The enterprise market for AI agents is on track to become one of the largest software categories of the next decade — analysts at McKinsey estimate that generative AI could add $2.6 to $4.4 trillion in annual value across industries, and the agent layer is where much of that value will be captured. The frontier labs have produced models that can reason, write, plan, and code at a level that would have seemed impossible three years ago. And yet, walk into any large enterprise today and the distance between what these models can do in a research demo and how employees are actually using them remains enormous.
That gap is where the value of this cycle gets built. Model capability is racing ahead; deployment inside real companies is not. Closing that distance is one of the hardest product, infrastructure, and change-management problems in software today, and the company that solves it — as Dust AI is doing — will be one of the defining businesses of the decade.
Dust closes the gap between models and the org chart
Dust is the platform for building, deploying, and running enterprise AI agents inside the organization. Employees can spin up custom agents grounded in their own company's data and tools, share them with colleagues, and compose them into workflows that get real work done. It is the missing layer between the model and the org chart.
What makes Dust different from the start is that it is multiplayer from the ground up. Agents are shared objects — you build them, share them, fork them, and improve them with your colleagues, not in isolation. That distinction matters more than it might seem: most enterprise AI tools today treat each user as an island, so the organization never accumulates collective intelligence. The platform connects to more than 100 data sources including Slack, Salesforce, Notion, Google Drive, and GitHub, so agents can draw on the full breadth of an organization's knowledge rather than a single user's context.
A founding team built for this exact problem
I have been closely following Dust since their seed, and few teams have impressed me more along the way. Co-founders Gabriel Hubert (CEO) and Stan Polu (Chief Technology Officer) are second-time founders who have been building together for over a decade. They first met at Stanford in 2007 and founded TOTEMS, a data analytics company that was acquired by Stripe in 2015. Both stayed there for years and became standouts inside one of the most demanding engineering cultures of the modern internet.
Stan went on to become a research engineer at OpenAI on the reasoning team, working on mathematical reasoning in large language models. Gabe went on to lead product at Alan, the European healthtech unicorn. Few founding pairs combine that depth of AI research with that level of product and operating craft, and even fewer have the muscle memory of a long working relationship to draw on. Gabe and Stan do.
The combination matters: enterprise AI agents require both frontier-model fluency and a deep understanding of how real organizations adopt software, and this team has earned both through a decade and a half of building together. When I evaluate founding teams, I look for evidence that they can ship fast, hold a high quality bar under pressure, and earn the trust of sophisticated buyers. Gabe and Stan check every box.
Taste, traction, and compounding product-market fit
What has always set Dust apart, beyond the team, is the taste. The product is multiplayer from the ground up: agents are objects you build, share, fork, and improve with your colleagues, not single-player toys. Quality compounds with usage, because every interaction adds context, connections, and judgment to the system; the more an organization uses Dust, the more uniquely valuable Dust becomes to that organization. And the product is genuinely, deeply usable, which sounds obvious until you sit through a procession of enterprise AI demos and realize how rare it is. Gabe and Stan obsess over the small details in a way that reminds me of the best consumer product teams, and it shows on every screen.
At Wing, we believe the next decade of enterprise software will be defined by the platforms that take general-purpose intelligence and turn it into specific, accountable work inside real companies. Dust is doing exactly that, with a level of conviction, craftsmanship, and customer love that is rare at any stage. We are proud to back Gabe, Stan, and the entire Dust team on their Series B.
If you want to join one of the most exciting teams in enterprise AI, check out Dust’s careers page.
