MilooshCustomer Support Pricing Benchmark 2026
Customer support software billing split into two families in 2025-2026: pure per-seat licensing, and seat pricing paired with a separate charge for AI-handled conversations. This benchmark checks all 16 products Miloosh's own catalog classifies under Customer Support, individually, against each vendor's own pricing page, to find out how many actually price AI usage separately -- and what that costs in practice.
Citation
Miloosh Research. “Customer Support Pricing Benchmark 2026.” Verified 2026-09-27. https://miloosh.com/research/customer-support-pricing-2026. Methodology: https://miloosh.com/research/customer-support-pricing-2026#methodology. Dataset licensed CC BY 4.0 -- reuse with attribution.
Revised 2026-09-27: a full re-verification against every vendor's live pricing page found that the original AI-usage-pricing count (5 of 16) had missed real, publicly disclosed AI charges at 4 vendors -- corrected to 9 of 16 below. This was a gap in the original research, not a set of overnight vendor price changes. See the dataset for the current figures.
9 of 16
Publish a distinct, separately billed AI-usage price
6 of 16
Bundle any AI features into seat tiers with no separate usage price disclosed
1 of 16
Have no public base seat price at all -- the price itself requires a sales conversation
4 of 16
Offer a free tier (not just a trial)
15 of 16
Publish a specific entry-tier dollar amount
1 of 9
Disclose an AI-usage price with no undisclosed seat price or allowance offsetting it (Intercom) -- see below
Vendors that bill AI usage separately
Every row was independently re-checked against the vendor's own pricing page on 2026-09-27. Units differ (per-resolution, per-session-block, per-conversation, per-action) and are not directly interchangeable -- see each note before comparing across rows.
| Vendor | AI usage rate | Unit | Note |
|---|---|---|---|
| Freshdesk | $0.49 | per Freddy AI Agent session (billed in blocks of 100 sessions at $49/block, after the first 500 sessions/month included) | "First 500 sessions included. $49 per 100 sessions" -- a session-block unit, not a per-resolution price. |
| Front | $0.05 | per conversation handled by the Autopilot AI agent add-on (a "starting at" rate, not confirmed flat) | "Autopilot -- Run complex customer work with our omnichannel AI agent -- Starting at $0.05 /conversation", billed separately from Front's per-seat plan pricing -- found live on 2026-09-27; missed in the original 2026-09-26 research. Front's other AI add-ons (Copilot, Smart QA, Smart CSAT) are seat-based, not usage-metered. Because the rate is stated as a "starting at" figure rather than a single confirmed number, this vendor is not used for the crossing-scenario calculation (see build.ts). |
| Gorgias | $0.90 | per AI Agent-resolved conversation, baked into each plan's headline price (rate varies $0.85-$1.00 by tier: Starter $1.00, Basic/Pro $0.90, Advanced $0.85) | Re-checked 2026-09-27: the per-interaction AI rate is not purely an overage charge -- it is the baseline rate pricing the AI Agent component of every plan's total, including its included allowance, with a separately noted (and possibly generic/non-tier-specific) $1.50/interaction overage rate beyond that allowance. The disclosed range widened from the original $0.90-$1.00 finding to $0.85-$1.00 once the Advanced tier's rate was confirmed. 30-day free trial, no credit card, also newly confirmed. |
| HappyFox | $0.33 | per resolution for the Chatbot product (a second, cheaper usage-based AI product, Autopilot, is billed "as low as 2 cents per successful action") | Correction, not a vendor price change: on 2026-09-26 HappyFox's pricing page could not be reached and this row was marked entirely UNKNOWN. Re-checked live on 2026-09-27: the page is reachable and discloses real prices throughout, including two separately metered, usage-based AI products (Chatbot at $0.33/resolution, Autopilot as low as $0.02/action) distinct from two flat per-seat AI add-ons (HappyFox AI $14/agent/month, Assist AI $1/user/month). Base Help Desk pricing starts at $24/agent/month. |
| Help Scout | $0.75 | per resolution via the AI Answers add-on (a resolution counts when the customer accepts the AI's answer without escalating, searching further, or asking more questions) | "AI Answers Add-on $0.75/resolution" -- billed monthly on top of the per-seat Standard/Plus plan price, with its own separate 3-month free trial. Found live on 2026-09-27; missed in the original 2026-09-26 research. |
| Intercom | $0.99 | per Fin AI outcome (a confirmed resolution or a completed workflow with human handoff) | "From $0.99 per Fin outcome", billed separately from the per-seat plan price on every tier -- confirmed live on intercom.com/pricing, 2026-09-26. |
| Kayako | $1.00 | per ticket Kay AI Agent fully resolves without human involvement (escalations and partial resolutions are free) | The AI usage rate is the ONLY dollar figure Kayako's pricing page discloses publicly; the base per-agent seat price is contact-sales-only. This is the inverse of Intercom's transparency pattern. |
| Re:amaze | $0.85 | per additional AI Agent resolution beyond the plan's monthly included allowance | "AI agent features include included monthly resolutions with per-resolution overage fees of $0.85 per additional resolution" -- confirmed live on reamaze.com/pricing, 2026-09-26. The size of the included monthly allowance itself is not disclosed on the pricing page, so a full crossing-point cannot be computed for Re:amaze the way it can for Intercom. |
| Tidio | $32.50 | for the Lyro AI Agent, sold as a separate product from the base live-chat plans (starting price for a block of 50 AI conversations/month, a bundle rate like Freshdesk's, not a marginal per-conversation price) | "Lyro AI Agent -- Starts at $32.50/mo. From 50 Lyro AI conversations", billed and sold separately from Tidio's base conversation-volume plans. The Premium plan additionally lists "Pay-per-resolution billing" as a feature, but no exact per-resolution dollar rate is published anywhere on the page -- that narrower claim is left unquantified rather than estimated. Billing basis was also reclassified: Tidio's plans are billed by conversation volume, not per seat (entryPerSeat corrected from unknown to false). |
Illustrative scenario: seat cost vs. AI usage cost
Intercom is the only vendor in the table above whose base seat price, AI usage price, and per-plan seat count are all public with no undisclosed included-usage allowance offsetting the arithmetic -- so it is the only one this scenario can be computed for from public information alone. This is arithmetic on published list prices for a hypothetical 3-seat team, not a customer invoice, a recommendation, or a claim about typical usage.
| AI-resolved conversations/month | Seat cost (3 seats) | AI usage cost | Total |
|---|---|---|---|
| 0 | $87.00 | $0.00 | $87.00 |
| 100 | $87.00 | $99.00 | $186.00 |
| 300 | $87.00 | $297.00 | $384.00 |
| 500 | $87.00 | $495.00 | $582.00 |
| 1000 | $87.00 | $990.00 | $1,077.00 |
The AI usage line alone crosses Intercom's 3-seat base fee ($87.00) at 88 resolutions/month ($87.00 / $0.99) -- a calculated threshold, not a claim that any specific business hits that volume.
Full dataset -- all 16 products
Sample: Every catalog product with category "customer-support" as of the compilation date -- not a hand-picked vendor list. Each row links to its vendor source.
| Vendor | Recorded starting price | Per-seat? | AI usage priced separately? | Free tier | Verified | Source |
|---|---|---|---|---|---|---|
| Crisp | $45 | No | No | Yes | 2026-09-27 | Source |
| Five9 | $119/seat/month | Yes | No | No | 2026-09-27 | Source |
| Freshdesk | $19/agent/month | Yes | Yes | No | 2026-09-27 | Source |
| Front | $25/seat/month | Yes | Yes | No | 2026-09-27 | Source |
| Genesys Cloud CX | $75/user/month | Yes | No | No | 2026-09-27 | Source |
| Gorgias | $40 | No | Yes | No | 2026-09-27 | Source |
| HappyFox | $24/agent/month | Yes | Yes | No | 2026-09-27 | Source |
| Help Scout | $25/user/month | Yes | Yes | Yes | 2026-09-27 | Source |
| Intercom | $29/seat/month | Yes | Yes | No | 2026-09-27 | Source |
| Kayako | Unknown -- contact sales | Unknown | Yes | No | 2026-09-27 | Source |
| LiveAgent | $15/agent/month | Yes | No | No | 2026-09-27 | Source |
| Re:amaze | $29/team member/month | Yes | Yes | No | 2026-09-27 | Source |
| Talkdesk | $85/user/month | Yes | No | No | 2026-09-27 | Source |
| Tidio | $24.17/month | No | Yes | Yes | 2026-09-27 | Source |
| Zendesk | $19 | Yes | No | No | 2026-09-27 | Source |
| Zoho Desk | $7 | Yes | No | Yes | 2026-09-27 | Source |
Methodology
- Sample
- Every catalog product with category "customer-support" as of the compilation date -- not a hand-picked vendor list. No vendor was added or removed to fit a preferred conclusion.
- Verification
- Each row's price fields are checked individually against that vendor's own pricing page (never a third-party aggregator, when an official source exists). AI-usage-pricing classification (whether a distinct, separately billed AI unit is publicly disclosed) was hand-verified per vendor, not inferred from keyword matching. All 16 rows were most recently re-verified on 2026-09-27; this page was compiled on 2026-09-26.
- Missing data
- All 16 products in this sample now have at least some publicly verified pricing information. That was not true at first publication: HappyFox's pricing page could not be reached during the original research and was marked entirely Unknown until the 2026-09-27 re-verification found it fully priced. One product (Kayako) still has no public base seat price -- see the table above.
- Why “benchmark,” not “index”
- This page computes no single composite score across vendors. It reports verified per-vendor facts and a small number of clearly labeled scenario calculations, which is what “benchmark” means. Calling it an index would imply a defensible weighted formula that does not exist here.
- Limitations
- Scenario calculations exclude annual-contract discounts, enterprise negotiated pricing, taxes, and non-USD currency conversion. Billable-unit definitions (an Intercom “outcome,” a Freshdesk “session,” a Help Scout or Re:amaze “resolution,” a Front “conversation,” a HappyFox “action”) are each set by that vendor and are not directly interchangeable across rows. This benchmark does not measure deflection rate, answer quality, or actual customer spend -- only published list prices and billing structures.
- Site-wide standards
- This page follows the same rules as every other page on Miloosh: see the Sources Policy for how facts are sourced and dated, the Editorial Policy for how commercial relationships are kept separate from what gets published, and the Corrections Policy to report an error in this dataset.
What this means for a buyer
- Ask every vendor with a disclosed AI-usage price what counts as a billable unit -- an Intercom “outcome” and a Freshdesk “session” are defined differently, and neither is quality-adjusted.
- A vendor that publishes only its AI-usage rate and not its base seat price (Kayako) still requires a sales conversation before any total cost is knowable -- the published number alone understates the bill.
- A vendor can also disclose the opposite way: Zendesk names a separate AI-usage billing unit (“Automated Resolutions”) without publishing its dollar rate -- ask for the actual number before assuming the base seat price is the whole bill.
- A “starting at” AI-usage rate (Front's Autopilot, at $0.05/conversation) is not a guaranteed flat rate the way Intercom's or Kayako's stated rate is -- confirm whether it can increase before modeling a total cost from it.
- Flat-workspace and conversation-volume-tiered pricing (Crisp, Tidio's base plans) cap headline exposure but shift the real decision to whether the included allowance actually covers the required volume.
- For a team expecting meaningful AI-resolution volume, model the AI-usage line item separately from the seat line item before comparing vendors -- collapsing them into one “starting price” figure hides exactly the divergence this benchmark measures.
How to compare unlike billing units
A per-seat price and a per-resolution price answer different questions, so putting them side by side as if they were the same kind of number is the most common mistake in this comparison. A seat price scales with headcount; a usage price scales with volume. The only fair comparison is a modeled total at your own expected seat count and usage volume -- not the two headline numbers on their own.
Why per-outcome and per-seat prices aren't directly comparable: a vendor charging $0.99 per AI outcome is not “more expensive” than one charging $29 per seat -- those numbers measure different things. A team with low ticket volume and several agents may pay far more under a per-seat model; a team with few agents but high automated-resolution volume may pay far more under a per-outcome model. Neither structure is inherently cheaper.
Questions worth asking before switching support platforms:
- What exactly counts as one billable unit (an “outcome,” a “session,” a “resolution”) -- and does it fire even when the customer wasn't actually helped?
- Is there a minimum seat count, a minimum contract term, or a required annual commitment to get the advertised price?
- If usage spikes in a single month, is the overage rate published, or does it require contacting sales after the fact?
- Does the plan's AI allowance reset monthly, and what happens to unused allowance -- is it lost, or does it roll over?
- Is the base seat price itself public, or does getting a real number require a sales conversation (as with Kayako in this dataset)?
Related reading
Get notified when this benchmark updates
Occasional emails when the dataset materially changes. Each vendor row carries its own verification date; compilation does not mean every vendor was rechecked today.