Dust

What it is

Dust is an AI/automation tool designed for "multiplayer AI" that facilitates human-agent collaboration. It connects company knowledge, existing tools, and teams to enable the creation, sharing, and running of AI agents within workflows. The platform allows users to utilize different AI models for various tasks, maintaining human oversight throughout the process.

Best for

Dust is positioned for "AI Operators" and teams within organizations who actively build, deploy, and manage AI tools. It aims to support various departments, including engineering, marketing, support, product, and sales, by integrating AI into their operational processes and fostering collaboration between people and AI agents.

Key strengths

• Comprehensive Context Integration: Dust connects to over 70 connectors (e.g., Slack, Notion, Google Drive, Salesforce, Zendesk) to provide AI agents with relevant company knowledge and data.
• Model Flexibility: Users can build agents using more than 20 frontier and open-source models from providers such as OpenAI, Anthropic, Google, and Mistral, selecting the most suitable model for each workflow.
• Multiplayer AI for Collaboration: The platform brings together people and AI agents in shared workflows, enabling coordination across departments, shared context, reusable skills, and human review for complex tasks.
• Enterprise-Ready Governance & Security: Dust offers AI-native governance with a dual-layer permission model, SCIM-synced groups, audit logs, and security features like SOC 2 Type II certification and data residency options.

Watch-outs

• The source does not specify the types of "tools" or "company systems" beyond general categories and a few examples (Snowflake, HubSpot, Zendesk, Salesforce, Slack, Notion, Google Drive, GitHub), making it unclear which specific applications are supported.
• No details are provided regarding the learning curve for building and adapting agents, or the technical expertise required for users to become effective "AI Operators."
• The text does not address the process for integrating custom or proprietary internal tools and data sources that may not be covered by the 70+ mentioned connectors.