With €34 million, DUST accelerates its push into multi-agent systems for enterprises.

Generative AI has spread across enterprises at a pace rarely seen in the history of software. Yet despite the rise of copilots and conversational assistants, the organization of work itself remains largely unchanged. Most AI usage is still individual, confined to private interfaces with little collective continuity. That is precisely the limitation Dust aims to overcome.

The company, which specializes in agentic AI systems, has raised $40 million, approximately €34 million, in a Series B round led by Abstract and Sequoia, with participation from Snowflake and Datadog. Since its inception, Dust has raised more than $60 million.

Dust is advancing a thesis that is becoming increasingly widespread across the AI ecosystem: models themselves are no longer the primary technological bottleneck. The real challenge now lies in enabling companies to coordinate collaboration between humans, agents, data, and workflows.

“With where the technology is right now and the capabilities of agents, humans are no longer the primary source of knowledge work,” explains Stanislas Polu, Dust co-founder and former OpenAI researcher. “So far, that has worked relatively well for individuals using AI assistants equipped with tools, skills, and memory. That’s what we call single-player AI.”

For Dust, this first phase of generative AI improves individual productivity without fundamentally transforming the organization itself. “The problem with single-player AI is that it doesn’t compound across the team. The agent I delegate work to doesn’t have the full picture of what’s happening across the company.”

This analysis directly targets current enterprise AI usage patterns. A salesperson uses an agent to prepare for a customer meeting, but the solutions engineer joining the next day starts from scratch. One marketing team generates a presentation with a copilot, while another produces content from a completely different version of the context. Productivity gains exist, but they remain fragmented.

According to Dust, coordination has now become the core bottleneck.

“As more humans delegate more work to more agents, humans and agents must become collaborators operating inside a shared workspace with shared context, artifacts, and goals, enabling seamless human-human, human-agent, and agent-agent collaboration.”

The company describes this approach as “multiplayer AI.” The objective is to build systems where humans and AI agents operate together inside the same operational environments, with shared context, aligned objectives, and collective visibility across workflows.

“The most complex work happening inside companies today, involving teams working across days and weeks, is never done by a single person. It involves multiple teams operating with different versions of context. That’s multiplayer AI. That’s Dust.”

Dust’s platform is therefore built around persistent shared workspaces where employees and AI agents operate simultaneously around the same projects, conversations, tasks, documents, and notifications. Unlike traditional conversational assistants, agents no longer function inside isolated sessions but within continuous workflows directly integrated into the organization of work.

Dust also places significant emphasis on its contextual intelligence layer. The startup argues that simply connecting a model to enterprise tools is no longer enough. Systems must now be capable of understanding the information circulating throughout the organization, synthesizing it, and acting on top of that context.

The platform connects more than one hundred enterprise data sources and software tools, enabling agents to interact with environments such as Slack, HubSpot, Notion, Gmail, Google Drive, and Snowflake.

Dust is also developing persistent memory systems and continuous improvement loops designed to progressively evolve agents based on real-world team usage. The objective is to transform AI into an organizational infrastructure capable of learning internal practices and improving workflows over time.

Governance represents another core pillar of the platform. Dust highlights granular permission systems, cost and usage monitoring, full audit trails, and analytics tools allowing enterprises to track agent activity with precision. The company also states that it is SOC 2 Type II certified, GDPR compliant, and contractually guarantees that customer data is never used to train models.

Dust now claims more than 3,000 customer organizations and over 300,000 deployed agents across its platform. The company also reports a 70% weekly active usage rate and zero churn since the beginning of 2025.

At Doctolib, Dust is being used as part of a company-wide AI strategy involving 3,000 employees. At Persona, eleven departments have deployed more than 300 AI agents. The company also cites deployments at Clay and Profound focused on go-to-market operations and customer intelligence management.

Dust also says it uses its own platform internally to coordinate a significant portion of its operations. To orchestrate the announcement of this funding round, the company’s marketing, operations, and data teams collaborated alongside multiple AI agents working simultaneously from the same datasets and tools. Agents synthesized information from Slack, Snowflake, HubSpot, Gmail, and Google Drive to generate content, coordinate approvals, and manage launch workflows.

The founders’ backgrounds also reinforce the project’s technical credibility. Gabriel Hubert and Stanislas Polu have worked together since meeting at Stanford University in 2007. They previously co-founded TOTEMS, a data analytics company acquired by Stripe in 2014.

After several years at Stripe, Polu joined OpenAI as a research engineer within the team led by Greg Brockman, where he worked on reasoning capabilities alongside Ilya Sutskever.

According to Dust, the company was founded on the conviction that models had already become powerful enough to economically transform organizations, but that the product layer required to integrate them into day-to-day operations still needed to be built.

“We are the multiplayer AI system for human-agent collaboration, empowering AI Operators to rewire how their organizations work,” the company says today.

This vision now appears to be gaining traction among a growing number of Silicon Valley investors. “There has never been a better moment to join the movement of AI Operators rewiring how companies operate,” Dust argues.

Beyond the funding round itself, the company illustrates a broader transformation underway in enterprise software. After ERP systems and SaaS platforms, a new generation of companies is now attempting to build hybrid operational environments where humans and AI agents work together continuously inside the same systems.