Dust Deep Dive — AI Agents Built for Teams

Opening

In the AI Agent space, founder backgrounds tell you a lot. Dust has two co-founders: Stanislas Polu, a former OpenAI researcher who spent three years at OpenAI working on mathematical reasoning and watched GPT-2 produce its first coherent paragraph; and Gabriel Hubert, former CPO at Stripe (via TOTEMS, which Stripe acquired) and Alan (one of Europe's fastest-growing insurtech companies). Both are Ecole Polytechnique + Stanford alumni. Sequoia Capital led both rounds, total funding sits at $21.5M, and a 66-person team is generating $7.3M ARR — that's $110K in annual recurring revenue per employee. I used Dust extensively while researching enterprise AI collaboration tools, and I've discussed how it stacks up against Microsoft Copilot with clients operating in European markets.

The Problem They Solve

Most AI Agent tools serve either individual users (personal productivity) or developers (building Agent systems). Dust is targeting a different scenario entirely: AI Agents for team collaboration.

Picture a 30-person customer success team processing hundreds of customer emails daily, needing to cross-reference internal documentation, historical tickets, and product release notes to craft accurate replies. Or a 50-person sales team producing weekly market analyses, competitive comparisons, and client proposals, pulling together CRM data, past case studies, and public information. The common thread: AI Agents need deep access to internal company knowledge, and they need to be shared and managed at the team level.

The core problem Dust solves: give teams a shared set of AI Agents connected to company knowledge bases that every member can use while admins maintain centralized control. Not giving individuals a ChatGPT Plus subscription, but giving the team a cluster of AI assistants that "know everything about the company."

Product Matrix

Core Products

Custom AI Agents: Dust's core offering is custom Agents. Each Agent has a defined role (customer support assistant, sales assistant, technical documentation assistant, etc.) and connects to specific knowledge sources and tools. Team members interact with Agents through a conversational interface, and Agents respond based on company data.

Company Knowledge Hub: Dust connects to a company's existing tools and data sources — Slack, Notion, Google Drive, GitHub, Intercom, and more. Data syncs and indexes automatically, allowing Agents to retrieve the latest information in real time. This resembles Glean's enterprise search, but Dust's emphasis is on "Agents using this data to execute tasks" rather than just "search."

Workspace Management: A team-level Agent management platform. Admins control who can create Agents, which data sources Agents can access, and Agent behavioral guardrails. Supports multiple workspaces so different departments can have independent Agent collections and data permissions.

Model Routing: Dust isn't locked to a single model — it supports GPT-4o, Claude, Gemini, Mistral, and more, with the ability to choose the best model per Agent. This model routing capability lets teams optimize for cost and quality based on task type.

Technical Differentiation

Dust differentiates on three levels. First, "team-first" product design — Agents aren't personal tools but shared "AI colleagues" for the team. Second, deep data connections with real-time sync — Agent responses are based on the company's latest data, not static knowledge bases. Third, a model-agnostic architecture that isn't locked to OpenAI or any single provider.

Stanislas Polu's OpenAI background adds direct value in prompt engineering and Agent behavior control — Dust's Agents outperform most comparable products in "hallucination control" and "citation accuracy."

Business Model

Pricing Strategy

Plan	Price	Target Customer
Free Trial	15 days free	Evaluation
Pro	€29/user/month	Small/mid teams
Enterprise	Custom	100+ users, multiple workspaces, SSO

Per-user, per-month pricing — simple and transparent. Compared to credit-based pricing (Relevance AI, Gumloop), Dust's costs are more predictable: no matter how many times the team uses an Agent, the monthly fee stays fixed.

Revenue Model

Pure SaaS per-seat pricing. With $7.3M ARR and an estimated average price of €29–40/user/month, this implies roughly 15,000–20,000 paying users. Growth flywheel: one department trials → other departments see results and want in → company-wide adoption → annual Enterprise contract.

Funding & Valuation

Round	Date	Amount	Lead
Seed	2023	€5M	Sequoia Capital
Series A	June 2024	$16M	Sequoia Capital

Total funding: $21.5M. Sequoia leading both rounds signals strong conviction in the team and direction. Valuation was approximately $100M+ at Series A. At $7.3M ARR, the current implied valuation is likely in the $100–150M range — a standout performance among European AI startups.

Customers & Market

Marquee Customers

Dust's customers are primarily European and American tech companies and growth-stage enterprises. Based on Sequoia's case analysis, the typical customer is a 50–500 person organization using a heavy SaaS stack (Slack + Notion + Google Workspace) with clear knowledge management pain points. Customer retention and expansion data are Dust's strengths — once a team connects its internal knowledge base, switching costs are high.

Market Size

The team collaboration AI market heavily overlaps with the enterprise search market. Glean (valued at $4.6B) has proven the market's scale. Dust's positioning is more focused than Glean's (Agents rather than search), more affordable (€29/user vs. Glean's $10–20/user, which requires large-scale deployment), and aims to reach the SMBs and teams that Glean can't serve.

Competitive Landscape

Dimension	Dust	Glean	Microsoft Copilot	Notion AI
Core Positioning	Team AI Agents	Enterprise AI Search	Microsoft Ecosystem AI Assistant	Knowledge Base AI
Target Customer	50–500 person teams	500+ person enterprises	Microsoft ecosystem enterprises	Notion users
Pricing	€29/user/month	$10–20/user/month	$30/user/month	$10/user/month
Custom Agents	Core feature	Limited	Copilot Studio	No
Data Connections	Broad (Slack/Notion/GitHub)	100+ enterprise apps	Primarily Microsoft ecosystem	Notion-internal
Model Selection	Multi-model	Proprietary + partnerships	GPT-4/OpenAI	Opaque

Dust occupies a unique position: lighter than Glean, more powerful than Notion AI, more flexible than Microsoft Copilot, and more deeply connected to company data than ChatGPT Team.

What I Actually Saw

The Good: Agent data connection quality is excellent. During testing, I connected a Notion workspace and several Google Drive folders. The Agent accurately cited specific documents and pages when answering questions — it could even note that "this information changed in the version updated last Wednesday." The model routing feature is practical too — use Gemini Flash for simple queries to cut costs, switch to Claude for complex reasoning. The admin dashboard is clear, showing which Agents are used most and which data sources are cited most frequently.

The Complicated: At €29/user/month, a 10-person team pays €290/month — not cheap. Scale that to 200 people company-wide, and you're looking at €5,800/month — at which point enterprises start wondering whether they should just use Microsoft Copilot (which is already bundled with their Microsoft subscription). Additionally, Dust's Agent capabilities currently lean toward "knowledge retrieval and Q&A," falling behind Relevance AI and n8n in "action execution" (sending emails, updating CRMs, creating tickets).

The Reality: A 66-person team, $7.3M ARR, Sequoia leading both rounds — this is a healthy early-growth profile. $110K ARR per employee is strong by SaaS standards. But Dust's biggest challenge is the Microsoft Copilot squeeze — Microsoft can bundle similar capabilities into Microsoft 365 subscriptions at $30/user/month, which includes Word, Excel, Teams, and the full suite. Dust needs to maintain a meaningful lead in "deep data connections" and "custom Agent capabilities."

My Verdict

Dust is the best "team-level AI Agent" product I've seen. The founders have exceptional pedigrees, Sequoia has backed them twice, and the unit economics are healthy. Its core value proposition — "turning company knowledge into Agent capability" — has strong PMF in knowledge-intensive teams. But pricing strategy and competition from Microsoft Copilot are two issues that need addressing.

✅ Good fit for: Knowledge-intensive teams of 50–500 people (customer success, sales, consulting); companies heavily using Slack + Notion + Google Workspace; scenarios requiring shared team AI Agents with deep data connections

❌ Skip if: You're already in the Microsoft ecosystem and satisfied with Copilot (high switching costs); you need Agents that execute actions (sending emails, updating CRMs, etc.) rather than primarily doing knowledge Q&A (look at n8n or Relevance AI); you're a team of fewer than 10 (the cost-benefit ratio doesn't pencil out)

Bottom line: Dust proves that "AI Agents aren't just developer tools — they're team infrastructure." The question is whether it can carve out its own space in the shadow of Microsoft Copilot.

Discussion

What AI collaboration tools is your team using? Dust, Glean, or Microsoft Copilot — which do you think has the most promise? Is "team-level AI Agents" a must-have for you? Let's discuss in the comments.