive than their predecessors, and they offer developers more deployment options. Until very recently, using large language models (LLMs) was expensive, especially when sending large amounts of data into the context window. This forced users to be judicious with their prompts and data inputs, limiting the potential applications of LLMs.  

Now, with costs dropping rapidly, contact centers can explore more creative uses of generative AI. For example, they can send larger volumes of customer interactions into LLMs for summarization, analyze longer conversations for sentiment analysis and key insights, or even use LLMs to generate synthetic training data for agent assistance. The reduced cost barrier also enables more experimentation and iteration, allowing contact centers to refine their applications of generative AI and unlock new benefits. 

Comparing the Cost of Closed vs. Open Source Models 

When deciding between using OpenAI’s ChatGPT-4o mini and an open-source model like Meta’s Llama 3.1 or Mistral Large 3, it’s important to consider your specific needs and estimate the costs involved. For OpenAI’s model, you’ll need to calculate the API costs based on the number of tokens used and requests made, as well as consider any additional costs such as data storage and processing fees.  

In contrast, for companies that don’t exceed the threshold of 700 million monthly active users per month, self-hosting Meta’s Llama 3.1 could be an option. Such a step requires calculating the infrastructure costs of running the model on your own servers or cloud infrastructure, including compute resources, storage, and maintenance expenses. You’ll also need to factor in the costs of deploying and integrating the model, processing and preparing your data, and potentially hiring personnel with expertise in AI and machine learning.  

While operating an open source LLM isn’t free, it comes with potential advantages. By leveraging open-source models, developers gain the ability to modify and tailor the model to specific needs. Additionally, open-source LLMs can be deployed on-premises or in a private cloud, ensuring data privacy and security. With access to the model’s architecture and training data, developers can also experiment and innovate more freely, leading to potentially more accurate and effective AI-powered solutions. Companies can also use Llama 3.1 405B to train smaller models capable of handling targeted tasks at minimal cost.

For those companies less enthusiastic about going the open source DIY route, both Llama 3.1 and Mistral Large 3 can be used via an API pay-as-you-go model, similar to OpenAI’s products. 

More Powerful and Affordable CCaaS Solutions Coming? 

CCaaS solution providers are actively integrating generative AI capabilities into their products to enhance customer experiences and improve agent efficiency. Many initially leveraged OpenAI’s models, such as ChatGPT, to power features like chatbots, sentiment analysis, and automated agent assistance. However, with the emergence of more affordable and flexible alternatives like Meta’s Llama 3.1, CCaaS providers are now exploring new options to stay competitive. By incorporating generative AI, these providers aim to offer more advanced and cost-effective solutions to their customers, including improved conversation analysis, personalized customer interactions, and streamlined agent workflows.  

As CCaaS providers adopt more cost-effective generative AI solutions, customers can expect to benefit from reduced costs and improved pricing models. With decreased expenses related to API fees or infrastructure, providers may pass the savings on to their customers, making it more affordable for them to leverage generative AI capabilities. This could lead to a more competitive market, driving innovation and better services for customers. 

As the cost savings are realized, CCaaS providers may also invest in developing more advanced features, improving customer support, or expanding their offerings. While the cost savings may not always be directly passed on to customers, the increased competition and innovation driven by more affordable generative AI solutions will ultimately benefit customers in the form of better value propositions and improved capabilities. 

Categories: Conversational Intelligence, Intelligent Assistants, Articles