Intercom Fin Review (July 2026): Pricing, Capabilities & Verdict

What is Intercom Fin in July 2026?

Intercom Fin is the customer-facing AI agent built by Intercom — rebranded as Fin in May 2026 and announced for acquisition by Salesforce on June 15, 2026. It answers support questions, executes multi-step workflows called Procedures, and hands off to humans when it’s not confident. Pricing is $0.99 per successful outcome, and the product runs on Fin’s proprietary Apex 1.0 model rather than a wrapped third-party LLM.

I spent the last two weeks of July 2026 going through Intercom’s pricing pages, the Fin product site, the public updates feed, the help-center docs on outcomes, the benchmarks tool, the trust center, customer stories, and the Wikipedia entry that traces the company’s history. My read: Fin is no longer “Intercom’s chatbot.” It’s the most explicitly priced, most aggressively benchmarked outcome-based AI agent on the market right now. Whether that’s a fit for you depends almost entirely on the quality of your knowledge base and how disciplined your team is about escalation design.

Pull quote: “Intercom [launched Fin] with a disruptive outcome-based pricing model, which they launched in early 2023…a year ahead of most competitors.” — Kyle Poyar, Editor at Growth Unhinged, quoted on the Fin homepage.

How much does Intercom Fin cost in 2026?

Intercom Fin costs $0.99 per outcome across every Intercom plan and as a standalone agent on your existing helpdesk, with a 50-outcome monthly minimum on the standalone SKU. Outcomes are resolutions, Procedure handoffs, and disqualifications. A successful Sales qualification is billed separately at $9.99 because it ties to a routed lead. There are no separate seat fees on the standalone SKU, no integration fees, and no platform fees.

The full pricing block, drawn from fin.ai/pricing and intercom.com/pricing:

• Fin with any helpdesk (Salesforce, HubSpot, Freshdesk, etc.): $0.99 per outcome, 50 outcomes/month minimum, no seats.
• Fin with Intercom: $0.99 per outcome, plus seat-based Intercom plans starting around $29/seat/month on Advanced.
• Copilot (inbox AI for human agents): $29 per agent/month billed annually, or $35/month-to-month. Each agent also gets 10 free Copilot conversations and 10 free AI Auto-translation conversations per month.
• Pro add-on (AI quality management): $99/month for analysis of 1,000 conversations. Includes CX Score, Topics, Recommendations, Monitors, and Custom Scorecards.
• Fin Voice: Custom pricing via sales.
• Outcome type: Qualification: $9.99 each (when Fin matches a prospect to your routing criteria and sends them to a team, trial, or self-serve flow).
• Trial: 14-day free trial with no credit card and unlimited Fin outcomes during the trial.

A small but important change: on March 12, 2026, Intercom renamed its core pricing metric from “Resolutions” to “Outcomes” to cover Procedure handoffs and Sales qualifications in the same unit (Fin product updates). If you’re comparing quotes from late 2025 against July 2026 quotes, normalize on “outcome” — not “resolution.”

How good is Intercom Fin at actually resolving support tickets?

Fin averages a 76% resolution rate across more than 12,000 customers, with top-10 performers in each industry hitting 85% resolution and 91% involvement. Intercom now lists Fin as the leader on G2 in the AI Customer Support Agents category with a 90% satisfaction score, beating the next vendor by 8 percentage points.

The cross-verified performance numbers I could confirm:

• 76% average resolution rate across 12,000+ customers (fin.ai).
• 2 million Fin resolutions per week company-wide (fin.ai).
• 78% / 85% / 91% automation / resolution / involvement averages for top-10 performers across 15 industries, based on 110M+ conversations (fin.ai/benchmarks).
• Top industry: Software & Technology at 98.0% automation, 98.2% resolution, 99.8% involvement among top performers (fin.ai/benchmarks).
• Anthropic customer story: 79% Fin resolution rate, 80% involvement, 63% automation, 560,000+ resolutions/month (Fin customer story).
• Best-case benchmark: Fin Apex 1.0 reports 2.8% higher resolution rate, 0.6s faster time-to-first-token, and a 65% reduction in hallucinations versus Anthropic’s Sonnet 4.6 (fin.ai).

The honest caveat: those are the company’s own numbers. The 76% average mixes software giants with regulated enterprises and a long tail of mid-market customers. In the same set of disclosures, the Fin trust center promises a $1M payout if a qualifying high-volume customer (over 250,000 monthly conversations, North America or Europe, currently Intercom-only) signs up for the Guarantee program and Fin fails to hit 65% resolution. That guarantee is a useful pricing signal — it’s the resolution rate Intercom is willing to put money behind.

What can Intercom Fin actually do?

Fin can answer questions grounded in your content, execute multi-step Procedures that touch third-party systems, hand off to humans with full context, and continuously improve from conversation data. In July 2026 it works across chat, email, voice, WhatsApp, SMS, Slack, Discord, Telegram, Instagram, Facebook Messenger, and an API surface for custom apps.

The product is organized around what Fin calls the Flywheel: Train, Test, Deploy, Analyze (fin.ai/capabilities). Each stage has a real tool behind it:

1. Train. Procedures (multi-step workflows written in natural language with branching logic, Python code blocks, and webhook pauses), Data Connectors (Shopify, Stripe, Salesforce, internal APIs), Knowledge Sources (Help Center, PDFs, URLs, snippets), Guidance (tone, vocabulary, style rules), Multilingual Fin (45+ languages with real-time translation).
2. Test. Simulations (run full AI-vs-AI conversations before going live), regression tests, audience testing, and Previews.
3. Deploy. Voice, email, Messenger, Slack, WhatsApp, SMS, Instagram, Facebook Messenger, Telegram, Discord, plus a Fin over API surface.
4. Analyze. Topics Explorer (AI-clustered topics and subtopics), CX Score (an AI quality score that doesn’t need surveys), AI Suggestions (one-click content improvements), Monitors (always-on QA with Custom Scorecards), and the Optimize Dashboard.

The most consequential feature shipped in 2026 is Procedures (fin.ai/procedures). Procedures let you describe a process the same way you’d brief a new hire — “verify the order, check the return window, offer store credit or refund, confirm with the customer, log the outcome” — and Fin runs it. You can add deterministic if/else branches, run Python for date math or eligibility, pause for a webhook (identity check, payment, bank linking), or “Ask a teammate” mid-flow. Procedures are the difference between Fin answering questions and Fin actually doing work.

Two other 2026 releases worth knowing about:

• Fin Apex 1.0 (March 2026) is the first customer-service model Intercom has post-trained itself rather than wrapping a third-party LLM (Fin updates; VentureBeat).
• Fin Operator (May 2026) is an internal agent that watches Fin’s performance, flags problems, tunes Fin, and keeps the knowledge base current (Fin updates; Wikipedia).
• Fin Voice 2 (June 2026) ships Apex Flash, a latency-tuned sibling of Apex 1.0, with claims of a 24.5% resolution lift over the prior Voice model (Fin updates).
• Fin Memory (July 2026) remembers prior conversations and customer context across channels (Fin updates).

How does Intercom Fin compare to Zendesk AI, Ada, and Decagon?

Fin is more expensive per conversation than most peers but is priced on outcomes, not seats or messages, and runs the most aggressive benchmark claims in the category. Zendesk AI is bundled inside a broader helpdesk; Ada is more enterprise-suite-shaped; Decagon competes hard on autonomous workflows. The right choice depends on whether you want a customer agent as a product (Fin, Decagon) or AI features inside an existing suite (Zendesk AI, Salesforce Agentforce).

A side-by-side view based on what each vendor publishes:

Dimension	Intercom Fin	Zendesk AI	Ada	Decagon
Primary pricing unit	$0.99 per outcome	Bundled with Zendesk Suite/AI add-on	Custom, conversation-based	Custom, conversation-based
Own model	Yes — Fin Apex 1.0 + Apex Flash	No (uses third-party LLMs)	No (multi-LLM orchestration)	No (multi-LLM orchestration)
Helpdesk flexibility	Intercom, Salesforce, HubSpot, Freshdesk, Dixa, Front, Zoho, Sprinklr, Gorgias	Zendesk-first	Primarily standalone	Primarily standalone
Outcome guarantee	Yes — $1M if below 65% resolution for qualifying customers	No public equivalent	No public equivalent	No public equivalent
Compliance posture	SOC 2 Type II, ISO 27001, ISO 42001, AIUC-1, ISO 27701, HIPAA	SOC 2, ISO 27001 (publicly stated)	SOC 2 (publicly stated)	SOC 2 (publicly stated)
Voice agent	Fin Voice 2 (Apex Flash)	Zendesk Voice	Voice in higher tiers	Voice in higher tiers
G2 category leadership	Leader in AI Customer Support Agents, 90% satisfaction	Strong in helpdesk	Strong in mid-market CX	Strong in autonomous AI

Source notes: Fin column from fin.ai/pricing, fin.ai/trust-reliability, and fin.ai/benchmarks. Competitor columns are summary-level and not deep-dive verified in this review; treat them as orientation, not a procurement matrix.

The defining choice for me is the pricing model. Fin charges per outcome; Zendesk AI and most peers charge per seat or per conversation. If your contact volume is lumpy — seasonality, product launches, viral tickets — outcome pricing means Fin costs scale with the value you get, not the volume you receive. If your volume is steady and your team is small, a seat-bundled suite can be cheaper.

Who should buy Intercom Fin in 2026?

Buy Fin if your support volume is high enough that outcomes become cheaper than seats, your knowledge base has an owner, and you either live on Intercom or are willing to run Fin on top of Salesforce, HubSpot, or Freshdesk. Skip it if your help content is a mess, you only want an in-inbox copilot, or your workload demands air-gapped, on-prem deployment.

Ideal users:

• Support leaders ready to invest in knowledge ops, not only a bot toggle. Procedures and Simulations make the difference; both require real content.
• Product-led SaaS and ecommerce with repetitive how-to, billing, shipping, and account questions — exactly the categories where Fin’s 76–85% resolution band holds up.
• Salesforce, HubSpot, or Freshdesk shops that want a customer agent without ripping out their helpdesk. Fin Standalone goes live “in under an hour” per Fin’s own documentation (fin.ai/pricing).
• Regulated industries that need a vendor with HIPAA, ISO 27701, SOC 2 Type II, and ISO 42001 — Fin checks each of those boxes (fin.ai/trust-reliability).
• Outcome-pricing curious teams who want the cost line to move with value, not with seat counts.

Non-ideal users:

• Teams with chaotic knowledge bases and no content owner. Fin won’t fix bad help articles — it’ll surface them faster.
• Buyers who want a copilot-only AI that summarizes tickets for human agents. That’s Copilot, not Fin.
• Organizations that need on-prem or air-gapped AI. Fin is a hosted SaaS; data residency is US, EU, or Australia (fin.ai/trust-reliability).
• Companies unwilling to monitor transcripts, CX Score, and reopen rates. Without that loop, even the best model drifts.
• Anyone uneasy about the Salesforce acquisition closing. The deal was announced June 15, 2026 (The Irish Times). Pricing and product names held steady through July 2026, but roadmap signals will get noisier until close.

How do you pilot Intercom Fin without burning budget?

Treat Fin like a 30-day operational pilot, not a one-week demo: clean your knowledge base, scope two or three low-risk categories, wire up Simulations, and gate every Procedures change behind regression tests. The cheapest wins come from restricting scope, not from tuning the model.

A pilot playbook worth following:

1. Clean the top 20 help articles. Fin’s ceiling is your content’s ceiling.
2. Pick two categories. “Where is my order” and “how do I reset my password” are the classics; both have high volume, low risk, and unambiguous ground truth.
3. Set hard usage limits. Fin lets you set usage reminders and hard caps (intercom.com/help outcomes article). Set both.
4. Use Simulations before going live. Build a Simulation library for your top intents and rerun it whenever you change Guidance or Procedures.
5. Watch the first 100 transcripts. The first week tells you more than any benchmark.
6. Track CX Score, not just resolution rate. A polite deflection isn’t the same as a satisfied customer.

The Anthropic customer story on Fin’s own site is a good template for what disciplined operation looks like in practice: weekly reviews of unanswered questions, quarterly content sprints, hackathons that produced a 10% resolution lift in a week, and a dedicated content/knowledge manager (Fin customer story).

What about governance, data, and the Salesforce deal?

Fin’s governance posture is mature: SOC 2 Type II, ISO 27001, ISO 42001, ISO 27701, AIUC-1, and HIPAA, backed by a 99.8% SLA with multi-region redundancy. The unanswered question for buyers in July 2026 is what the Salesforce acquisition — announced June 15, 2026 — means for the product.

Three concrete things worth pinning down before you sign anything:

• Salesforce acquisition context. On June 15, 2026, Salesforce announced an agreement to acquire Fin (formerly Intercom) for approximately $3.6 billion, expected to close before the end of January 2027 (The Irish Times; Wikipedia). The Intercom/Fin brands, pricing, and product names held steady through July 2026, but Salesforce already has its own AI agent (Agentforce). Ask for written assurances on pricing continuity and product roadmap.
• Data handling. Fin’s third-party providers are contractually restricted from using your data for training; Fin itself can use anonymized data for fine-tuning unless you opt out (data is deleted within 30 days) (fin.ai/trust-reliability). Data residency: US, EU, or Australia.
• Hallucination controls. The Fin AI Engine validates queries, retrieves grounded content, reranks for accuracy, and does a final validation pass; if confidence falls below thresholds, Fin escalates rather than guessing (fin.ai/trust-reliability). Fin reports a 65% reduction in hallucinations versus Sonnet 4.6 in its own benchmarks (fin.ai).

Verdict: is Intercom Fin worth it in 2026?

Yes for the right buyer, with caveats. Fin is the most aggressive outcome-priced, most benchmarked, most helpdesk-agnostic AI customer agent you can buy in July 2026. If your team has clean content, a clear escalation policy, and the discipline to monitor transcripts, Fin will return more value per dollar than any seat-bundled AI add-on.

Hold off if your knowledge base is broken, you only need an in-inbox copilot, your data can’t leave your VPC, or the Salesforce deal makes a multi-year commitment feel risky.

Score: 86/100 — strong on practical value, product quality, and differentiation; weaker on transparency (the headline 76% is self-reported) and on support docs for non-Intercom helpdesks.

Frequently asked questions about Intercom Fin

What is Intercom Fin?

Intercom Fin is an AI customer agent built by Intercom (rebranded to Fin in May 2026) that resolves support tickets end-to-end, executes multi-step Procedures, and hands off to humans with context. Pricing is $0.99 per successful outcome (fin.ai/pricing).

How much does Intercom Fin cost per resolution?

$0.99 per outcome for resolutions, Procedure handoffs, and disqualifications, with a 50-outcome monthly minimum on the standalone SKU. A Sales qualification is $9.99 because it ties to a routed lead (intercom.com/pricing; Intercom help on outcomes).

Is Intercom Fin worth it in 2026?

If your contact volume is high, your knowledge base has an owner, and you either live on Intercom or want to add AI to Salesforce, HubSpot, or Freshdesk — yes. Fin’s published resolution rate is 76% across 12,000+ customers, with top-10 performers hitting 85% (fin.ai; fin.ai/benchmarks).

What is the difference between Intercom Fin and Copilot?

Fin is the customer-facing AI agent that talks to your end users. Copilot is the in-inbox AI assistant for your human support agents, priced at $29/agent/month on annual billing (intercom.com/pricing). They’re complementary, not substitutes.

Can Fin work with Salesforce or HubSpot?

Yes. Fin Standalone integrates with Salesforce Service Cloud, HubSpot, Freshdesk, Dixa, Front, Zoho, Sprinklr, and Gorgias, and is sold for $0.99 per outcome with no seats (fin.ai/pricing). Intercom itself was rebranded as Fin in May 2026.

What certifications does Intercom Fin hold?

SOC 2 Type II, ISO 27001, ISO 27701, ISO 27018, ISO 42001, HIPAA, GDPR/CCPA, and AIUC-1, documented at fin.ai/trust-reliability and trust.intercom.com.

Is Fin the same as the new company called Fin?

Yes. Intercom, Inc. rebranded as Fin in May 2026 and announced a Salesforce acquisition on June 15, 2026 (The Irish Times; Wikipedia). The product line and pricing held steady through July 2026.

What is the Fin Million Dollar Guarantee?

For qualifying high-volume customers (over 250,000 monthly conversations in North America or Europe) who join the Fin Guarantee Success Program, Intercom guarantees at least a 65% resolution rate or pays $1 million (fin.ai/trust-reliability). Phase one is for customers on Intercom.

What is Fin Apex 1.0?

Fin Apex 1.0 is Intercom’s first post-trained customer-service language model, launched in March 2026. It generates Fin’s final answer and, per Intercom, beats GPT-5.4 and Claude Sonnet 4.6 on internal customer-service resolution benchmarks (VentureBeat; fin.ai/capabilities). Apex Flash is a latency-tuned sibling used in Fin Voice 2.

What languages does Intercom Fin support?

45+ languages with real-time translation, plus the ability to control which languages you serve per region (fin.ai/updates).