Try Dust
Overview
Dust builds the layer between a company's scattered internal knowledge and the people who need it. Agents are grounded in your own sources — Notion, Slack, Google Drive, GitHub, Confluence and the rest — so an answer about the refund policy comes from your refund policy rather than from a model's general impression of how refund policies usually work. Sequoia led a $40 million Series B in May 2026, which is a meaningful vote for a European company in a market dominated by American incumbents.
The product's better idea is that agents should be specific. Rather than one company assistant that vaguely knows everything, teams build purpose-built agents — one that answers support questions from the help centre and past tickets, one that helps sales find the right case study, one that onboards engineers into a codebase. Narrow agents with clear sources are dramatically more reliable than broad ones, and Dust's structure pushes teams toward that.
Being European is not incidental. For organisations in the EU where data residency and GDPR are board-level concerns rather than checkboxes, a Paris-headquartered vendor is a materially easier conversation than an American one. That, plus a genuine focus on making agent building accessible to non-engineers, is the position it occupies against Glean and the Microsoft ecosystem.
Key Features
Purpose-Built Agents
Teams create narrow agents for specific jobs with specific sources, which is far more reliable than one assistant expected to know everything.
Broad Knowledge Connectors
Notion, Slack, Google Drive, GitHub, Confluence and more, so agents answer from your actual documentation rather than from model priors.
Accessible to Non-Engineers
Agent building does not require a developer, which is what determines whether adoption spreads beyond the team that bought it.
European Data Residency
A Paris-headquartered vendor, which is a substantially simpler GDPR and data residency conversation for EU organisations.
Model-Agnostic
Works across frontier model providers rather than locking you to one vendor's roadmap.
Source Citation
Answers cite the internal documents they came from, so people can verify rather than trust.
Pros & Cons
Advantages
- Grounding in company knowledge makes answers verifiable
- Narrow purpose-built agents are more reliable than one broad assistant
- European base simplifies GDPR and data residency questions
- Non-engineers can build agents, which drives real adoption
- Sequoia-backed with $40M raised — reasonable continuity confidence
Disadvantages
- Answer quality is capped by how good your internal documentation is
- Smaller than Glean and the Microsoft ecosystem it competes with
- Connector coverage is good but not exhaustive
- Requires someone to own agent curation or agents go stale
Pricing Plans
| Plan | Price | Key Features |
|---|---|---|
| Free Trial | $0 | Evaluation period with core features |
| Pro | From ~$29 / user / month | Agent building, connectors, team workspace |
| Enterprise | Custom | SSO, advanced security, data residency, support |
Best Use Cases
Dust Excels At:
- European organisations with data residency requirements
- Companies whose knowledge is scattered across many tools
- Teams wanting business users to build their own agents
- Onboarding, internal support and knowledge retrieval use cases
May Not Be Ideal For:
- Organisations with poor or non-existent internal documentation
- Companies fully standardised on Microsoft who will use Copilot anyway
- Teams unwilling to maintain agents after launch
How It Compares
Dust vs Glean
Glean is the enterprise search incumbent with over $100 million ARR and deeper large-enterprise features; Dust is more agent-centric, easier for business users to build in, and European. Data residency and team size usually decide it.
Dust vs Microsoft Copilot
If your company runs entirely on Microsoft 365, Copilot's integration is hard to argue with. Dust wins on multi-tool environments, agent specificity and EU data residency.
Final Verdict
Our Recommendation
Dust is the strongest European answer in the enterprise agent category, and the Sequoia-led round says the market agrees. The design instinct — many narrow agents with clear sources rather than one omniscient assistant — is correct, and it is the single biggest predictor of whether internal AI actually gets used past month two. Two honest caveats: the agents are only as good as the documentation behind them, so this rewards companies that already write things down, and someone has to own keeping agents current. For EU organisations with data residency constraints, it should be on the shortlist.