Put an AI agent to work inside Intercom
Connect Intercom in one click and delegate the busywork: an autonomous agent that triages conversations, drafts replies grounded in your help center, escalates the ones that matter with full context, and acts across the rest of your stack. You set the autonomy, it does the work.
4-day trial · Cancel anytime
Fleece AI connects to Intercom through managed OAuth and lets autonomous agents read and triage conversations, draft replies grounded in your help-center articles, tag and segment, and escalate the ones that matter. Agents combine Intercom with 3,000+ other apps — Zendesk, Slack, HubSpot, Gmail — so a conversation can update a CRM, ping the right owner, or raise an engineering ticket, under the approval rules you set. It complements Intercom Fin rather than replacing it.
At a glance
| Category | Support |
|---|---|
| Availability | Pro plan and up — included in the 4-day trial |
| Connects to | 3,000+ apps via managed OAuth |
| Setup | Connect your tools — no code required |
| Autonomy | Suggest-only to fully autonomous, with approval gates |
| Pricing | 4-day trial (card required, no charge today), then paid plans |
By Loïc Jané · Updated June 30, 2026
What a Fleece agent does with Intercom
Intercom is where your customer conversations live — the inbox, tickets, help-center articles, outbound messages, and the segments that group your users. Fin, Intercom's AI agent, already answers a lot of front-line questions inside Intercom. But plenty of work sits around those conversations: reading a thread and routing it to the right person, pulling the customer's account context from your CRM, escalating a churn risk to the right channel, updating a ticket, and turning a week of conversations into an insights report. A Fleece agent does that work, and it does it across your whole stack rather than inside Intercom alone.
Under the hood, the connection runs through managed OAuth — you authorize Intercom once, and Fleece handles tokens, scopes, and refresh securely. The agent can read conversations and the inbox, draft and send replies, open and update tickets, apply tags and segments, look up help-center articles, and send outbound messages. Event triggers let it react in real time when a conversation opens, a reply lands, or a tag is applied — no polling, no delay — so triage and escalation happen in seconds.
What makes this different from a fixed rule or a stand-alone bot is judgment plus reach. A Fleece agent reads the actual conversation, decides what it is, drafts an answer grounded in your help-center articles, and coordinates other tools to finish the job — updating HubSpot, pinging Slack, raising a GitHub issue. And because Fleece agents work as a hierarchy — a lead agent delegating to specialized child agents — one Intercom-facing agent can hand research, CRM, or engineering work to a child agent and report the outcome back on the conversation.
What the agent can do in Intercom
Conversation triage
Watches the inbox, reads each incoming conversation, and routes it by topic, urgency, and customer tier so the right person picks up the right thread.
Grounded draft replies
Drafts replies grounded in your help-center articles, cites the source, and either leaves them for a teammate or sends them automatically on routine questions.
Escalations with context
Escalates VIP and churn-risk conversations to Slack with the customer's CRM context attached, so the right owner responds fast and informed.
Tagging and segments
Tags conversations, applies segments, and keeps records consistent so routing, reporting, and outbound targeting stay accurate.
Real-time triggers
Reacts the moment a conversation opens, a reply lands, or a tag is applied, through event-based triggers rather than polling.
Approval gates
Anything customer-facing — sending a reply, closing a ticket, messaging a segment — can pause for your one-click sign-off before it happens.
Automations teams run on Intercom
These are concrete setups you can describe to a Fleece agent in plain language. Each one combines Intercom with other connected apps.
Support: triage and route with judgment
The agent watches the Intercom inbox, reads each new conversation, and routes it by topic, urgency, and customer tier — billing questions to one teammate, suspected bugs flagged for engineering, VIP threads to a senior agent. It tags each conversation so segments and reporting stay accurate, and posts a Slack note the moment a thread needs a human now rather than a queued reply.
Support: replies grounded in your help center
For routine questions, the agent drafts a reply grounded in your Intercom help-center articles, cites the article, and checks the customer's plan in HubSpot so the answer fits their tier. It either leaves the draft for a teammate to send or sends it under your rules. When no article covers the question, it says so and flags the gap instead of inventing an answer.
Success: escalate churn risk with CRM context
When a high-value or at-risk customer opens a conversation, the agent recognizes them from HubSpot — plan, renewal date, open deals — and escalates to a Slack channel with that context and a one-line summary attached. The owner sees who it is and why it matters before they open Intercom, so the first response is fast and already informed.
Leadership: a weekly support insights digest
Every Monday the agent reads the week's Intercom conversations, groups them by theme, and produces a support insights digest: volumes, response times, recurring issues, and the documentation gaps where neither Fin nor the team had an article to point to. It posts the digest to Slack so the team knows exactly what to document and fix next.
How to connect Intercom to Fleece AI
Create your Fleece account
Sign up and start the 4-day trial. Your Intercom agent, its flows, and every other integration live together in one workspace.
Connect Intercom via managed OAuth
Pick Intercom from the integrations catalog and authorize it in one click. Fleece manages tokens and scopes for you; you can revoke access at any time from Intercom or from Fleece.
Describe the job in plain language
Create an agent and tell it what to watch and what to do — "triage new conversations, draft replies from our help center, escalate churn-risk customers to #success". No flowchart building required.
Set autonomy and approval gates
Choose what the agent may do on its own and what waits for your sign-off. Customer-facing actions like sending a reply can pause for one-click approval.
Run it on triggers or a schedule
Let the agent react to Intercom events in real time, or schedule recurring flows — an end-of-day handoff summary, a weekly insights digest — that run without you.
Intercom works better with the rest of your stack
Intercom automations get powerful when they cross app boundaries. Pair Intercom with HubSpot so replies and escalations carry real account context, with Slack to route the conversations that need a human now, with Zendesk to keep tickets consistent across tools, or with Gmail to fold email threads into the same triage — all through the same agent, all under the same approval rules. Fleece connects to 3,000+ apps, so the stack you already use is almost certainly covered.
Frequently asked questions
Only if you allow it. You set approval gates per agent: routine actions like tagging or drafting a reply can run automatically, while customer-facing ones — sending a reply, closing a ticket, messaging a segment — pause for your one-click sign-off. You choose the autonomy level, from suggest-only to fully autonomous, and can tighten it at any time.
The connection uses managed OAuth: you authorize Intercom once through the official consent screen, and tokens are stored and refreshed securely — you never paste API keys. Access is scoped to what you grant, you can revoke it from either side at any time, and security-relevant actions are recorded in an audit log so you can see what the agent did.
Yes. Event triggers fire the moment a conversation opens, a reply lands, or a tag is applied, so triage and escalations happen in seconds rather than on a delay. You can combine real-time triggers with scheduled flows — for example instant VIP escalation to Slack plus a weekly support insights digest that runs every Monday.
Fin answers front-line questions inside Intercom, and it does that well. A Fleece agent complements it: it reads conversations and acts across your whole stack — updating HubSpot, escalating to Slack, raising engineering tickets, and reporting on trends. Think of Fin as answering in the inbox and the Fleece agent as handling everything that has to happen elsewhere as a result.
Intercom is part of the Pro tier — available on Pro and Business plans, where Pro unlocks 1,000+ apps and Business the full 3,000+ catalog. You can try it during the 4-day trial of those plans. Starter covers about 20 essential apps and does not include Intercom, so pick Pro or Business to connect it.
Yes — that's the core design. One agent can read an Intercom conversation, pull context from HubSpot, escalate to Slack, and raise a ticket in Zendesk in a single run. For bigger jobs, a lead agent can delegate to specialized child agents and report the outcome back on the Intercom conversation.
Runs are recorded step by step, so you can see exactly what the agent did and why. Failed flows retry sensibly and pause themselves after repeated failures instead of looping, and you are notified so nothing fails silently. For tools without a stable API, browser automation is available as a fallback path.
Yes. The agent grounds draft replies in your Intercom help-center articles and cites the source it used. When no article covers the question, it says so and flags the gap instead of inventing an answer, and sensitive replies can wait for your approval before they reach a customer. Accuracy and human sign-off stay under your control.
Go deeper on the blog
Put Intercom on autopilot
Connect Intercom in one click, describe the job in plain language, and let an autonomous agent triage, draft, and escalate across your stack. 4-day trial, cancel anytime.
Powered by Fleece AI · autonomous agents for 3,000+ apps