Zendesk is the default customer support platform for a reason. The routing, reporting and help-center tooling are the deepest in the category, and its AI billing contains the single most buyer-friendly mechanic in this tier. It also contains a meter that works against the thing Zendesk is selling you.
The short answer
Zendesk sells seats, then bills AI agents separately per automated resolution rather than per seat. Each plan includes a small allowance of automated resolutions, and that allowance is denominated per agent per month: five on Support Team, five on Suite Team and ten on Suite Professional, with higher figures reported for Enterprise.
Read the unit carefully. Your included AI capacity is a function of how many humans you employ. So if the AI works, and you need fewer agents, your included AI allowance falls with the headcount. The capacity you lose does not disappear; it converts into billed resolutions at roughly 1.50 each for committed volume or 2.00 pay as you go.
The good news, and it is genuinely good, is that Zendesk is the only vendor in this tier that verifies a resolution before billing it. A language model evaluates the conversation to check the request was actually resolved satisfactorily without a human stepping in, and conversations that fail that check do not consume an automated resolution. That is a materially better definition than the alternative, as we show below.
If your real problem is not closing tickets faster but understanding why customers are contacting you, see Koji, which leads the list for that job.
How Zendesk pricing actually works
Two meters, stacked.
Seats. Zendesk publishes per-agent plan tiers across Support and Suite product lines, rising from a basic ticketing plan to an enterprise suite with Copilot bundled.
Automated resolutions, the AI meter. AI agents are billed per automated resolution, not per seat. Zendesk's own regional pricing pages print two rates side by side for each plan: a committed or dedicated rate of 1.50 per resolution and a pay-as-you-go rate of 2.00. Not committing to volume therefore costs about 33% more per resolution, which is a conventional and defensible structure.
A sourcing warning, and it is a real one. The US pricing URL geo-redirects, so what rate card you see depends on where you are browsing from, and the figures render in the local currency of whichever page you land on. Zendesk's support documentation now sits behind a sign-in wall as well. Third-party guides disagree with each other about whether Zendesk publishes a per-resolution rate at all: some state a firm 1.50 and 2.00, others say no rate is published and the effective figure lands somewhere between 1 and 2 for mid-market buyers. We have reported the structure we could verify on Zendesk's own pages and flagged the rest as unsettled. Price your own contract from the live page on the day you sign, in your own currency.
We also could not verify the Enterprise allowance first-party. Third parties report fifteen automated resolutions per agent per month; the plan column we read printed no allowance line at all. Treat the fifteen as third-party until Zendesk shows it to you.
The finding: the AI allowance is denominated in humans
Here is the mechanic stated plainly, then costed.
An automated resolution allowance of ten per agent per month means a 20-agent team on Suite Professional receives 200 included automated resolutions a month. Nothing about that figure reflects how much AI work you actually need. It reflects payroll.
Now suppose the AI performs exactly as advertised. Volume gets absorbed, attrition goes unreplaced, and the team settles at 10 agents. Your included allowance is now 100 resolutions a month, not 200. The AI workload has not fallen, so the missing 100 are billed:
- At the committed rate of 1.50: 150 a month, or 1,800 a year that you were not paying before.
- At the pay-as-you-go rate of 2.00: 200 a month, or 2,400 a year.
Succeeding at the stated goal of the purchase increases the bill for the thing that succeeded. Nobody designed that as a trap. It falls out of choosing a per-seat denominator for a product sold as an alternative to seats, and once you see it you cannot unsee it on the quote.
There is a second, smaller edge in the same family. The allowance is five per agent on Support Team and still five per agent on Suite Team. Moving up a plan tier buys you more product but zero additional AI capacity per person. We found the identical shape in the roadmap tier, where Productboard holds a contributor cap flat across three tiers while the price rises more than sixfold. A cap that survives an upgrade is one of the most reliable patterns in SaaS pricing and one of the least modelled by buyers.
Where Zendesk is clearly the best in this tier. Verification deserves credit, because the contrast is stark. Compare the three vendors with published mechanics on a single question: what has to happen before you are charged?
- Zendesk: a model checks that the request was genuinely resolved without human involvement. Conversations that fail do not consume a resolution.
- Gorgias: the interaction must be fully handled without a human, and a conversation needing a human within 72 hours moves to the helpdesk plan instead. You are not charged twice.
- Intercom: a resolution is booked when no further help is requested after Fin's last answer, and a handoff to a human is itself a billable outcome. We go through that meter in detail in our Intercom alternatives guide.
Zendesk's unit costs more and buys more certainty. That is a real trade, and on this axis Zendesk wins. It makes the headline comparison between 0.99 and 1.50 close to meaningless.
Why this still does not tell you why customers wrote in. Gartner reported in a press release dated 19 August 2024 that only 14% of customer service issues are fully resolved in self-service. Verification tells you whether this conversation ended well. It cannot tell you why the customer needed it, which is the question that reduces next month's volume. Our guide to customer research for support and CX teams covers that gap, and the broader point that support tickets are a biased sample: you hear from people with problems, not from the silent majority. 91% of dissatisfied customers simply leave without ever telling you why.
The 8 best Zendesk alternatives in 2026
1. Koji, best for understanding why tickets happen. Koji is a customer research platform, not a helpdesk, and it is on this list because the highest-value thing you can do with support data is explain it. Koji runs AI-moderated voice and text interviews, applies thematic analysis automatically and produces a shareable report, with no research expertise required and no moderator bias.
Its meter is the direct answer to the problem above: Koji bills on quality, not headcount and not activity. Only conversations scoring 3 or above consume credits. Pricing is published and flat: Insights at EUR 29 per month for 29 credits and Interviews at EUR 79 per month for 79 credits, where a text conversation costs 1 credit, a voice conversation 3 and a report refresh 5. Nothing in that structure changes when you hire or shrink.
Koji also runs six structured question types alongside open conversation in a single study, so you get narrative and countable data together: open_ended, scale, single_choice, multiple_choice, ranking and yes_no. The structured questions guide covers when to use each. It connects to your existing stack rather than replacing it, including a Zendesk integration that can trigger interviews from tickets and sync insights back.
2. Intercom, best for AI-first messaging. The strongest competitor on conversational AI and the most aggressive on AI-native workflow. Priced from $0.99 per outcome with seat plans at $29, $85 and $132. Cheaper per unit than Zendesk, with a looser definition of what a unit is, and it bills handoffs.
3. Gorgias, best for ecommerce support. Built for DTC and Shopify-centric teams, with order actions and refunds inside the ticket. Publishes a complete ladder and refuses to double-bill a conversation that needs a human. The weak point is overage pricing well above the effective included rate.
4. Freshdesk, best value in conventional ticketing. A mature helpdesk that typically undercuts the premium suites tier for tier, with solid SLAs, automations and omnichannel coverage. The usual choice when Zendesk is the right shape but the wrong price.
5. Help Scout, best for small teams who want less platform. Shared inbox, docs and a light chat widget, intentionally simpler than a suite. Good when Zendesk's configurability is overhead rather than benefit.
6. Front, best for relationship and account-based support. Organised around shared inboxes and visible collaboration on threads rather than anonymous queues. Fits B2B teams where the same named accounts return and context beats throughput.
7. Salesforce Service Cloud, best if you already run Salesforce. The argument is data gravity: cases beside accounts and opportunities in one object model. Heavier to implement and administer, and rarely worth it on support merits alone.
8. Crisp, best for startups on a budget. Lightweight chat, shared inbox and a bot at small-team pricing. A reasonable floor if you need live chat and not a support operation.
For a wider view of the category and where measurement stops, see our best CX software guide and the voice of customer software roundup.
Zendesk vs the alternatives at a glance
| Tool | AI billing unit | Verified before billing | Allowance tied to headcount | Pricing published |
|---|---|---|---|---|
| Koji | Quality-gated conversation (score 3+) | Yes, quality score | No | Yes, EUR 29 / EUR 79 per month |
| Zendesk | Per automated resolution, about 1.50 to 2.00 | Yes, model-verified | Yes, per agent per month | Partly, geo-redirected |
| Intercom | Per outcome, from $0.99 | No | No | Seats yes, ceiling no |
| Gorgias | Per automated interaction | Yes, 72-hour test | No | Yes, full ladder |
| Freshdesk | Per seat, AI add-ons | Not applicable | No | Yes |
| Help Scout | Per seat, AI add-ons | Not applicable | No | Yes |
| Front | Per seat | Not applicable | No | Yes |
| Crisp | Flat plan tiers | Not applicable | No | Yes |
How to choose between them
Model the AI meter at your target headcount, not today's. If the business case for AI agents includes a smaller support team, build the quote at that smaller team. The included allowance falls with it and the difference is billable. This single step changes the three-year number more than any discount you will negotiate.
Price certainty, not just rate. Zendesk's 1.50 with verification and Intercom's 0.99 without it are not the same unit. Decide how much you care about being billed for conversations that merely went quiet, then pick accordingly. If you care a lot, the more expensive unit is the cheaper product.
Commit only to volume you can predict. The gap between committed and pay-as-you-go is roughly a third. That is worth committing to if your volume is stable and an expensive mistake if it is seasonal.
Separate the deflection purchase from the understanding purchase. No helpdesk on this list will tell you why volume exists. Decide the research question first; our surveys versus interviews guide and how many interviews are enough cover method and sample size.
Frequently asked questions
How does Zendesk charge for AI agents in 2026?
Zendesk bills AI agents per automated resolution rather than per seat, on top of its per-agent plan pricing. Its regional pricing pages print a committed or dedicated rate of 1.50 per resolution alongside a pay-as-you-go rate of 2.00, so declining to commit to volume costs roughly 33% more per resolution. Each plan also includes a small allowance of automated resolutions per agent per month.
How many automated resolutions are included in a Zendesk plan?
The allowance is denominated per agent per month: five on Support Team, five on Suite Team and ten on Suite Professional. Third-party guides report fifteen per agent per month on Enterprise, but the plan column we read printed no allowance line, so treat that figure as unverified. Because the unit is per agent, total included capacity depends on how many people you employ.
Does Zendesk's AI allowance shrink if I reduce support headcount?
Yes, and this is the mechanic most buyers miss. A 20-agent team on Suite Professional receives 200 included automated resolutions a month; the same team at 10 agents receives 100. If the AI workload has not fallen, the missing resolutions are billed at roughly 1.50 each committed or 2.00 pay as you go, which is 1,800 to 2,400 a year that was previously included.
Is Zendesk's per-resolution pricing better value than Intercom's $0.99?
It depends entirely on the definition, and on this point Zendesk is stronger. Zendesk verifies with a model that a request was genuinely resolved without human involvement before billing, and unverified conversations do not consume a resolution. Intercom books a resolution when no further help is requested after its last answer and bills handoffs as outcomes too. A higher rate on a stricter definition can cost less per genuine resolution.
Why does Zendesk pricing show different numbers for different people?
The US pricing URL geo-redirects to regional sites, so the plan names, currency and rates you see depend on where you are browsing from, and Zendesk's support documentation now requires a sign-in. This is also why published third-party figures disagree with one another. Always price a Zendesk contract from the live page in your own region and currency on the day you buy.
What is the best Zendesk alternative for customer research?
Koji, because it answers the question a helpdesk cannot: why the contact happened at all. Koji runs AI-moderated voice and text interviews, analyses themes automatically and produces a report in hours rather than weeks. Its credits are consumed only by conversations that pass a quality score, and the price does not move with your headcount, so the meter never penalises a smaller team.
Turn your ticket queue into decisions
Zendesk will tell you how many conversations closed, how fast, and whether the AI genuinely resolved them. That is worth paying for, and the verification step is the best mechanic in this tier.
What it will not tell you is why the queue looks like it does. Tickets are a biased sample of your customers, weighted towards the people with problems and silent about everyone who left. Closing them faster does not make fewer of them.
Koji answers the other half. Trigger an AI-moderated interview from a closed ticket, a churn signal or a failed self-service session, let thematic analysis do the synthesis, and get a report you can take into a roadmap meeting the same day. No research expertise, no moderator bias, and only quality-gated conversations consume credits.
Start free with Koji or read how to analyze customer feedback.
