Fin Review: Intercom Renamed Itself, Then Sold for $3.6B

On May 12, 2026, a fifteen-year-old company looked at its own name and decided it wasn't worth keeping. Intercom, the customer messaging platform that basically invented the in-app chat bubble, renamed itself Fin, after an AI agent that didn't exist five years earlier. Five weeks later, Salesforce agreed to pay $3.6 billion for the whole company.

That is not a rebrand for the sake of a new logo. It's what happens when a feature quietly becomes the business.

Fin is an AI customer service agent that resolves support conversations on its own, across chat, email, voice, WhatsApp, and Slack, and it now closes out roughly 76% of the conversations it touches across more than 12,000 businesses, according to the company's own published numbers. This review covers what Fin actually does, what it costs, how that resolution rate holds up against independent testing, and whether it's worth signing up for a product that's mid-acquisition.

Short answer: Fin is one of the few AI support agents that publishes real, per-resolution pricing instead of hiding behind "contact sales," and the resolution numbers are genuinely strong. The catch is that "strong" and "guaranteed" aren't the same thing, and you're buying from a company that renamed and sold itself in the same quarter.

Quick Verdict

What Works Well / What To Watch

What works well:

• Genuinely transparent pricing. Fin publishes its per-resolution rate on its website, while most enterprise competitors make you sit through a sales call to find out what anything costs.

• A real money-back guarantee. Fin's Million Dollar Guarantee pays out up to $1,000,000 to new customers unsatisfied in the first 90 days, and separately promises at least a 65% resolution rate to qualifying high-volume customers or refunds them $1,000,000.

• Multi-channel from the start. Chat, email, voice, WhatsApp, SMS, and Slack all run through the same agent, so you're not stitching together separate bots per channel.

• Fast to launch. Fin can go live off your existing help center content in under an hour, without engineering resources, which matters if you don't have a dedicated automation team.

• It keeps improving on its own. Fin's resolution rate climbs roughly 1% a month per customer as the underlying models and content improve, so day-one performance understates where an account typically lands.

What to watch:

• The advertised 76% resolution rate is a company-wide average, not a promise. Independent reviewers who've run their own ticket batches through Fin report meaningfully lower numbers, in some cases down around 38%, depending on how messy the knowledge base is.

• You're buying into an acquisition. Salesforce's $3.6 billion deal to acquire Fin was signed in June 2026 and isn't expected to close until early 2027. Leadership says pricing and the roadmap won't change before then, but "before then" is doing some work in that sentence.

• Outcome-based pricing can get expensive at real volume. $0.99 a resolution sounds cheap until a high-traffic support team is resolving tens of thousands of conversations a month, at which point the bill scales with success, which is the point, but budget for it.

• Running Fin fully requires Intercom's own helpdesk (at $29 per seat, per month) to unlock everything. Pairing it with a different helpdesk works but costs a separate $49 monthly base fee on top of the per-resolution charge.

• Add-ons aren't included. Real analytics and QA tooling (Operator, CX Score, AI Topics) sit behind a Pro add-on starting at $99 a month, and agent-assist tooling (Copilot) is another $35 per user, per month.

What Is Fin?

Fin started as a feature inside Intercom, the messaging platform three co-founders launched to help SaaS companies talk to their customers without email. For most of Intercom's life, the company was known for chat widgets and product tours, not AI. Fin changed that quietly, then all at once.

By 2026, Fin's AI agent business had crossed $100 million in annual recurring revenue on its own, growing at roughly 3.5x year over year, while the rest of Intercom's legacy software business sat around $300 million and wasn't growing nearly as fast. Fin, in other words, had become responsible for almost all of the company's growth while still being a minority of its revenue. CEO Eoghan McCabe's answer to that math was to rename the entire 15-year-old company Fin on May 12, 2026, saying the new name carried no baggage: "they have no baggage, they don't need to convince anyone of their new position in the market, because they never had an old one." The Intercom name lives on as the label for the underlying helpdesk software, but the company itself is now Fin.

Five weeks later, on June 15, 2026, Salesforce signed a definitive agreement to acquire Fin for approximately $3.6 billion, bringing more than 30,000 business customers into Salesforce's fold. Salesforce CEO Marc Benioff framed it as filling a gap in Agentforce, Salesforce's own AI agent platform: "Fin brings proven agent technology, a deep commitment to customer success, and an incredible AI team that will complement Agentforce with powerful service agent capabilities." McCabe is staying on as CEO through the transition, and Salesforce expects the deal to close by the end of its fiscal Q4 2027, pending regulatory approval. Notable current customers include Anthropic, Rocket Money, Clay, and Gamma, two tools we've reviewed here ourselves.

Key Features

• Multi-channel resolution. Fin handles chat, email, voice, WhatsApp, SMS, Slack, and social messaging through one agent, rather than requiring a separate bot per channel.

• Procedures. A no-code workflow builder that lets Fin follow multi-step processes, like verifying an order, checking eligibility, then issuing a refund, instead of just answering FAQs.

• Roles. Fin can be configured to run Service, Sales, or Ecommerce roles, switching automatically depending on what the conversation actually needs, so the same agent can resolve a support ticket and qualify a sales lead.

• Operator. A newer, separate AI agent whose only job is monitoring and improving Fin's own performance in the background, flagging content gaps and tuning behavior without a human doing it manually.

• Copilot. An add-on that gives human support agents real-time suggested replies and instant translation while they work a ticket Fin didn't resolve.

• Custom AI models. Fin runs on Intercom's own Apex 1.0 and Apex Flash models, trained specifically on customer service interactions rather than a general-purpose foundation model, which the company says produces higher resolution rates and fewer hallucinations than off-the-shelf alternatives, a claim worth testing yourself rather than taking at face value.

• Enterprise security. SOC 2, ISO 27001, ISO 42001, HIPAA support, SSO, and regional data hosting in the US, EU, or Australia, which matters if support conversations touch regulated data.

How Fin Performs: The Vendor Number vs. The Real Number

Here's the thing about that 76% headline resolution rate: it's real, it's published, and it's also an average across 12,000+ very different businesses with very different help documentation. Your mileage depends entirely on how good your knowledge base is before Fin ever sees it.

Fin's own customer stories back up the top end of that range. Anthropic, which deployed Fin for its own support team, saved more than 1,700 hours of agent time in its first month and grew its resolution rate to 58% across roughly 50,000 monthly conversations, at one point lifting it 5.5 percentage points in a single week through a content-focused hackathon. Rocket Money reports a 68% resolution rate and about $1 million in annual ROI from Fin. Ecommerce customers like ZayZoon report resolution rates around 80% and savings in the millions.

But that's the curated case-study end of the distribution. Independent write-ups that ran their own ticket batches through Fin report a wider, messier range, with real-world resolution rates clustering anywhere from roughly 38% to 72% depending on setup quality, well short of the 76% headline. One outside review that tested Fin against 500 real support tickets landed at 38%. None of that makes Fin's number dishonest, an average is still an average, but it's a good reason to run your own pilot before committing budget, rather than assuming you'll land at the top of the range on day one.

Fin Pricing: What You'll Actually Pay

Fin uses outcome-based pricing, meaning you pay when it actually resolves something, not a flat seat fee for a tool that might sit idle. Verified directly on Fin's pricing page today:

A billable "outcome" is a resolution, a procedure handoff, or a disqualification at $0.99 each; a qualified sales lead is billed separately at $9.99. There's no charge for conversations Fin doesn't resolve. Eligible early-stage startups can get 93% off Intercom and a free year of Fin through Intercom's Early Stage Program, and every new account gets a 14-day free trial with unlimited outcomes and no credit card required.

Is it worth the price? For a team that already has decent help center content, yes, the math tends to work: paying under a dollar for a resolved ticket is cheap next to the fully loaded cost of a human agent handling the same conversation. For a team with thin or outdated documentation, budget for a slower ramp, since Fin's accuracy is only as good as what it's trained on.

Fin vs. Zendesk AI vs. Ada vs. Decagon

Fin isn't the only AI agent selling outcome-based pricing, but it's one of the only ones that actually publishes a number.

Our Zendesk AI review covers the closest direct comparison in more depth: Zendesk's AI is arguably a stronger pick if you're already committed to Zendesk's ticketing platform, since the AI is native rather than bolted on, but Fin's per-resolution price is lower and its pricing page doesn't require a sales call to read. Ada and Decagon both skip public pricing entirely, which tends to mean bigger minimum commitments and longer procurement cycles, fine for a large enterprise, overkill for most mid-market teams.

Who Should Use Fin

Best for:

• Support teams that already have solid help center documentation and want a fast, usage-based way to automate a real share of ticket volume.

• Startups and mid-market companies that want to test AI support without committing to a six-figure annual contract, thanks to the transparent pricing and free trial.

• Teams handling support across multiple channels (chat, email, voice, WhatsApp) that don't want to run a separate bot for each one.

Skip it if:

• Your knowledge base is thin, outdated, or scattered across tools Fin can't easily ingest. You'll pay for a lot of low-quality resolutions before it improves.

• You need a locked-in, multi-year vendor relationship with zero organizational risk. Fin is mid-acquisition, and while leadership says nothing changes before the deal closes in early 2027, that's still a live variable worth flagging to whoever signs the contract.

• You're already deep into Zendesk or Salesforce Service Cloud and would rather keep AI native to that platform than add a separate outcome-based layer.

Final Verdict

Fin earns its reputation honestly: it's one of the few AI customer service agents that tells you what it costs before you talk to a salesperson, backs that up with an actual money-back guarantee, and has the case studies (Anthropic among them) to show the resolution rates aren't purely marketing. The 76% headline number is real, but it's a company average, not a promise for your account on week one, so run your own pilot before you plan a budget around it.

The bigger open question isn't the product, it's the company underneath it. Fin renamed itself out of a decade and a half of brand equity, then agreed to be bought by Salesforce for $3.6 billion five weeks later. That's a strong signal the technology works. It's also a genuine reason to ask what happens to your account, your pricing, and your support relationship once Salesforce actually takes the wheel in 2027. If you can live with that uncertainty, the 14-day free trial is a low-risk way to find out whether Fin resolves your tickets as well as it resolves everyone else's.