Grounded Support Replies
Every reply starts as a grounded draft that already knows the customer’s history, so agents edit instead of writing from scratch. A ModuleX agent reads the Intercom conversation and contact, drafts a reply, and leaves it as a note for a human to approve.
Intercom
6 steps1 integrationRuns on demand
Context is scattered across the record. What writing every reply from scratch costs a full queue.
- 01A good reply needs the history: past asks, the plan, what is tagged on the record.
- 02Pulling that together for every conversation is slow, so replies go generic instead.
- 03Across a full queue that means slower first responses that miss context the customer already gave.
- Read threadsecond ask about SSO
- Pull contextMax plan · 3 seats
- Draft replygrounded in the docs
- Approvewaiting on you
One workflow, six steps. ModuleX drafts this graph from a single sentence, and every step stays editable.
- 01Open the customer conversationIntercom
- 02Read the contact and their historyIntercom
- 03Check tags and past conversationsIntercom
- 04Draft a grounded replyModuleX
- 05Leave it as an internal noteIntercom
- 06Send only after a human approvesIntercom
The draft never leaves the note on its own. reply_to_conversation runs only after a human approves, and the workflow keeps that gate however you edit the rest of the graph.
Every reply, grounded. Paste this into ModuleX chat, or hit Use this workflow. It drafts the graph, connects your accounts, and schedules it.
When a customer replies in Intercom, read their contact history and past conversations, draft a reply that answers using that context, and leave it as a note. Nothing goes to the customer until I approve it.
to: composerOr skip the workflow entirely. Ask the assistant in chat and it runs the same job agentically: it picks the right actions, chains them, and reports back. The composer is for when you want it repeatable.
Read the latest reply in this Intercom conversation and the contact’s history, draft an answer grounded in what they already told us, and leave it as an internal note for me to review.
- intercom · get_conversation · thread
- intercom · get_contact · history
- intercom · create_note · draft
Draft ready as a note: it answers their second SSO question, references the seat limit they hit last month, and links the setup steps. Nothing goes to the customer until you approve.
Where this fits. The tools it runs on, and the teams that run it.
Keep the momentum. Nearby workflows that reuse the accounts you just connected.
Support Ticket TriageNew tickets get read, prioritized, and routed the moment they land, and the urgent ones surface in Slack instead of sitting in a queue. A ModuleX agent classifies each Freshdesk ticket, sets its priority and tags, assigns it, and posts the ones that cannot wait.
Answer an order question with the order in handA reply that answers with the real order status instead of asking the customer for their order number. Built for an ecommerce team answering the same three questions about where a package is.- Answer support questions using your own documentationA drafted answer grounded in your own material, showing what it drew from, behind an approval step. Built for support teams whose answers already exist, just not where the question arrives.
About this use case.
No. The draft is posted as an internal note with create_note, and it only reaches the customer through reply_to_conversation after a human approves it.