In this episode, "Minimalism to optimize AI performance," Cisco technical writers Ravi T R and Deepak Ruhil discuss the intersection of artificial intelligence and technical documentation. Deepak notes that because AI can generate infinite amounts of content, the human skill of minimalism is highly valuable for filtering noise and reducing cognitive load for users.
The guests explain how strict structural frameworks like DITA (Darwin Information Typing Architecture) are used give guardrails AI models and prevent hallucinations. By assigning metadata and maintaining a single user intent per topic, writers can give LLMs exactly the right amount of context needed to deliver accurate answers. Deepak suggests that this modular, metadata-rich documentation will eventually replace traditional user guides and lead to the "death of the search bar," as users increasingly rely on chatbots to find specific steps.
The conversation also covers stigmergy, comparing specialized AI agents to ant colonies. Instead of relying on a centralized manager, AI agents can accomplish complex tasks autonomously if technical writers leave clear, well-structured information nuggets (similar to pheromone trails) for them to follow.
Finally, the episode offers practical frameworks for optimizing AI. Deepak shares guidelines from his working AI style guide, such as writing explicit intent and front-loading important information for AI token reading. Meanwhile, Ravi explains how DITA data maps can be exported as JSON objects to create knowledge graphs. These knowledge graphs give AI agents the deterministic context they need to navigate systems and assist users accurately.
Bio
Ravi Tarikere Ramesh
Ravi is a Senior Information Developer at Cisco with over 10 years of experience in information development. He holds an M.Tech in Software Engineering and a postgraduate degree in Artificial Intelligence and Machine Learning.
Deepak Ruhil
Deepak is a technical writer at Cisco who brings nearly 15 years of experience specializing in DITA-based documentation and structured authoring. He has worked across domains including software, security, networking, and aviation, delivering clear, user-focused content for global organizations.