Back to the newsroom

Insights

Our in-house LLM Wiki and AI workflows

Wicked Storm is gathering in-house knowledge that used to be scattered into an LLM Wiki. The goal is to keep documents, discussions, and answers to recurring questions in one place, so they are ready to use whenever they are needed.

Concept image: scattered documents, tables, and conversations on the left are organized in the middle into a knowledge graph (ontology) of concepts and relationships, leading on the right to a set of wiki pages and an AI that hands over answers
Scattered documents and discussions are woven into concepts and relationships (an ontology) and gathered in the wiki, where AI finds materials and creates drafts (concept image made with AI).

Why gather it in one place

When many projects run at the same time, the same questions come up again and again, and the answers tend to scatter across individual documents and conversations. The more time it takes to find the knowledge we need, the less time we have for new work. The LLM Wiki is our effort to gather scattered documents and discussions in one place so that anyone can ask for and find the knowledge they need.

Building AI into our workflows

On top of this, we are building AI into our workflows to speed up the process of finding materials, organizing them, and creating drafts. AI finds and gathers relevant materials and creates a draft; the person in charge then reviews, edits, and finalizes the content.

The same principle as our products

This approach follows the same principle Wicked Storm keeps in its products. In Lecognizer and LearnHubble AI, AI finds and analyzes signals and suggests drafts, while administrators and instructors judge and apply. In our own work, too, AI takes on finding, organizing, and drafting, and people decide what to use and how to decide.

We are going beyond adopting tools and continuing to experiment with refining the way we work itself. We will keep sharing what we learn along the way on this board.

Back to the newsroom