Dust

Dust is the enterprise AI assistant platform built on a conviction that has aged well: your company's AI should know your company. The product connects to the systems where organizational knowledge lives — Slack, Google Drive, Notion, GitHub, Linear, Jira, and dozens more — and builds continuously updated indexes that every assistant and workflow can ground against. Ask 'what did we decide about the pricing change?' and Dust answers from the actual Slack threads and docs where it happened, with citations to sources — replacing the archaeology of enterprise search with conversation.

The platform's depth is in building on that foundation: custom assistants (with model choice across OpenAI, Anthropic, Mistral) tuned to specific jobs — support triage, sales-call prep, engineering onboarding — plus agent workflows that connect systems programmatically, like syncing CRM entries from call transcripts. The admin surface takes enterprise security seriously: granular permissions that inherit source-system access, data never used for training, and audit trails that satisfy the reviewers who block most AI rollouts. In our deployment review, the permission mirroring — assistants only see what the asking user could see — was implemented more rigorously than competitors.

Honest limits: index quality depends on source hygiene — messy wikis produce confident answers citing chaos; pricing (~$29+/user/month, sales-led) targets companies past startup stage; and building genuinely useful custom assistants takes iteration. For mid-size and larger organizations ready to make company knowledge conversational under real governance, Dust is among the most thoughtful platforms in a crowded category.