Multiplayer AI for the whole company
HQ: Paris, France / San Francisco, CA, United States | Founded: 2023 | Employees: ~50–100 (est.) | Stage: Series B — $40M co-led by Abstract and Sequoia (May 2026); $60M+ raised total | Website: https://dust.tt
Dust helps companies build their own AI workforce. The platform connects to a company's internal knowledge and tools — Notion, Slack, Google Drive, GitHub, databases, support desks — and lets teams compose custom assistants and multi-step agents on top: a support agent that drafts replies from past tickets, a sales agent that preps account briefs, an ops agent that reconciles spreadsheets. The 'multiplayer' part is the point: assistants are shared, forked and improved by the whole company instead of living in one person's chat window.
The company was founded in early 2023 by Gabriel Hubert and Stanislas Polu, who had worked together at Stripe (Polu also spent three years at OpenAI). From day one they designed for a multi-model world — wrap every frontier model, let customers pick and switch — on the conviction that no single model would win and the durable layer would be the one holding company context.
Three years on, Dust reports 51,000 workers using it across more than 3,000 companies, with the May 2026 Series B ($40M co-led by Abstract and Sequoia) funding an enterprise push: governance, security review and the integrations that make agents genuinely useful at work.
Dust started from a contrarian premise in 2023: nobody wants another chatbot — companies want agents that know their business and that the whole team can build and share. Gabriel Hubert and Stanislas Polu (both ex-Stripe; Polu also spent years at OpenAI) built a platform where a company connects its documents, tickets, Slack and databases once, and then anyone can assemble assistants on top of that context — what the founders call 'multiplayer AI'.
The bet that mattered early was model-agnosticism. Dust launched insisting no single model would win, so it wrapped every frontier model and let customers switch. That looked risky when OpenAI dominated the conversation; it looks prescient now that enterprises run different models for different tasks and want an orchestration layer above them all.
By 2026 the numbers validated the shape: 51,000 workers at more than 3,000 companies building and running agents — unusual breadth for enterprise AI — and a $40M Series B in May 2026 co-led by Abstract and Sequoia Capital, taking total funding past $60M.
The hard part is ahead. Dust competes in the most contested layer of enterprise AI: Microsoft and Google bundle copilots into agreements, Glean owns 'answers from company knowledge', and a dozen well-funded startups (Lindy, Sierra, Decagon, CrewAI) attack adjacent slices. Dust's edge is its builder culture — the product treats every employee as a potential agent developer — and its independence from any single model vendor. The question is whether that stays a wedge or becomes a feature.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki