Cassidy

Sweet spot: a 25–250 person company where the docs are mostly in Notion / Drive / Confluence, the support team is drowning in repetitive questions, and there is no platform team to build a custom RAG chatbot. Cassidy's free Starter tier lets you prototype the actual assistant in an afternoon, and the path from prototype to production runs through the same UI rather than a rewrite. Failure modes. Knowledge quality is the hidden tax — Cassidy can only retrieve what is documented, and the more disorganised your Drive is, the more "I don't know" answers you get. AI-credit metering surprises early-stage teams who underestimate how chatty their assistants are. The Workflow builder is great until you need 30+ steps with complex branching, at which point the visual canvas becomes harder to debug than code. Glean overlap is real for buyers above 1,000 employees — at that scale, search quality and governance depth tilt toward Glean. What to pilot. Free Starter, ingest your top 5 most-referenced doc folders, build one Slack assistant for the team that asks the most repetitive questions (usually support or HR). Track deflection rate over two weeks — if 40%+ of questions are answered without a human, it is working. If under 20%, your docs need investment before the AI does.

Cassidy is an AI-automation platform built around a single insight: most useful AI work for an operations or support team is grounded in the company's own knowledge — Notion docs, Google Drive folders, Confluence pages, Gmail threads, support tickets. Cassidy syncs that knowledge continuously, then lets you build assistants and agentic workflows that draft replies, qualify leads, summarise meetings, and run multi-system actions on top of it. The core surface is the assistant builder — pick a knowledge base, write a system prompt, optionally hand the assistant tool access (HubSpot, Salesforce, Zendesk, Gmail, Slack), and deploy via Slack, Microsoft Teams, or a Chrome / Word / Excel / Outlook extension. The Workflows surface adds multi-step automations: triggers, conditional steps, loops, and tool calls — closer to a no-code agent than a Zap. The "Meetings" feature turns conversations into action items wired into the rest of the stack. Compared with Glean — the gold standard for enterprise knowledge search but priced and shaped for 1,000+ employee orgs — Cassidy is the SMB-and-mid-market answer. Compared with ChatGPT Team — generic chatbot with file uploads, no real action layer — Cassidy actually executes inside your tools. The 20,000+ team count, SOC 2 / GDPR / HIPAA / CASA compliance, and the free Starter tier make it easy to try without procurement. For a 10–500 person company that wants knowledge-grounded AI ops without an enterprise-platform price tag, Cassidy is one of the most pragmatic choices in 2026.