AI Support’s Per-Resolution Pricing, Decoded: Intercom Fin vs Zendesk AI vs Freshdesk

Disclosure: This article contains no affiliate links. None of the vendors covered — Intercom (Fin), Zendesk, or Freshworks — is an affiliate partner of Automated Wealth Lab, and we earn nothing from any of them. Every price below was checked against published pricing pages in October 2026; where a number rests on third-party reporting rather than a page we could read directly, we say so.

“Per-resolution pricing” is the pitch of the year in support software: don’t pay for seats or messages, pay only when the AI actually solves something. It sounds like the fairest pricing model ever invented. And it can be — if the word “resolution” means what you think it means. Across the three biggest vendors, it does not mean the same thing once. One of them bills you when the customer goes silent. One bills you only when its own verification check agrees the issue is solved. One doesn’t bill per resolution at all, but per conversation window, win or lose.

This article decodes the three models from the vendors’ own pricing pages — Intercom’s Fin, Zendesk’s AI agents, and Freshdesk’s Freddy AI Agent — and then runs one honest model across all three: the same ticket volumes (500, 2,000, and 10,000 tickets a month), the same deflection assumption, the same team sizes, so the totals can actually be compared. Every assumption is printed. None of the deflection rates below is a vendor promise; they are planning assumptions, labeled as such, because the single most common budgeting error in this category is typing a marketing department’s resolution rate into a spreadsheet.

One thing to know before the numbers: every vendor in this article writes the definition of its own billable unit. That is not corruption, it is just the model — but it means the definition section of each pricing page is worth more of your attention than the price itself.

The short answer: Intercom Fin charges $0.99 per outcome — including conversations where the customer simply stopped replying. Zendesk charges $1.50 (committed) to $2.00 (pay-as-you-go) per automated resolution, but since May 2026 counts only resolutions its own checks verify. Freshdesk’s Freddy charges $0.49 per session — per conversation window, resolved or not. At 2,000 tickets a month and a 40% planning deflection rate, the modeled totals are Fin $937, Freshdesk $1,125, and Zendesk $1,437.50 (committed). The unit definition matters more than the sticker price.

Three vendors, three different billable units

Vendor	Billable unit	Unit price	What triggers a charge	Base cost underneath
Intercom Fin	An “outcome” (max 1 per conversation)	$0.99 (qualification outcomes: $9.99)	Resolution (confirmed or assumed), procedure handoff, disqualification, self-serve routing	Intercom seats: $29/$85/$132 per seat/mo annual — or no seats on another helpdesk, with minimum commitments
Zendesk AI agents	An automated resolution (verified only, since May 2026)	$1.50 committed / $2.00 pay-as-you-go, after a per-agent allowance	The AI resolves a request without human escalation, and Zendesk’s verification confirms it	Suite seats: $55 (Team) / $115 (Professional) per agent/mo annual; Support Team $19
Freshdesk Freddy AI Agent	A session (24h window for chat, 72h for email)	$49 per 100 sessions = $0.49	The AI agent interacts with a customer at all — resolved or escalated	Freshdesk Omni seats: $29/$79/$119 per agent/mo annual

Keep this table in mind through the model below. Fin’s unit is the broadest (four different events bill, one of them is silence). Zendesk’s is the narrowest (a verified solve, nothing else). Freddy’s is not an outcome at all — it is a meter on attempts. Everything about how these bills behave at scale follows from those three sentences.

Intercom Fin: $0.99 per outcome, and “outcome” is doing heavy lifting

Fin’s pricing page states the model plainly: $0.99 per outcome, charged at most once per conversation, with no setup, integration, or platform fees, and a 14-day free trial with unlimited outcomes. The definition is where it gets interesting. A billable outcome is any of:

• Resolution — “no further help is requested after Fin’s last answer.” Fin’s own materials split this into confirmed resolutions (the customer says it helped) and assumed resolutions (the customer leaves without asking again). Both cost $0.99.
• Procedure handoff — Fin completes a workflow you configured that ends in handing off to a human. Note: a human still does the work afterwards; you pay $0.99 for the handoff itself.
• Disqualification and self-serve routing — sales-side outcomes, $0.99 each.
• Qualification — Fin qualifies a sales lead: $9.99, ten times the support price. Irrelevant to pure support teams; a budget hazard if Fin touches your sales chat.

The genuinely fair part: if Fin cannot resolve and passes the conversation to your team without one of those outcomes, you pay nothing. The structural problem is the assumed resolution. A customer who reads a wrong answer, gives up, and closes the tab looks identical to a customer whose problem was solved — both stopped asking. Fin is not being dishonest here; the definition is printed on its pricing page. But it means your billable “resolutions” will always exceed your true solved count by some margin you can only discover by auditing outcomes against CSAT in the first months.

Underneath the outcomes sit the seats, if you run Fin inside Intercom: Essential $29, Advanced $85, Expert $132 per seat/month on annual billing ($39/$99/$139 monthly). Copilot — the AI assistant for your human agents — is a separate $29/agent/month (annual; $35 monthly). Fin also runs on other helpdesks (Salesforce, HubSpot, Zendesk, Freshworks, and others) at the same $0.99 with no seat cost, but there minimum commitments apply — reported as a 50-outcome monthly minimum, about $49.50, on standalone deployments. Two more budgeting notes: Intercom has been price-testing Essential (a $19 seat price has been spotted in the wild), so read your seat price off the page the day you sign; and Fin’s parent company changed hands in 2026 — Salesforce completed its acquisition of Intercom (now branded Fin) on September 10, 2026, for roughly $3.6 billion. Pricing was unchanged in checks after the close, but a multi-year contract signed today is a bet on another company’s packaging roadmap.

Zendesk AI: the strictest definition, the highest unit price

Zendesk includes AI agents in every Suite and Support plan and bills them per automated resolution: a customer request the AI resolves without escalation to a human. Seats run $19 (Support Team), $55 (Suite Team), and $115 (Suite Professional) per agent/month paid yearly ($25/$69/$149 monthly), and the Copilot add-on for human agents is a separate $50/agent/month.

Each plan includes a small monthly allowance of resolutions per agent — 5 per agent on Support Team and Suite Team, 10 per agent on Suite Professional. Beyond the allowance, resolutions cost $1.50 each on a committed volume or $2.00 pay-as-you-go. (Zendesk’s pricing page presents these rates inside its plan comparison; we corroborated them across multiple current pricing breakdowns, as noted in the Methodology.)

The definition, though, is where Zendesk now beats Fin. Until 2026, Zendesk effectively counted a resolution when the customer went quiet — the old model treated 72 hours of silence as success, the same silence-billing Fin still uses. Since May 2026, only Verified Resolutions are billable: Zendesk runs an LLM-based check on whether the customer’s issue was actually resolved, and the other outcomes — “Assisted Escalations” (the AI helped, a human finished) and “Contained Resolutions” — are free. That is the most honest meter in this article, and it is not close. The trade is the price: at $1.50–$2.00, a verified Zendesk resolution costs 1.5–2× a Fin outcome, and the per-agent allowance is small enough to be a rounding error at real volume (a 15-agent team gets 75 included resolutions against thousands resolved). Also budget for the fact that testing your AI agent on a live account consumes resolutions too — keep experiments in a sandbox. And treat Zendesk’s own marketing line on its pricing page — agents that “resolve up to 80%+ of even the most complex service issues” — as what it is: a vendor claim, not a planning number.

Freshdesk Freddy: you pay per session, win or lose

Freshworks declined to play the resolution game entirely. Freddy AI Agent bills per session, at $49 per 100 sessions ($0.49 each), on top of Freshdesk seats — $29/$79/$119 per agent/month (annual) on Freshdesk Omni, with the classic Freshdesk Support Desk tiers starting lower (Growth $19/agent/month) under the same Freddy session pricing. Freddy AI Copilot for human agents is another $29/agent/month, and only on Pro and Enterprise plans.

A session, per Freshworks’ own pricing-page FAQ, is all interactions between a customer and the AI agent within 24 hours of initiation for chat, or a 72-hour window for email. Follow-ups inside the window are one session; a customer who returns on day two of a chat issue, or day four of an email thread, starts a new one. The pricing page also advertises “first 500 sessions included” on Growth, Pro, and Enterprise — and the FAQ on the same page quietly adds the two words that change the math: the 500 free sessions are limited to once per account. They are a trial budget, not a monthly allowance. When they are gone, every session is $0.49, whether the AI resolved the issue, escalated it, or produced an answer the customer ignored. Session packs expire at the end of your billing cycle with no rollover, and per Freshworks’ session documentation, even building and previewing the agent consumes sessions — your testing comes out of the same wallet.

So Freddy’s meter is the cheapest per unit ($0.49) and the least outcome-sensitive: it is a per-attempt price. Whether that is a good deal depends entirely on your resolution rate — which is exactly what the model below makes visible.

The model: 500, 2,000, and 10,000 tickets a month

Planning assumptions (ours, not the vendors’):

• The AI resolves 40% of tickets. This is a planning assumption for a competent deployment on typical SaaS/e-commerce questions — not a benchmark and not any vendor’s claim. The sensitivity table below shows 25% and 60%.
• Reference teams: 2 agents at 500 tickets/month, 5 at 2,000, 15 at 10,000.
• Annual-billing rates throughout. Fin is modeled on Intercom Essential seats ($29). Zendesk on Suite Team ($55, $1.50 committed rate). Freshdesk on Omni Growth ($29) with the AI agent as front door, at roughly one session per ticket (the 24h/72h windows merge most follow-ups).
• Freshdesk is modeled at steady state: the one-time 500 free sessions are excluded. In month one, they would cut the 500-ticket scenario’s session cost from $245 to $0.

Tickets/month	Intercom Fin (Essential seats)	Zendesk Suite Team (committed)	Freshdesk Omni Growth
500 (200 AI-resolved)	200 outcomes × $0.99 = $198 + $58 seats = $256	190 billable* × $1.50 = $285 + $110 seats = $395	500 sessions × $0.49 = $245 + $58 seats = $303
2,000 (800 AI-resolved)	800 × $0.99 = $792 + $145 seats = $937	775 billable* × $1.50 = $1,162.50 + $275 seats = $1,437.50	2,000 sessions = $980 + $145 seats = $1,125
10,000 (4,000 AI-resolved)	4,000 × $0.99 = $3,960 + $435 seats = $4,395	3,925 billable* × $1.50 = $5,887.50 + $825 seats = $6,712.50	10,000 sessions = $4,900 + $435 seats = $5,335

*Zendesk billable = AI-resolved minus the included allowance (5/agent/month: 10, 25, and 75 resolutions at our team sizes). At the $2.00 pay-as-you-go rate instead of committed, Zendesk’s totals become $490 / $1,825 / $8,675.

Three readings of this table matter more than the ranking. First, the AI line, not the seats, becomes the bill: at 10,000 tickets, Fin’s outcomes are 90% of its total; Zendesk’s resolutions are 88% of its total. Headcount-based budgeting will miss this completely. Second, the effective cost per AI-resolved ticket at our 40% assumption is: Fin $0.99 (by definition), Zendesk $1.50–$2.00, Freshdesk $1.23 ($0.49 per session ÷ 40% resolution). Freddy’s cheap unit is not the cheapest outcome, because you fund the 60% of sessions that escalate. Third, the ranking is assumption-sensitive. Push deflection to 60% at 2,000 tickets and Fin rises to $1,333 (1,200 outcomes) while Freshdesk stays $1,125 — the session model wins when the AI resolves more, because its cost is pinned to tickets, not wins. Drop deflection to 25% and Fin falls to $640 (500 outcomes) while Freshdesk still charges for every attempt. Fin’s bill breathes with performance in both directions; Freddy’s only breathes with volume; Zendesk’s breathes with verified performance, at the highest rate.

The incentive traps, collected

• The vendor grades its own homework. All three define the billable unit, run the counter, and send the invoice. Zendesk’s May 2026 move to verified-only billing is the one structural reform in the category that reduces the conflict — the check is still Zendesk’s, but at least “customer gave up” no longer bills.
• Silence bills at Fin. An assumed resolution is a customer who stopped replying. Some were helped; some gave up; some found a competitor. All cost $0.99. Audit billable outcomes against satisfaction scores before you trust any deflection ROI.
• The success tax. Per-outcome pricing means the better the AI performs, the more you pay — the opposite of a saving until resolution quality is proven. Budget AI support as a rising cost curve tied to deflection, not as a fixed saving.
• Allowances are sized to look meaningful. Zendesk’s 5–10 included resolutions per agent cover 2–4% of a working deployment’s resolutions. The committed-vs-pay-as-you-go spread ($1.50 vs $2.00) is a 33% charge for flexibility.
• The 500 “free” sessions are a one-time trial. Freshdesk’s most misread line. Month two is when the real Freddy bill starts, and session packs expire whether or not you used them — a December slump doesn’t bank credit for a January spike.
• Windows cut both ways. Freddy’s 24-hour chat window can split one slow issue into two sessions; its 72-hour email window can merge two issues into one. Your session count is partly an artifact of customer reply speed, which you do not control.
• Every vendor sells a second AI. Fin’s Copilot ($29/agent), Zendesk’s Copilot ($50/agent), Freddy Copilot ($29/agent) — the agent-facing assistant is always a separate line, priced per human seat, on top of the customer-facing meter.

FAQ

What exactly counts as a Fin “resolution”?

Per Fin’s pricing page: no further help is requested after Fin’s last answer. That covers a customer explicitly confirming they were helped and a customer who simply leaves. You are charged once per conversation at most, and not at all when Fin hands off to your team without completing an outcome.

Is Zendesk’s per-resolution price actually published?

The structure is on Zendesk’s pricing page — AI agents included in every plan, billed per automated resolution, with per-agent allowances — and current pricing breakdowns put the rates at $1.50 per resolution on committed volume and $2.00 pay-as-you-go, with 5 included resolutions per agent/month on Team plans and 10 on Professional. Confirm the rate in your quote before budgeting; Zendesk has changed this model twice in two years.

Why do people misread Freshdesk’s 500 free sessions?

Because “first 500 sessions included” reads as a monthly allowance, and Freshworks’ own FAQ clarifies it is once per account, ever. At 2,000 tickets a month, the free sessions cover about a week of AI traffic.

Which vendor is cheapest at 2,000 tickets a month?

Under our stated assumptions, Fin at $937/month, then Freshdesk at $1,125, then Zendesk at $1,437.50 committed. Change the deflection rate, the team size, or the share of tickets the AI touches, and the order can change — the model above shows exactly which lever moves which bill.

Can I cap what the AI spends?

Partially. Fin offers configurable usage alerts, limits, and real-time reporting. Freshdesk’s session packs are a hard cap by construction — when packs run out, the agent stops, which is a service risk of its own. Zendesk’s committed volumes fix the unit price, not the total.

Do I pay when the AI fails to solve the problem?

Fin: no, if it escalates without an outcome — yes, if it completes a procedure handoff or the customer goes silent after its answer. Zendesk: no — only verified resolutions bill. Freshdesk: yes — the session is consumed either way.

Methodology

What we did: Read the official pricing pages of all three vendors in full in October 2026 — Fin’s pricing page (outcome definitions, $0.99/$9.99 rates, trial terms), Zendesk’s pricing page (seat prices, Copilot add-on, plan structure), and Freshworks’ Freshdesk Omni pricing page (seat prices, $49/100 session pricing, and its FAQ defining sessions and the one-time 500 free sessions). Took Intercom’s seat prices and the standalone 50-outcome minimum from a current Intercom pricing breakdown that documents its checks against Intercom’s and Fin’s pricing pages.

What we did NOT do: We created no accounts and ran no trials, so nothing here measures resolution quality — a vendor whose AI resolves 40% of tickets badly is not cheaper than one resolving 40% well. Zendesk’s automated-resolution rates and allowances do not render in the text of its pricing page; we took them from multiple independent current breakdowns of that page and flagged them in the text accordingly. All deflection rates, team sizes, and session-per-ticket figures in the model are our labeled planning assumptions, not vendor data.

Conflict disclosure: None. No vendor in this article is an affiliate partner of this site.

AWL verdict

Pick your meter before you pick your vendor. Fin is the best-engineered deal for Intercom-native teams — $0.99 is a fair price for a real resolution — provided you audit assumed resolutions against satisfaction scores, because you will be billed for some customers who simply gave up. Zendesk charges the most per resolution and defines it the most honestly; for teams already on Zendesk whose tickets are complex enough that a “verified” check matters, that premium buys the only meter in the category aligned with actual outcomes. Freshdesk is the budget front door: unbeatable per interaction, indifferent to results, and ideal when your questions are simple and your volume is the problem. Whichever you choose, write your own deflection rate into the budget — never the vendor’s.

Sources

• Fin (Intercom) official pricing page — outcome definitions, $0.99/$9.99 rates, trial and standalone terms (verified October 2026): https://fin.ai/pricing
• Zendesk official pricing page — Suite and Support seat prices, Copilot add-on (verified October 2026): https://www.zendesk.com/pricing/
• Freshworks official Freshdesk Omni pricing page — seat prices, $49/100 sessions, session definition, one-time 500 free sessions (verified October 2026): https://www.freshworks.com/freshdesk/omni/pricing/
• Intercom pricing breakdown documenting seat prices, Copilot, and the standalone outcome minimum against Intercom’s and Fin’s pricing pages (verified October 2026): https://www.getmacha.com/blog/intercom-pricing-explained
• Zendesk AI chatbot pricing breakdown — automated resolution rates, allowances, and the May 2026 verified-resolution change (verified October 2026): https://www.eesel.ai/blog/zendesk-ai-chatbot
• Freddy AI pricing breakdown — Freshdesk seat tiers and session-pack mechanics (verified October 2026): https://www.eesel.ai/blog/freddy-ai-pricing