Fin Operator is an agent for your customer operations that gives every teammate capabilities that have never existed before - from compressing weeks of operational work into minutes, to building, testing, and shipping improvements to Fin in seconds.

It's like adding a team of data analysts, knowledge managers, agent builders, and operations leads to your org - except it's one agent, available to everyone on your team, instantly. One place to go, regardless of the task. One agent that knows your business, your system, and can act on it - fast.

Who it's built for

CX and CS leaders: Stop context-switching between reports, content tools, and your inbox. Surface what's breaking, fix it, and move on, all in one place.
AI leaders: Build and fine-tune how Fin behaves, there is no coding required. When Fin gives an unexpected answer, Fin Operator shows you exactly why it responded that way so you can fix it. You can also run multi-step tasks in a single request, and every change you make comes with a full preview before it goes live.
C-level: Turn your support operation into a strategic asset. Fin Operator surfaces what customers actually need, so you can act on product insights, not just ticket counts.

Fin Operator capabilities

Fin Operator works across multiple skill areas in a single conversation. You can go from analyzing data to fixing a content gap without switching context.

Analysis & reporting	Ask Fin Operator any question about your data or support operation and get structured answers with charts and breakdowns Set Fin Operator to run any analysis on a recurring schedule, delivered automatically
Content management	Paste in a product update or policy change and Fin Operator finds every affected article across your knowledge base Fin Operator drafts new articles or updates to existing ones, matching your tone of voice and content structure Fin Operator proactively identifies underperforming, outdated, or contradictory content Fin Operator remembers your style guides, tone of voice, and content structure across sessions Upload product docs, call transcripts, or briefs directly into a thread to give Fin Operator additional context about content you'd like to create
Fin configuration & optimization	Share a conversation and Fin Operator debugs it — identifies what went wrong and why Fin Operator proposes Guidance changes, Procedure builds, and workflow updates Fin Operator tests proposed changes against real historical conversations before shipping Fin Operator monitors changes after they ship and flags when something needs attention All changes are formatted as proposals, making it easy for you to review and approve before going live
QA monitors	Ask Fin Operator to analyze QA monitor results and identify which conversations are failing and why Get a breakdown of pass/fail rates by monitor, evaluator, or topic to pinpoint quality gaps Fin Operator surfaces patterns in failing conversations and suggests content or guidance fixes to address them

How Fin Operator works

Every configuration change Fin Operator makes goes through a proposal.

When you ask Fin Operator to update an article, create a guidance rule, or modify a procedure, it doesn't make the change directly. Instead:

Fin Operator proposes a change.
You see a diff or summary right in the chat — what's being added, removed, or modified.
You approve, edit, or reject the proposal.
Only after your approval does the change go live.

Think of it like a pull request for your support setup. You always review before anything is merged.

Note: Fin Operator never publishes, deletes, or modifies content directly. Every change flows through this review step, so you stay in control.

See Operator's thinking

After each response, you'll see a collapsible summary you can expand using the chevron (›). It comes in two forms depending on what Fin Operator did:

Thought for X seconds — Fin Operator was reasoning and planning, but no tools were used. Expand to read the full reasoning — not a summary, but the actual thought process: what Fin Operator understood from your request, what it considered, and why it made the decisions it did.
Worked for X seconds — Fin Operator actively used tools and took actions. Expand to see a bullet list of exactly what it did (such as 'Proposed changes to [article name]'), plus a scrollable Thought section with the full internal reasoning behind those actions.

Note: While Fin Operator is thinking or working, the composer (text input) is locked. You won't be able to send a new message until it's finished. This is expected — Fin Operator is mid-task and needs to complete before it can take a new instruction.

This matters for a few reasons:

Catch misunderstandings early — If Fin Operator has interpreted your request differently to what you intended, you'll see it in the thinking before a proposal lands. You can course-correct mid-task without having to unpick a change.
Learn to prompt better — Seeing how Fin Operator interprets your language teaches you how to ask more precisely next time. It's a feedback loop built into every conversation.
Understand proposals, not just changes — When a proposal arrives, the thinking tells you why Fin Operator made specific decisions. You're reviewing the reasoning, not just the diff.
No black box — Especially for new users, the thinking panel demystifies what's happening. You can follow the logic at every step.

Memory and preferences

Fin Operator saves preferences that persist across conversations. If your team has a specific style guide, preferred terminology, or standard reporting preferences, tell Fin Operator once and it remembers for future sessions.

For example: "Remember that we use British English in all customer-facing content" — and every future content proposal will follow that convention. You can also save reporting preferences, default date ranges, and other workspace-level settings.

Note: To save something to memory, explicitly tell Fin Operator to save it — for example, "Save this to workspace memory." Fin Operator won't automatically persist preferences unless you confirm you want them saved.

Known limitations

Conversation analysis limit — Real-time analysis of conversations is currently limited to 250 conversation samples.
External content sources — Can't edit scraped pages, PDFs, or other external content sources. It works with content you manage directly in your workspace.
Configuration boundaries — Can't modify workflows, metadata, Fin settings, or channel settings. It focuses on content, guidance, procedures, and reporting.
Aggregate data — Works with aggregate data and trends, not individual customer records.

Note: If Operator's conversation stops responding, re-engage by sending a new message in the same thread. If the issue persists, contact our support team.

Coming soon

Data connectors — Walk through the setup of data connectors to be used by Fin directly or via procedures.
Scheduled threads — Set up recurring or one-off tasks that run automatically (e.g., "Send me a weekly Fin performance summary every Monday at 9am").
Custom MCP servers — Connect Fin Operator to external tools and services, enabling it to take actions across your toolstack.

💡Tip

Need more help? Get support from our Community Forum
Find answers and get help from Intercom Support and Community Experts

How to use Fin Operator

Use Fin Operator for procedures

Use Fin Operator for conversation debugging

Use Fin Operator for guidance rules

Fin Operator FAQs