Recently, we've been working with a large global services firm in the finance and legal space - helping them automate their workflows and applying LLMs to some of their drafting and management processes. The AI products we're building require detailed prompting flows and interfaces that reflect what the model is thinking. 

Current prompt analysis, versioning, and analytics tools are not designed to empower the most important stakeholders... non-technical domain experts.

Taking a step back for a second, there are promising prompt management and analytics tools in the space such as Langfuse, Latitude, and Braintrust - from what we've gathered they serve as great tools for AI product teams to develop well structured prompts and conduct prompt experiments. However they all seem to be neglecting a vital part of the AI product development process, making it easy for domain experts (the actual end-users of these tools) to understand how the model is responding and how to steer it.

For almost a decade, we’ve advocated for the role of great product design in making AI more accessible to domain experts without technical backgrounds. We lean heavily on our clients and their domain knowledge to help us build meaningful AI products and systems. Therefore, we built a tool that sits as an AI explainability layer on top of any product that displays dynamic AI generated content / decision making.