Fin
Fin, formerly Intercom, is an autonomous support agent that resolves customer conversations end to end: grounded in your help content, working across chat and email, and priced by outcome rather than seat, so you pay per resolution it actually completes. It layers over existing helpdesks rather than demanding a migration.
The operating loop is the product: simulate answer quality before customers see it, deploy on routine volume, escalate the rest to humans with full context attached, and mine unresolved questions for the content gaps behind them.
Real-world resolution settles well under half of volume for most teams, which is the honest baseline to plan staffing and economics around.
- Cost
- Paid only
- Ease
- Intermediate
- Model
- Hosted service
- Checked
- July 2026
Prices, plans and model versions change fast: this is a mid-2026 snapshot; check the tool's official site for the latest.
Best for
- Autonomous end-to-end resolution of routine conversations
- Outcome pricing: pay per resolution, not per seat
- Layering over an existing helpdesk without migration
- Escalation to humans with full context attached
- Mining unresolved cases for missing help content
Less suited to
Outcome pricing defies flat-rate budgeting: costs scale with volume resolved, and forecasting takes modelling rather than a price list. Realistic resolution sits well under half of conversations for most teams, so humans remain the majority channel.
Answer quality also rides entirely on help-content quality: thin documentation produces a confidently unhelpful agent.
Costs & data, in short
Fin charges per resolved conversation (currently $0.99), so the bill scales with how well it performs. Used inside Intercom it also needs at least one paid seat; used on another helpdesk there is a small monthly minimum. Model your real ticket volume before committing.
Fin answers from your help content and past conversations under Fin's commercial terms. One strategic note: Salesforce has agreed to acquire the company, with Fin expected to fold into its Agentforce line, which is worth weighing in any long-term platform decision.
In practice
How Fin is used, area by area.
Customer support
Fin is the support agent to beat on autonomous resolution. It answers from your help content and past conversations, resolves a large share of routine volume end to end across chat and email, and charges only when it succeeds, with the operating loop as the real product: simulate answer quality before customers see it, deploy on routine volume, escalate the rest with full context, and mine unresolved questions for the content gaps behind them. It also layers over a helpdesk you already have, including Zendesk, so the bot decouples from the platform decision. It belongs with teams whose goal is reducing ticket volume, paying for outcomes rather than seats. Bills scale with success and resist forecasting, and a resolution is counted when the customer stops replying, not only when they confirm the fix, so audit resolved conversations. The pending Salesforce acquisition belongs in any long-term platform decision.
Example tasks
- Resolve routine questions end to end across chat and email
- Run Fin on top of your existing helpdesk without a migration
- Escalate unresolved cases to humans with the full conversation attached
- Use its reporting to find the content gaps behind failed answers
- Simulate answer quality on real past conversations before launch
Limits
Bills scale with success and can be hard to forecast, since a resolution is counted when the customer stops replying, not only when they confirm the fix. Budget-sensitive teams should model real volume first.
Compares
| vs | Pick Fin when | Pick the other when |
|---|---|---|
| Zendesk AIFull comparison → | Fin typically resolves a higher share of conversations autonomously | deep ticketing workflow and reporting matter more than the resolution rate |
| Tidio LyroFull comparison → | Fin is the stronger and more expensive engine | the budget and the ticket volume are both small |
Founders & entrepreneurs
Fin lets a founder ship real customer support without a support hire. It answers from your docs and past conversations, resolves the majority of routine volume end to end, and charges only when it succeeds, which fits founder economics precisely: cost scales with customers rather than headcount, and the first support hire gets deferred until volume genuinely justifies one. It meets founders at a specific moment: the support inbox outgrowing personal reply capacity. The preconditions matter: with no help content it has nothing to answer from, so write the docs first, and pre-launch products with ten users need a founder replying personally more than they need deflection. Outcome pricing means success costs money, so model resolution volume against runway, and note the pending Salesforce acquisition before signing anything long-term.
Example tasks
- Stand up an agent that answers from your help docs from day one
- Resolve routine billing, how-to and account questions end to end
- Escalate the conversations that genuinely need the founder, with context
- Mine unresolved questions for the docs you have not written yet
- Scale support volume without the first support hire
Limits
With no help content it has nothing to answer from, so write the docs first; and pre-launch products with ten users need a founder replying personally more than they need deflection.
Compares
| vs | Pick Fin when | Pick the other when |
|---|---|---|
| Zendesk AIFull comparison → | Fin runs standalone or atop whatever helpdesk you adopt later | you are committing to its full platform early |
| Tidio LyroFull comparison → | Fin is the stronger engine at higher cost per resolution | budget rules and volume is small |
Customer support & chatbots
Fin is the closest thing this category has to a benchmark. Its published resolution rates lead the mainstream field, it deploys across chat, email and voice, and unlike most rivals it does not force a platform migration: it operates as the AI layer on whatever helpdesk you run, which decouples the bot decision from the helpdesk decision entirely. Teams choosing on resolution performance start here, with the baseline stated plainly: real-world resolution settles well under half of volume for most, so humans remain the majority channel. Outcome pricing defies flat budgeting, since a good month costs more, and what counts as resolved deserves auditing, because abandoned is not answered. The pending Salesforce acquisition makes the long-term roadmap a fair pre-signature question, and cost per resolution deserves monthly tracking against a human-handled baseline.
Example tasks
- Deploy an agent that resolves the majority of routine conversations
- Layer Fin over Zendesk, Salesforce or HubSpot without moving platforms
- Test answer quality in simulation before customers see it
- Extend the same agent from chat into email and voice channels
- Track resolution quality, not just deflection volume
Limits
Outcome pricing means a fixed budget is impossible to guarantee; a good month costs more. Teams wanting a flat, predictable line item should weigh seat-priced or platform-included alternatives.
Compares
| vs | Pick Fin when | Pick the other when |
|---|---|---|
| Zendesk AIFull comparison → | Fin is the specialist with the stronger autonomous engine | you want the bot and the helpdesk from one vendor |
| Freshdesk FreddyFull comparison → | Fin leads on resolution capability | you are already on Freshworks and want good-enough AI at lower complexity |
Where to start
Not sure what to adopt first?
Five quick questions about your job, task and constraints. We'll suggest your top three tools, plus the one to try first.
Alternatives
Same category, different strengths.
- ChatGPTWriting, research, analysis and everyday professional productivity
- ClaudeAnalysing complex documents, research and large amounts of source material
- Freshdesk FreddyNo-code support agents answering from the knowledge base
- Google GeminiA genuinely capable free assistant with paid depth when needed
Where it fits
Explore this tool in context.
For your job
By task
Common questions
What is Fin best at?
Fin is strongest for autonomous end-to-end resolution of routine conversations; outcome pricing: pay per resolution, not per seat; layering over an existing helpdesk without migration; escalation to humans with full context attached; mining unresolved cases for missing help content.
What is Fin not good for?
Outcome pricing defies flat-rate budgeting: costs scale with volume resolved, and forecasting takes modelling rather than a price list. Realistic resolution sits well under half of conversations for most teams, so humans remain the majority channel. Answer quality also rides entirely on help-content quality: thin documentation produces a confidently unhelpful agent.
Is Fin free?
No: Fin is a paid product, with plans for individuals and teams.
Where does Fin fit best?
Fin fits best in Customer support and Founders & entrepreneurs; see its practice notes for how.
Before sharing confidential or personal data, check this tool's data-governance and training policies. They differ between providers and can change.
Last checked: July 2026