$0.99 per resolved conversation. Intercom built Fin's pricing to feel like a no-brainer - you pay when the AI works, nothing when it doesn't. No per-seat AI license, no monthly minimum tied to agent headcount. For a support leader evaluating AI tools, it reads as genuinely simple.
Run the numbers past 10,000 monthly resolutions, and you're looking at a different calculation entirely.
What Fin Gets Right
Fin is a legitimately good product for the right use case. Dismissing it as expensive or poorly built would be wrong, and it would waste the time of anyone who fits the target profile.
Intercom has built an AI support tool that deploys in days rather than months. Fin integrates directly with Intercom's inbox, conversation history, and customer records. Support teams can update the knowledge base, adjust response tone, and modify routing behavior through a configuration interface - no engineering tickets required. It handles 80+ languages and resolves the common-question volume that used to require additional headcount.
For SaaS companies with well-documented help centers, Fin's ability to answer "how do I export my data?" and "what is your refund policy?" at 3am without human involvement is genuinely useful. The self-serve configuration is a real operational advantage for lean teams - a support lead making changes in a UI beats waiting a week for an engineering sprint. That matters, especially for companies whose support operations are already running inside Intercom's ecosystem alongside 25,000+ other businesses.
Fin's pricing transparency is also better than most of the field. Industry pricing comparisons show that many competing platforms bury AI costs inside opaque seat licenses or charge flat monthly fees regardless of how often the AI actually resolves anything. $0.99 per resolution ties the cost directly to the value delivered. Minami's 2026 Fin pricing breakdown confirms the model is cleaner than most enterprise AI pricing - the problem is not opacity, it is what the math looks like as volume grows.
Deployment speed matters too. A custom-built AI agent takes weeks to months to design, build, test, and deploy. Fin takes days. For a company needing AI support coverage before a product launch or heading into a high-volume season, that timeline gap has real dollar value.
The $0.99 Math That Changes Everything
Here is the catch: $0.99 per resolution is a growth penalty dressed as a fair deal.
The per-resolution model sounds reasonable at low volume. At moderate volume it becomes a significant budget line. At high volume it is expensive by most measures. It also scales in exactly the wrong direction - as your business grows and generates more customer contacts, your AI cost grows in lockstep. The entire point of automation is to decouple support costs from revenue growth. Fin's pricing structure doesn't do that.
| Monthly Resolutions | Monthly Fin Cost | Annual Fin Cost |
|---|---|---|
| 1,000 | $990 | $11,880 |
| 5,000 | $4,950 | $59,400 |
| 10,000 | $9,900 | $118,800 |
| 20,000 | $19,800 | $237,600 |
| 50,000 | $49,500 | $594,000 |
| 100,000 | $99,000 | $1,188,000 |
These are Fin fees only. On top of them: Intercom's base platform runs $29-$132 per human agent seat per month depending on plan tier. Features like advanced reporting, certain routing capabilities, and higher API limits require higher tier subscriptions. Flowgent's 2026 Intercom pricing guide shows how these layers combine in practice; the headline $0.99 understates total cost at any meaningful volume.
The demand spike problem compounds this. Retail companies see holiday volume 3-5x above baseline. SaaS companies face support surges during product launches and outages. Healthcare operations experience seasonal volume spikes around flu season. In every case, per-resolution pricing means the AI bill spikes at exactly the moment when automation is supposed to hold costs stable. The pricing model inverts the benefit you are trying to capture.
Run a three-year projection for a company starting at 20,000 monthly resolutions and growing at 40% year-over-year:
- Year 1 (20K resolutions/mo): ~$237,600 in Fin resolution fees
- Year 2 (28K resolutions/mo): ~$332,640 in Fin resolution fees
- Year 3 (39K resolutions/mo): ~$464,100 in Fin resolution fees
- Three-year total: ~$1.03M in resolution fees
That is before Intercom platform costs. Before seat licenses for the humans handling escalations. And at the end of year three, you own nothing.
The Ecosystem Wall
Fin's native Intercom integration is its sharpest competitive advantage and its hardest constraint.
Within Intercom's world - conversations starting via email or chat, resolved inside the Intercom inbox, drawing on customer data that Intercom holds - Fin has clean access and performs well. The integration is deep by design.
The wall appears when AI needs to work across systems, or when the support workflow itself crosses into infrastructure Intercom doesn't control.
A retail company handles "where is my order?" questions at scale - Fin handles those well. But a return process that requires checking nearest-store inventory, scheduling a pickup window, adjusting loyalty point balances, notifying warehouse fulfillment, and updating the order management system is a multi-system orchestration problem. Intercom doesn't have deep integrations with any of those systems. Fin can field the customer's opening question; it cannot complete the workflow.
A B2B SaaS company wants AI triaging bug reports: pull relevant monitoring data, check whether the issue matches a known deployment state, determine severity, then route to the right engineering team with full context included. The customer conversation is 20% of the value; the multi-system triage is the other 80%. Fin handles the conversation side. The orchestration has to happen somewhere else.
A media company running AI beyond subscriber support - ad operations, content scheduling, rights management queries - finds that each workflow lives in different systems with incompatible data models. Fin doesn't span them.
Fini Labs' 2026 comparison of agentic support platforms categorizes Fin accurately as a support deflection tool. That is a description, not a criticism. Support deflection and cross-system business process automation are different categories of problem, and Fin is built for one of them.
Cross-platform comparisons of Intercom, Drift AI, and HubSpot AI show this ceiling is consistent across the category - platforms built inside CRM or helpdesk products tend to stop where the helpdesk stops. When the business need outgrows support deflection, Fin doesn't extend. The typical outcome is buying a separate tool for each additional workflow, and the total system becomes fragmented and expensive to maintain.
What You Own After Three Years
Most cost comparisons stop at monthly spend. The ownership question is just as important, and it almost never gets asked.
Every per-resolution payment to Intercom buys access to a service you are renting. The knowledge base you built, the routing logic you refined, the edge cases you taught Fin to handle through months of careful configuration - all of it lives on Intercom's servers, governed by Intercom's terms, subject to Intercom's pricing decisions.
Inkeep's architecture comparison with Fin identifies portability as a structural distinction between Fin and alternatives built with platform-independent integration in mind. Fin's behaviors and learned configurations don't transfer when you change platforms. The iterations that moved your AI resolution rate from 40% to 72%, the custom flows for enterprise customers, the escalation logic tuned over a year - they are settings inside a SaaS product, not assets you own.
When a company migrates from Intercom to Zendesk, Freshdesk, or Salesforce Service Cloud, the Fin investment resets to zero. You start over regardless of what you spent.
A custom-built AI agent works differently. The codebase belongs to the company. The agent runs in infrastructure you control. It connects to the systems you specify, adapts when those systems change, and survives any vendor migration intact. A flat monthly retainer covers ongoing development and maintenance. When the engagement ends, you keep what was built.
After three years of Fin at scale, you have invoices and a configuration inside a vendor platform. After three years of custom development at a flat retainer, you have software.
How to Choose
This is not a case where one approach is universally better. The right choice depends on volume, use case scope, and how much vendor dependency matters to your business.
| Intercom Fin | Custom AI Agent | |
|---|---|---|
| Pricing model | $0.99/resolution + platform fees | Flat monthly retainer |
| Cost at 50K resolutions/mo | ~$49,500/mo in resolution fees alone | Fixed monthly fee |
| Software ownership | No - lives in Intercom | Yes |
| Works outside Intercom | No | Any system |
| Use cases | Customer support deflection | Any business workflow |
| Deployment speed | Days | Weeks to months |
| Configuration requires engineering? | No | Yes |
| Best for | Moderate-volume SaaS on Intercom | High volume or cross-system automation |
Choose Fin when:
- You are already on Intercom and plan to stay
- Support volume is under 10,000 resolutions per month
- AI needs are limited to answering support questions from a knowledge base
- You need to be live in days, not weeks
- Your team doesn't have engineering capacity for custom development right now
Choose custom when:
- Per-resolution fees are approaching or exceeding what a flat monthly retainer would cost
- You need AI across multiple systems, not just the helpdesk
- Use cases extend beyond support into operations, compliance, scheduling, or internal workflows
- You want to own what you pay to build
- You are growing fast enough that predictable, fixed automation costs matter more than low entry price
The inflection point is roughly 10,000 monthly resolutions. Below that, Fin's speed and simplicity are often worth the cost structure. Above it, the math starts favoring alternatives - particularly once multi-year growth is in the picture.
For companies running 30,000+ monthly resolutions, the question worth answering honestly is: "Are we comfortable paying over $400,000 per year for AI we will never own, locked to a single vendor's platform?"
Want to see what custom AI agents would cost for your volume and use cases? Talk to us.