We recently implemented our first batch of AI-enabled features a few months back. This venture into generative AI was a learning experience for us at incident.io, as we navigated uncharted territory. In this post, I will share some insights we gained while utilizing LLMs to enhance our products. These insights are particularly relevant for companies operating at the application layer, integrating AI capabilities into their offerings.
One crucial lesson we learned early on was the importance of investing in our team upfront. With the vast possibilities AI presents, it can be overwhelming to know where to begin. To address this, we conducted several short experiments to explore the technical capabilities available. By iterating quickly and utilizing tools like Jupyter Notebooks, we gained a better understanding of what was feasible before committing to full-scale development.
Our experimentation involved various techniques such as RAG, embeddings, multi-shot prompts, code generation, and function calling. This exploration helped us identify the most compelling applications for AI-powered features. The initial release of four key features allowed us to cover a wide range of functionalities and gather valuable insights on user preferences.
Investing in robust tools and developer experience was just as crucial as nurturing our team’s skills. For instance, we developed a CLI tool to test OpenAI models with different incident scenarios, enabling us to refine our outputs efficiently. This emphasis on tooling streamlined our development process and accelerated our iterations, leading to high adoption rates among customers.
Maintaining a principled approach to product development was another key takeaway. While AI presents exciting possibilities, it’s essential to prioritize building products that resonate with customers. We discovered that incorporating AI into existing workflows and providing automation tools, rather than standalone chatbots, yielded better user engagement and retention.
The rapid evolution of foundation models and developer tools underscores the need to focus on the bigger picture and iterate swiftly. Embracing platform updates and leveraging tools like clustering and LLMs during development can enhance product capabilities and drive innovation.
Launching AI-enabled features in phases and closely monitoring customer feedback proved invaluable in refining our offerings. By addressing unexpected challenges, such as language support issues, early on, we were able to enhance the user experience and maintain control over product quality.
In conclusion, the potential of AI to transform products and services is immense, and companies should not hesitate to embrace this technology. Despite the ever-changing landscape of AI, starting now can position businesses as industry leaders and pioneers in user-centric innovation.
For more insights on practical AI applications and future updates, stay connected with us by subscribing to our newsletter. Your journey into the world of AI-driven solutions awaits!
