Generative artificial intelligence has spread across businesses at a speed rarely seen in the history of software. However, despite the proliferation of co-pilots and conversational assistants, the organization of work still remains largely unchanged. Uses remain predominantly individual, compartmentalized in private interfaces, without real collective continuity. It is precisely this limit that Dust seeks to overcome.
The company specializing in agentic systems has just raised $40 million, or around €34 million, in a Series B round led by Abstract and Sequoia, with the participation of Snowflake and Datadog. Since its inception, the company has raised more than $60 million.
Dust defends a thesis that is increasingly widespread in the AI ecosystem: models no longer constitute the main technological obstacle. The real challenge now lies in the ability of companies to organize collaboration between humans, agents, data and workflows.
“With the current level of technology and the capabilities of agents, humans are no longer the main source of cognitive work,” explains Stanislas Polu, co-founder of Dust and former employee of OpenAI. “So far, it works relatively well for individuals using AI assistants with tools, skills and memory. This is what we call “single-player AI”. »
For Dust, this first phase of generative AI improves individual productivity without truly transforming the organization. “The problem with single-player AI is that it doesn’t produce a cumulative team-wide effect. The agent to whom I delegate work does not have a complete view of what is happening in the company. »
This analysis directly targets the current uses of AI assistants in business. A salesperson uses an agent to prepare for a customer meeting, but the pre-sales engineer who comes in afterwards starts from scratch. One marketing team generates a presentation with a co-pilot, while another produces content from a different context. The gains exist, but they remain fragmented.
According to Dust, the main obstacle now becomes that of coordination. “As more humans delegate more work to more agents, humans and agents must become collaborators working in a shared space with common context, artifacts, and goals to enable seamless collaborations between humans, between humans and agents, and between agents themselves. »
The company presents this approach as “multiplayer AI”. The goal is to build systems where humans and AI agents work together in the same operational spaces with a shared context, common goals and collective visibility into workflows.
“The most complex jobs carried out today in companies, involving teams over several days or weeks, are never carried out by a single person. They involve multiple teams with different versions of the context. This is AI multiplayer. That’s Dust. »
The platform developed by Dust is therefore based on persistent workspaces in which collaborators and AI agents operate simultaneously around the same projects, conversations, tasks, documents and notifications. Unlike traditional conversational assistants, agents no longer operate in isolated sessions but in continuous workflows integrated directly into the work organization.
Dust also strongly highlights its contextual layer; the startup considers that connecting a model to business tools is no longer enough. Systems must now be able to understand the information circulating within the company, synthesize it and act from this context.
The platform thus connects more than a hundred data sources and business tools to allow agents to interact with environments such as Slack, HubSpot, Notion, Gmail, Google Drive or Snowflake.
Dust is also developing persistent memory mechanisms and continuous improvement loops intended to gradually evolve agents based on real team usage. The goal is to transform AI into organizational infrastructure that can learn from internal practices and improve workflows over time.
Governance constitutes another central axis of the product and Dust highlights granular permissions, cost and usage monitoring systems, complete audit logs as well as analytical tools allowing the precise monitoring of the activity of agents deployed in the company. The company also specifies that it is SOC 2 Type II certified, GDPR compliant and guarantees that no customer data is used to train the models.
Dust now claims more than 3,000 customer organizations and more than 300,000 agents deployed on its platform. The company also claims to have recorded a weekly active usage rate of 70% as well as zero churn since the start of 2025.
At Doctolib, Dust is used as part of an AI strategy deployed to 3,000 employees. At Persona, eleven departments have deployed more than 300 AI agents. The company also mentions uses at Clay or Profound around commercial operations and customer knowledge management.
Dust also claims to use its own platform to coordinate a significant part of its internal operations. To orchestrate the announcement of this fundraising, the marketing, operations and data teams collaborated with several AI agents working simultaneously from the same data and tools. The agents notably synthesized information from Slack, Snowflake, HubSpot, Gmail or Google Drive in order to generate content, coordinate validations and monitor launch workflows.
The profile of the founders also contributes to the technical credibility of the project. Gabriel Hubert and Stanislas Polu have worked together since they met at Stanford University in 2007.
After several years at Stripe, Stanislas Polu joined OpenAI as a research engineer within the team led by Greg Brockman. He notably participates in research work on the reasoning capabilities of models with Ilya Sutskever.
According to Dust, the company’s creation was based on the belief that the models were already powerful enough to economically transform organizations, but the product layer to actually integrate them into daily operations had yet to be built.
“We are the AI multiplayer system dedicated to collaboration between humans and agents, allowing AI Operators to reinvent the way their organizations operate,” the company says today.
This vision now seems to convince a growing portion of Silicon Valley investors. “There has never been a better time to join the movement of AI Operators who are reorganizing the way their businesses operate,” said Dust.
Beyond this fundraising, the company illustrates a more profound change in business software. After ERPs and then SaaS platforms, a new generation of players is now seeking to build hybrid environments where humans and AI agents work together continuously in the same operational systems.
