Unthread

What it is

Unthread is an AI-powered helpdesk built natively into Slack and Microsoft Teams. It automates ticketing and resolution for IT, HR, customer support, and other operations teams by detecting requests where employees already work, triaging intent, and resolving issues end to end through API and MCP integrations — no portal required. Customers including Intuit, Lemonade, and Automattic are hitting 40% automatic deflection across IT, HR, Legal, and Finance. Founded in 2022 by Tom Bachant and Jake McCloskey through Y Combinator, Unthread was built on a single observation: employees already live in Slack, and any helpdesk that requires them to switch contexts loses the request half the time. The product layers ticketing infrastructure, task assignment, and agentic automation — each layer unlocking the next. The result is an operations stack that reduces coordination overhead without adding process overhead. Enterprise customers can bring their own LLM API key so all AI traffic stays inside their existing vendor agreements. Human-in-the-loop approval gates handle destructive actions like MFA resets and contract reviews, keeping the system safe without sacrificing speed.

Who it's for

IT, HR, and operations teams at mid-market companies (roughly 100 to 2,000 employees) already running on Slack who are drowning in ticket volume, losing requests in chat threads, or spending more time routing work than doing it. Particularly strong for teams migrating off Jira Service Management or Zendesk who want faster deployment, native Slack UX, and AI resolution out of the box.

Why it's better

• •Slack-native intake means requests start where employees already work — no portal friction, no context-switching, no lost tickets buried in threads.
• •40% deflection rate means nearly half of all requests are resolved end to end by AI before a human ever touches them — across IT, HR, Legal, and Finance.
• •MCP and API integrations let the AI actually do things in connected systems — resetting passwords, provisioning access, pulling data — not just suggest what a human should do.
• •BYO-LLM architecture lets enterprise customers plug in their own OpenAI or Anthropic API key so AI traffic stays inside their existing legal agreements rather than flowing through a vendor.
• •Human-in-the-loop approval gates for destructive or sensitive actions mean the system is fast where speed is safe and cautious where it needs to be.