Use Cases

A short tour of agent shapes that map well to real business problems. Each pattern links to the building blocks you'd use to assemble it.

Customer support copilot

What it does. Answers customer questions in your product or in Slack/WhatsApp, escalates to humans when it's unsure, logs every conversation against the customer record.

Building blocks

• A Knowledge Base of help-center articles and product docs (Assets)
• The Chat integration for multi-turn state
• A Slack or WhatsApp connector (Connectors)
• A Salesforce or HubSpot integration to pull customer context and write notes
• A sub-agent for escalation triage that routes hard cases to a human

Sales research & enrichment

What it does. Given a company name or domain, returns a structured profile: people, recent news, tech stack, signals worth following up on.

Building blocks

• Tavily or BrightData for live web search
• Apollo / G2 integrations for firmographic data
• An Agent Step with structured output to shape the result into your CRM schema
• Optional: a sub-agent that drafts a tailored outreach email

Data analyst on call

What it does. Engineers and PMs ask plain-English questions; the agent translates them into SQL, runs them safely, and returns charts/tables.

Building blocks

• PostgreSQL, MySQL, or MongoDB integrations (use Sandbox Mode for read-only safety — see Sandbox Mode)
• A schema-aware system prompt that knows your tables
• The Code Executor for charting and post-processing
• A Slack connector so the team can ask questions in-channel

Document & contract Q&A

What it does. Upload a PDF (RFP, contract, policy doc) and answer questions against it with citations.

Building blocks

• Assets with vectorization enabled for retrieval (Knowledge Base)
• The File integration to ingest user-supplied documents at runtime
• An Agent Step that returns answers with source spans

Ticket triage

What it does. New tickets in Jira / Linear / Zendesk get auto-classified, assigned, and tagged based on content and history.

Building blocks

• Jira, Linear, Zendesk integrations
• An Agent Step with a classification prompt and structured output
• Memory to learn assignment patterns over time
• A scheduled trigger or webhook connector

Internal ops bot

What it does. A Slack bot teammates can ask: "what's the status of order 12345?", "who owns the X repo?", "schedule a sync with the design team."

Building blocks

• Slack connector
• Integrations into the systems you reference: Shopify, Stripe, GitHub, Google Calendar, Notion
• Sub-agents per system, with a router agent on top
• Access requests so new users go through an approval flow before they can ask

Research briefings

What it does. Daily or on-demand briefings on a topic — competitive intel, regulatory updates, customer signals — delivered to email or Slack.

Building blocks

• Tavily or BrightData for fresh sources
• A multi-step flow: gather → cluster → summarize → format
• Task scheduling to run on a cron schedule
• An Email or Slack connector for delivery

Lead qualification

What it does. New form submissions get scored, enriched, and routed to the right rep — or politely declined.

Building blocks

• Typeform or your own form via webhook
• Apollo / HubSpot for enrichment
• Structured output to score and tag
• Sub-agent to draft the follow-up email

RAG over your wiki

What it does. Your team's living source-of-truth Q&A — across Confluence, Notion, GitHub, Google Drive — with answers that cite the source page.

Building blocks

• Confluence, Notion, Google Drive, GitHub integrations as data sources for the Knowledge Base
• A retrieval-augmented Agent Step
• A Slack connector

---

Templates: Solutions

Many of the patterns above ship as Solutions — pre-built agent templates you can clone and customize. Browse the Solutions page in the app to start from a working agent instead of an empty canvas.

Don't see your use case?

Most agents are some combination of the patterns above. Start with the quickstart, then layer in the tools, connectors, and knowledge you need.