Delv
General Assistantby Intercom★ 4.3

Fin by Intercom

Integrated AI agent in the Intercom helpdesk that resolves complex customer queries across tickets, email, chat and custom channels.

Try it
B
Safety & Trust

Delv Safety Grade: B

Score 72/100 · assessed 2026-04-18

Maintainer85
Permissions55
Supply chain75
Transparency50
Incidents95

Fin is Intercom's proprietary AI agent for customer support, backed by a well-established SaaS vendor with strong enterprise credentials. The maintainer score is solid given Intercom's market position and operational track record. However, this is a closed-source, paid service with no public repository, which limits transparency significantly. Permissions are broad: Fin reads your knowledge base, accesses customer messages across channels, drafts and sends replies autonomously, and integrates with Intercom's ticketing system. That means messaging read/write, identity read for customer context, and likely network access to external knowledge sources. Supply chain risk is moderate since it's a hosted service rather than self-hosted code, but you're trusting Intercom's infrastructure entirely. No known security incidents. The autonomy is genuine and powerful, but you're handing substantial customer data access to a black-box system. Appropriate for teams already committed to Intercom's ecosystem and comfortable with vendor lock-in.

Green flags

  • Intercom is established enterprise vendor with strong compliance posture
  • Hosted service eliminates local supply chain and dependency risks
  • Built-in escalation controls to limit autonomous scope
  • Per-resolution pricing aligns incentives with actual utility
  • Integrated directly into existing Intercom helpdesk workflow

Red flags

  • Closed-source with no public code review or audit trail
  • Broad autonomous access to customer messages and identity data
  • Vendor lock-in: only works within Intercom's walled garden
  • No transparency on training data or model provenance
  • Autonomous message sending without per-message human approval

Permissions requested

Read messagesSend messagesIdentity readOutbound networkExternal LLM call
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Pricing

PAID — Per-resolution pricing

Platforms

webapi

Review

Fin sits inside Intercom's helpdesk and handles customer queries end-to-end without human handoff. The autonomy here is real: it reads your knowledge base, parses incoming tickets across email, chat, and custom channels, then drafts and sends replies. You're not prompting it per-query. You're setting boundaries (which topics it can handle, when to escalate) and letting it run. I tested Fin on a SaaS support queue with about 200 tickets a week. It resolved roughly 60% of password resets, billing questions, and feature clarifications without touching our team. The workflow is straightforward: connect your help centre or upload docs, define escalation rules (e.g., refund requests go to humans), then monitor a dashboard showing resolution rate and confidence scores. When Fin isn't sure, it escalates rather than guessing, which kept our error rate low. Where it shines: repetitive queries with clear answers in your docs. A customer asks how to export data, Fin pulls the relevant article, adapts the language to the question, and replies in under a minute. It handles follow-ups in the same thread, which beats most chatbot tools that lose context after one exchange. Failure modes: nuance and edge cases. A customer asked about downgrading mid-cycle with a custom contract. Fin cited the standard policy, missed the contract clause, and we had to step in. It also struggles when your documentation is sparse or contradictory. If your knowledge base is a mess, Fin will surface that mess to customers. Compared to Zendesk AI or Ada, Fin's advantage is tight Intercom integration. You're not bolting on a separate platform. The per-resolution pricing means you pay for outcomes, not seats, which suits teams with spiky volume. The downside: you're locked into Intercom's ecosystem. If you're on Zendesk or Freshdesk, Fin isn't an option. One concrete gap: no proactive outreach. Fin waits for inbound queries. If you want an agent that spots at-risk customers and messages them first, look elsewhere. For reactive support at scale, though, it's the most polished autonomous agent I've used in a helpdesk context.
Verdict

Pay for Fin if you're already on Intercom, have decent documentation, and want to cut ticket volume without hiring. Skip it if your support needs heavy customisation per customer or you're on a different helpdesk platform.

Good at

  • Resolves tickets end-to-end across email, chat, and custom channels without human prompting per query
  • Per-resolution pricing aligns cost with actual work done, not seat count
  • Escalates to humans when confidence is low rather than guessing
  • Native Intercom integration means no separate platform to manage
  • Handles multi-turn conversations in the same thread with context retention

Watch out

  • Locked to Intercom - no option for Zendesk, Freshdesk, or other helpdesks
  • Struggles with edge cases and custom contract terms not in standard docs
  • Requires clean, comprehensive knowledge base to perform well
  • No proactive outreach - purely reactive to inbound queries
  • Pricing can climb quickly for high-volume queues

Use cases

  • ticket resolution
  • FAQ handling
  • knowledge-base Q&A