The pitfalls of implementing AI from a technical writer's perspective

Generative models have become a powerful tool for technical writers, enabling them to translate technical documentation and improve text quality. However, when it comes to implementing AI within a company, the process is not without its challenges. A myriad of issues arises, particularly concerning the accuracy of the text used for training. The situation is exacerbated when the information needed to create the text is poorly structured, partially missing, or scattered across inaccessible sources.

In my presentation, I will analyse a real case of our company and share the main pains of a technical writer when implementing generative models and methods of overcoming bottlenecks of this process. I will delve into the intricacies of AI through chatbots, the most common medium for communication with both internal and external clients. This approach not only provides the broadest range of problematic scenarios but also offers rich insights for improving the model training process.