Customer-AI pricing pages rarely say it this cleanly: not seats, not messages—$0.99 per outcome. Fin meters only when a conversation ends in a billable result.

On June 15, 2026, Salesforce signed a definitive agreement for roughly $3.6 billion to acquire Fin (formerly Intercom) into the Agentforce story. The deal is not closed; public guidance points to around Q4 of Salesforce’s FY2027. The product teardown is not M&A gossip. It is how outcome billing turns “resolution” into an invoice line—and how much a platform will pay for that unit of measure—especially whether Agentforce packaging later rewrites the definition.

For buyers, the sticker has a psychological edge: it sounds like “no help, no charge.” What you actually need to tear down is how the vendor defines “help,” and which way the bill skews when that definition sits with the seller.

Mechanism: an outcome is not “the ticket vanished”

Fin’s pricing page is blunt: with your current helpdesk (Salesforce, HubSpot, and others), the core meter is $0.99 / outcome, with a 50-outcome monthly minimum (help center: ~$49 / month base including 50, then $0.99 overage). On Intercom’s own helpdesk, seat fees stack on top (Essential from about $29 / seat / month; Advanced / Expert higher). At most one outcome per conversation, even if Fin answers many questions or runs many steps. Easier to forecast than per-message or per-tool-call—while the complexity hides inside “what counts as success.”

Help-center definitions are sharper than the pricing page. Billable types:

Type	Price	Rough trigger
Resolution	$0.99	After Fin’s last answer, the user confirms help—or asks for nothing further (including assumed resolution)
Procedure handoff	$0.99	Fin finishes a Procedure you configured to end in a human/workflow handoff
Disqualification	$0.99	Sales: prospect fails your criteria
Qualification	$9.99	Sales: prospect matches criteria and is routed (book a call, trial, self-serve onboarding, etc.)

Resolution bites on silence or confirmation after Fin answered—not on “ticket closed.” Help center:

• Confirmed: affirmative replies (“thanks,” “that helped”).
• Assumed: ~24 hours with no further ask after Fin’s last answer → billed.
• Not billed: greetings-only; clarifying question left unanswered (no resolution state—abandoned); explicit ask for a human; Procedure failure; workspace-default / frustration escalations.
• Deduction: charged resolution later reopens with more help needed—even across billing periods.
• Teammate joins: if the customer already got Fin’s answer and did not ask for more help/a person, the charge may already stand; if they asked for a human, usually not charged.

Procedure handoff vs ordinary escalation is the definition-power miss:

Scenario	Status	Billed?
Procedure guidance / @handoff routes to a team	Procedure handoff	Yes ($0.99)
User asks for a human / frustration default	Escalation	No
Workspace escalation rules (Train → Escalation)	Escalation	No
Procedure technical/logic failure	Escalation	No
Clarifying question, user silent, auto-close	Abandoned	No

Direct implication: put “hand to human” inside a Procedure Fin completes → billable outcome; rely on global escalation for hard tickets → usually free. Configuration rewrites the bill, not only the support experience.

Getmacha’s sharpest point is the assumed resolution: silence after Fin’s last answer can still bill. The invoice meters platform-defined outcomes, not CSAT. Treating “resolution rate” as “customers were happy” systematically overstates ROI. In week one, put assumed-resolution share, reopen deductions, and “Fin answered but the user went elsewhere” on the same page—or your deflection chart will look healthier than the floor. Help center even tips: to avoid “left angry but quiet” becoming assumed resolution, configure frustration escalations so default behavior is a non-billable escalation, not silent billing.

The other miss is sales: Qualification is $9.99, 10× a normal outcome; Fin for Sales qualification-style outcomes are mainly documented on Intercom’s own helpdesk—not as clean parity on Fin for platforms. Harmless for pure support; lethal to a model that only multiplies deflection × $0.99 once lead routing is on.

Business model: seat + outcome all-in math

The public pitch: no setup, integration, or platform fees—almost all variable cost sits on outcomes; on external helpdesks, “unlimited teammates, no AI seat fee” is part of the story. Help-center floor math is blunt: resolve 5 tickets in a month and you still pay $49 (inside the included 50); resolve 100 and you pay $99.50 (50 included + 50 × $0.99). Low-volume teams are buying a minimum commitment, not true pay-as-you-go-to-zero.

Stack seats (Getmacha-style: resolutions only, no $9.99 quals; Advanced seats ~$85 annual):

Scale	Path	Seats + outcomes	~Monthly	Fin share
External trial	Platforms	0 seats + 50-outcome floor	$49	100% (almost all floor)
External, real volume	Platforms	0 seats + 500	~$495	100%
Small on Intercom	Intercom	3 seats + 500	~$750 ($255 seats + $495 Fin)	~66%
Mid	Intercom	10 seats + 2,000	~$2,830 ($850 + $1,980)	~70%
High	Intercom	25 seats + 10,000	Fin alone $9,900—can run ~3× seat cost	Fin-dominated

Two formulas:

• External helpdesk (Fin for platforms): Fin ≈ max(50, billable outcomes) × $0.99 (public story: no AI seat fee; minimum applies).
• On Intercom: all-in ≈ seats × seat price + outcomes × $0.99 (public Intercom path usually does not stress the same 50-outcome floor; unit price stays $0.99).

No public volume break, no rollover—scale makes the linear slope sting. That curve fights seat-priced AI: seats rise with headcount; outcomes rise almost linearly with conversation volume. The more traffic you successfully hand Fin, the more visible the variable line becomes—by design. Buyers vote with cash on whether automation is worth it; ARR tracks penetration instead of stalling at a seat ceiling.

Also stack Copilot (agent-side)—about $35 / user / month on the public page (third-party annual quotes sometimes land lower—trust your account). Getmacha’s mid scenario still adds hundreds of dollars/month for 10 Copilot seats. Customer-facing Fin + agent-facing Copilot is two meters on one stack. TCO models that merge them into one “AI budget” land closer to reality.

Comparisons help negotiation. Intercom’s own public pricing comparison puts Salesforce Agentforce list at about $2 / conversation (or a Flex Credits path). Fin counters with “pay for success, no platform fee.” Macro industry ranges (~$0.50–$0.70 AI interaction vs ~$6–$8 human) are context, not Fin contract numbers—trial outcome counts still belong in the appendix. The 14-day unlimited, no-card trial exists to get that count—not to put the marketed ~76% autonomous handling / resolution figure (vendor average; includes assumed resolutions) on the ROI slide.

Salesforce packaging risk: buying the meter, or the channel?

Salesforce’s announcement and TechCrunch-style coverage align on ~$3.6B, with Fin as the packaged, fast-to-deploy customer agent beside Agentforce’s deeper build. Product reality today: Fin already runs on Salesforce as a helpdesk—the deal is not “first contact with Salesforce,” it is folding a market-validated outcome meter onto the Agentforce shelf.

Read packaging risk on three lines:

1. SKU rewrite: today fin.ai is a clear $0.99 / outcome + ~$49 external floor. On an Agentforce price list, common paths are bundles (seat packs, cloud credits, industry packs), tiers, and rewritten minimums. Multi-year deals that lock “today’s unit price” without locking “outcome definition and metering boundaries” lock a moving target.
2. Definition drifts upward: assumed-resolution window, whether Procedure handoffs bill, whether sales Qualification opens on the Salesforce path—these belong in the contract appendix by name. Vendor resolution rate already includes confirmed and assumed; if Agentforce packaging changes the gloss, buyer ROI models break.
3. Price-pressure from the peer meter: public comps put Agentforce around ~$2 / conversation; Fin’s narrative edge is “success only, ~$0.99.” If the acquirer wants the two meters to converge, buyers may see “outcomes get dearer / conversations get cheaper / fold into credits”—all three need re-eval clauses, not faith that linear friendliness survives forever.

The deal sells scalable outcomes; multi-year buyers should watch whether Agentforce packaging rewrites the definition. “Price looks unchanged pre-close” does not mean “packaging stays buyer-friendly post-close.”

Where intuition fails

Intuition 1: per-outcome always beats seats.
Often true at low volume; at high volume with no list discount, outcomes can swallow the seat story. Reframe: estimate AI-eligible conversations × an honest resolution rate before picking a shape. Mid-tier math: 2,000 × $0.99 = $1,980—already more than double ten Advanced seats. On the external path, well under 50 outcomes/month, you pay the floor—not true pay-to-zero.

Intuition 2: vendor resolution rate belongs on the ROI slide.
76% is a coverage average that includes assumed resolutions. Thin KBs and messy emotion tickets need trial numbers in the contract appendix.

Intuition 3: price is frozen until close—and only improves after.
$0.99 is still on fin.ai today; Agentforce packaging often means bundles, tiers, and minimums. Multi-year deals need re-eval clauses on outcome definition and unit price.

Sources

• Fin AI Agent Pricing (fin.ai)
• Fin pricing: Outcomes (Fin Help)
• Intercom Fin Pricing: Per-Resolution Cost Breakdown (Getmacha)