Draft customer support replies grounded in your apps

By General Input

Pull the ticket, the customer record, and matching knowledge base articles, then return a ready-to-review reply plus internal handling notes.

Integrations

  • HubSpot
  • Notion
  • Google Drive
  • Dropbox
  • Box
  • Gmail
  • Microsoft Outlook
  • Slack
  • Microsoft Teams
  • Asana
  • ClickUp
  • Linear

Type

Agentic Task

Categories

  • Customer Support

Build me an on-demand agent workflow that drafts a customer support reply for a specific ticket or conversation. The agent should be triggered manually (I invoke it when I want a draft) and given either a reference to the conversation (ticket ID, ticket URL, email thread reference, or task link) or pasted/uploaded source materials.

Source priority: the agent must always try the connected apps first and only fall back to uploaded materials when those sources are unavailable or incomplete.

Step 1: Pull ticket context. Look up the conversation in whichever support, CRM, or work app holds it. Use HubSpot (Get Ticket) for support tickets, Asana (Get a Task) for tasks tracking customer issues, ClickUp (Get Task) for customer-facing work, or Linear (Get Issue) when support escalates issues there. Capture the full history, current status, owner, and any linked records.

Step 2: Pull account context. Use HubSpot (Get Contact) to look up the customer record by email or ID. Capture plan or tier, recent activity, prior tickets, and any notes that affect how we respond.

Step 3: Pull knowledge base context. Search Notion (Search by Title and Retrieve Page as Markdown), Google Drive (List Files with a query, then Download File Content), Dropbox (Search Files), or Box (Search Content) for help articles, playbooks, or runbooks relevant to the customer's question. Cite which document each fact came from.

Step 4: Fallback to uploads. Only if a needed source is not connected, returns nothing useful, or the requested record cannot be found, fall back to materials I attach to the request (pasted transcript, uploaded PDF, screenshot text). Note clearly when a fact comes from an upload rather than the canonical app.

Step 5: Draft three outputs.

A. Customer-ready reply. Written in our team's tone, directly answering the customer's question, with a clear signoff. Should be ready to send with light editing.

B. Internal handling notes. A short block for the agent: who this customer is and any account context worth knowing, escalation guidance if the issue is urgent, billing related, security related, or outside our normal scope, and concrete next steps if the reply needs follow-up.

C. Missing evidence and open questions. A bulleted list of facts the agent should verify, information the customer has not provided, and decisions that need a human before sending.

Step 6: Deliver the draft somewhere reviewable. Default behavior: save it as a draft addressed to the customer using Gmail (Create a Draft) or Microsoft Outlook (Create Draft Message or Create Reply Draft). Alternative behavior, if my team reviews drafts in chat: post the reply, internal notes, and open questions to a designated Slack channel (Send a Message) or Microsoft Teams channel (Create Channel Message). Optionally also leave the internal notes on the source record using HubSpot (Create Note) or Linear (Add Comment to Issue).

Never send the reply directly to the customer. The agent's job ends at producing a draft for human review.

If no connected app returns the referenced ticket and no materials were uploaded, the agent should stop and ask for either a valid reference or attached source materials rather than guessing.

When the workflow runs, ask me at setup which apps to use as the primary ticket source, the primary knowledge source, and the primary draft destination, plus a short voice or style guide for the customer reply and our escalation rules.

Related prompts

Explore more prompts
Call overdue Xero customers with an AI collections agentWin back LiveChat visitors whose chats went unansweredTurn Mailjet email clicks into ranked HubSpot follow-upsiMessage campaign console with pre-flight checks and delivery boardChat quality review board for LiveChat support leadsA searchable RFP answer library your bid team drafts fromSee which target accounts just started advertising on LinkedInAccount health board that puts product usage next to your CRMLook up a customer's full chat history mid conversationChase the paperwork every new client and vendor still owes