What happens when you bolt AI onto tools that were designed for human-to-human chat? According to Hasura, you get fragmentation, context loss, and inaccurate results. Their answer is PromptQL, an AI-native workspace that aims to replace the "Slack plus Claude plus Glean plus custom agents" stack that many teams are running today.
The argument is simple: communication tools like Slack were built for humans passing messages to each other. When you layer AI assistants, search tools, and agents on top of that foundation, every tool operates in its own silo. Context doesn't flow between them. The AI in your search tool doesn't know what your AI coding assistant just built, and neither knows what was decided in your Slack thread yesterday.
PromptQL's alternative has three pieces: a multiplayer AI surface that works across web, desktop, and mobile; an AI-assisted wiki for shared knowledge; and a permissions system using row-level security and attribute-based access control to manage who (and what AI) can see what data.
The most interesting technical choice: rather than using RAG (retrieval-augmented generation, where AI searches through documents to find answers) or chaining together tool calls, PromptQL has the AI write code to solve problems directly. Hasura's bet is that code generation produces more reliable results than having an AI summarize search results.
There are no published accuracy benchmarks comparing PromptQL against the Slack-plus-tools approach, so the "killing Slack" claim is more vision statement than proven result. The product is live but early. For teams frustrated by juggling five different AI-adjacent tools that don't talk to each other, the pitch makes intuitive sense. Whether a single AI workspace can actually replace the flexibility of choosing best-in-class tools for each job is the harder question.