Thena Enhances AI Capabilities with OpenAI Model and New Endpoints

December 21, 2023

We're thrilled to announce the successful completion of a major update to our AI technology stack here at Thena. Our latest enhancements involve the seamless integration of the most cutting-edge OpenAI models into both our Aggressive and Conservative Strategies, making our AI capabilities more powerful and reliable than ever.

Technical Advancements

Necessity for Upgrade

With the impending deprecation of the gpt-3 based models (text-davinci-003) and Completion API endpoint by OpenAI and Azure OpenAI on January 4, 2024, we recognized the urgency of upgrading our systems. Key components of our existing frameworks, including Request Detection, AI Metadata Generation, AI Tags Matching, Custom Field Assignment, and User Tag Assignment, were dependent on these models slated for deprecation.

Strategic Migration Solution

To address this challenge, we successfully transitioned to more advanced and potent models, specifically gpt-3.5-turbo-16k and gpt-4-32k. These models not only offer extended context lengths but also boast superior response capabilities. Furthermore, we integrated the newly released Chat Completion API endpoint to elevate our AI-driven workflows to new heights.

Robust Fallback Mechanism

Recognizing the importance of uninterrupted service, we implemented a comprehensive 3-step fallback protocol for each AI workflow. In case of a disruption in OpenAI's service, our system will automatically reroute via Portkey and fallback to Azure OpenAI's models. If both primary services are unavailable, a seamless switch to the Bayes Classifier ensures continuous operation, reaffirming our commitment to reliability.

Operational Impact

Enhanced Request Detection

These updates have significantly bolstered our Request Detection methodologies, reducing the likelihood of downtime even in scenarios where OpenAI APIs are temporarily unavailable. Our system's ability to adapt and reroute ensures uninterrupted service for our users.

Minimal Trade-offs

While the implementation of the new models and fallback mechanisms has resulted in a slight increase in false positives within the Conservative Strategy (estimated at 1-2%), we believe this is a manageable trade-off considering the remarkable improvements in the GPT models operating in the background.

Comparative Accuracy

Notably, our Conservative Strategy continues to outperform the Aggressive Strategy in accurately identifying requests. This underscores our commitment to delivering AI solutions that strike the right balance between precision and responsiveness.

In conclusion, these enhancements represent a significant leap forward in the capabilities of Thena's AI technology stack. We're confident that these upgrades will not only ensure the uninterrupted delivery of our services but also set the stage for even more sophisticated and reliable AI solutions in the future. Thank you for your continued support, and we look forward to bringing you more exciting developments in the world of AI at Thena.

Acquire, Service & Grow customers in Slack
Share this blog :

Start managing your customers faster and easier from Slack.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.