On Zendesk's current pricing page, Suite Team and Suite Professional include 5 and 10 automated resolutions per agent a month, and each one beyond that costs $2.00 pay-as-you-go or $1.50 committed, as listed when this tutorial was written. A three-agent Suite Team therefore gets 15 included, and each resolution above that adds $2 to the bill.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowThe unit to understand is the "resolution": a conversation the AI agent handled without a person, which Zendesk then counts and bills. It is not a ticket, a message or a chat. A customer who asks, gets an answer and leaves counts as one. A customer who asks, gets passed to a human and gets help there does not. That single definition decides your invoice, so the sections below work through what it does to a small team's month.
What does Zendesk's pricing page show for the AI agent?
Zendesk prices its AI agent in automated resolutions: Suite Team and Suite Professional include 5 and 10 per agent a month, then each extra resolution costs $2.00 pay-as-you-go. Zendesk's pricing page, read in November 2026, lists the following for annual billing, per agent per month.
| Plan | Seat price (annual) | Included resolutions per agent | Committed rate | Pay-as-you-go rate |
|---|---|---|---|---|
| Support Team | $19 | 5 | $1.50 | $2.00 |
| Suite Team | $55 | 5 | $1.50 | $2.00 |
| Suite Professional | $115 | 10 | $1.50 | $2.00 |
| Suite Enterprise with Copilot | Talk to sales | Not stated | Not stated | Not stated |
The page also notes that the included allowance is shown as "up to 10k per year", which I read as a ceiling on what the allowance can reach rather than a figure for a small team. It adds that resolutions beyond the allowance are billed, and that you can add committed resolutions or pay as you go. The Copilot add-on, a separate product that helps your agents rather than customers, costs $50 per agent a month paid yearly and can be added to Professional and higher.
One honest warning. Zendesk's help article on automated resolution tiers describes the allowance as a flexible pool measured in currency, with a dashboard that shows spend, whereas the pricing page speaks in counts per agent. Both views may be true at once, because the tiers change what a resolution costs. Because I could not reconcile them from public pages, treat 5 and 10 per agent as the headline figures and ask Zendesk to put your allowance in writing.
How much does a three-agent Suite Team pay in a normal month?
In the illustrative garage example, a three-agent Suite Team pays $405 a month when the AI agent resolves about 30% of 450 conversations. Take a car repair garage with three people answering customers, using Suite Team with the AI agent on its website chat. In a typical month it receives about 450 conversations. After a few weeks of tuning, the AI agent fully resolves around 30% of them, so roughly 135 resolutions. Here is the bill.
- Seats: 3 × $55 = $165 a month.
- Included resolutions: 3 × 5 = 15.
- Resolutions over the allowance: 135 minus 15 = 120.
- Pay-as-you-go cost: 120 × $2.00 = $240.
- Total: $165 + $240 = $405 a month.
Compare that with committing to a volume. If the garage commits to 120 resolutions at $1.50, the cost of those is $180 instead of $240, a saving of $60 a month, or $720 a year. But a commitment carries a risk that PAYG does not: if the AI agent resolves only 80 in a quiet month, a fixed commitment of 120 may leave you paying for 40 unused ones, subject to Zendesk's terms on how commitments are billed. The pricing page does not spell out the mechanics, so ask before you commit.
How much does the Zendesk bill change from month to month?
The Zendesk bill swings with the resolution count: for the same garage, a quiet month costs $185, a normal month $405 and a busy month $735. Here are three scenarios for the same garage, using the same $55 seats and PAYG rates.
| Scenario | Resolutions a month | Over the 15 included | Resolution cost at $2 | Total with $165 seats |
|---|---|---|---|---|
| Quiet start, bot barely trained | 25 | 10 | $20 | $185 |
| Normal month | 135 | 120 | $240 | $405 |
| Busy month after a recall notice | 300 | 285 | $570 | $735 |
The spread is the point. The same team, on the same plan, can pay $185 or $735 depending on what customers ask. The seat price is fixed and predictable; the resolution line is not. Budget for the middle scenario and decide in advance what happens in the busy one, including whether overages are switched on. Zendesk's help article says admins can allow or disallow overages; if disallowed, you can pause the features that need automated resolutions instead of paying for extra ones.
What do five included resolutions per agent actually cover?
Five included resolutions per agent cover very little: for a three-agent team that is 15 a month, which is 0.5 a day, or about 3% of the garage's 450 conversations. It is a trial allowance, enough to see the AI agent working, not enough to run a channel on. A larger team gets proportionally more, because the allowance scales with agents. A single-agent shop gets 5.
That makes the first decision a different question from "is it included?". The better question is: what is each resolution worth to you? Suppose a human agent handles a simple order-status chat in six minutes, and an agent's loaded cost is $24 an hour. A six-minute chat costs $2.40 of staff time. At $2.00 a resolution, the AI saves about $0.40 per chat on the face of it, which is thin. The saving is real only if the freed minutes go to something that earns money or protects customers, or if the AI handles volume you would otherwise hire for. At $1.50 committed the margin is better, $0.90 a chat. Those figures are illustrations; use your own handling time and wage.
For a wider look at how these costs sit with seats across a plan, How Much Is Zendesk? Seats, AI Resolutions and Real Monthly Totals works through complete monthly totals, and Zendesk Suite Pricing for Small Businesses covers a three-agent example on the seat side.
When does Zendesk count a conversation as an automated resolution?
Zendesk records an automated resolution when a conversation ends in a way it judges to be resolved without a human. Zendesk's earlier published rules said that for messaging, a conversation ends after two hours of inactivity by default, adjustable up to 72, and for email and web forms after 72 hours, and that an AI relevance check also applies. The account's resolution tiers, introduced in May 2026 according to Zendesk's help notes, can change what each resolution counts for. The current help page does not restate the trigger, so verify the rule that applies to your account.
Why does the clock matter? Because a customer who goes quiet after a half-answer can be counted as resolved. If your chat is busy at lunchtime and your customers often wander off mid-conversation, a short inactivity window may count unsatisfied customers as resolved, and you pay $2 for the privilege. A longer window protects you from that, at the cost of counting later. Whoever sets it should also read a sample of counted conversations every month, as in the review sheet below.
How do you check that each counted resolution was a real one?
Spot-check a sample of the resolutions Zendesk counts each month, because paying per resolution means each one should deserve the name. Here is a sample for the garage, with illustrative figures.
| Check | Sample result | Action |
|---|---|---|
| Resolutions counted in the month | 135 | Record against budget of 120 |
| Sample read (every tenth) | 13 conversations | Read the full transcript |
| Genuinely answered the question | 10 of 13 | Good |
| Customer left without an answer | 2 of 13 | Lengthen the inactivity window |
| Wrong answer given confidently | 1 of 13 | Fix the article the AI used |
| Cost of the 3 poor ones, extrapolated | about 31 × $2 = $62 | Note, then act on the cause |
Three poor counts in 13 is about 23%, and across 135 resolutions that is roughly 31 questionable ones, which at $2 each is $62 a month. The number is a prompt to improve the help articles the AI draws on, not a reason to switch it off. Each fixed article raises the share of genuinely good resolutions and lowers waste.
If you keep the monthly review sheet in a spreadsheet, try The Complete Excel Course with AI Integration on Coding Liquids.
A busier shop: the online clothing store
An illustrative online clothing store has six agents on Suite Professional ($115 each, $690 a month), with 1,800 conversations a month and an AI agent resolving 40%, so 720. Included allowance: 6 × 10 = 60. Over the allowance: 660. At pay-as-you-go: 660 × $2 = $1,320. At committed: 660 × $1.50 = $990, a $330 monthly saving for committing to 660. The total with seats is $2,010 on PAYG or $1,680 committed.
At that volume, the commitment question deserves real attention. Returns season can add a third more conversations in a couple of months, and the commitment can be sized to the quiet-month baseline rather than the peak. A reasonable pattern is to commit to about 80% of your lowest recent month and pay as you go above it. That captures most of the saving without paying for a volume you may not reach. Check Zendesk's current commitment terms before copying that plan.
A one-agent pest-control firm: when the allowance is nearly irrelevant
Consider an illustrative pest-control firm with one person answering the phone and chat, on Suite Team at $55 a month. The allowance is 5 resolutions. The firm gets about 60 chats a month, mostly "do you cover my area?" and "how much for wasps?". If the AI agent resolves half, that is 30 resolutions: 5 included and 25 at $2, so $50. The total is $105 a month for a business that previously paid $55.
What did the extra $50 buy? Perhaps 25 chats the owner did not have to answer, at roughly five minutes each, which is about two hours. If she values her time at $30 an hour, that is $60 of time for $50 of resolutions: almost break-even in money, with a benefit in hours given back and in chats answered out of hours. A one-agent firm should therefore ask a plain question: does the AI agent answer at night and weekends when nobody is there? If yes, the value is in the enquiries you would otherwise have lost, not the minutes saved.
Is the Zendesk AI agent cheaper than hiring a part-time agent?
The Zendesk AI agent is cheaper per simple conversation than a part-time agent only for conversations it actually resolves, so compare its per-resolution price with the cost of a person. Suppose a part-time agent costs $1,800 a month fully loaded and handles 600 simple conversations in that time, which is $3 a conversation. The AI agent at $2 pay-as-you-go is cheaper per simple conversation, at $1.50 committed it is half the price. But the comparison holds only for conversations the AI actually resolves. If it resolves 40% of what it sees and hands the rest on, the people still do the harder 60%. The sensible reading is that the AI agent reduces the number of hours you need to buy; it rarely replaces a role outright. Run the sum with your own wage, your own handling time and a realistic resolution share, and treat any figure much better than half a conversation of savings with suspicion.
How do you improve a help article so more resolutions are real ones?
A Zendesk AI agent resolves only as well as the help articles it draws on, so rewrite vague articles into specific answers. Here is an illustrative before and after for an article used by the garage's chat.
BEFORE
Title: Bookings
We offer servicing and inspections. Call us to arrange a time.
AFTER
Title: How to book a service or inspection
You can book online at any time or call during opening hours,
Monday to Saturday, 8am to 6pm. A standard service takes about
three hours; a full inspection takes five. We will text you when
the car is ready. If you need a courtesy car, tell us when you book,
because there are only two. To change a booking, reply to your
confirmation text at least 24 hours before.
The "before" article forces the AI agent either to guess or to hand the customer to a person, so the conversation costs staff time and may not count as resolved. The "after" answers four common questions: how to book, how long it takes, what happens at the end, and how to change it. After a week of such fixes, the share of conversations the agent genuinely resolves rises, and the quality of what you pay for rises with it. The same edit costs nothing and can be done in twenty minutes per article.
What should you ask Zendesk in writing before committing to a volume?
Before committing to a volume, send your figures to Zendesk's sales or support contact and ask for the answers in writing. A filled-in message looks like this.
Hello,
We have 3 agents on Suite Team (annual). Over the last 3 months
we averaged 450 conversations a month in web chat. Please confirm
in writing:
1. Our included automated resolutions per month, and whether that
is a count or a currency pool.
2. The rate for resolutions above the allowance if we commit to
nothing, and the committed rate at 120 a month.
3. How a commitment is billed in a month we use fewer than 120.
4. The inactivity window used to end a chat for counting.
We expect to enable the AI agent on website chat only at first.
Thanks, [your name]
A reply in writing is worth keeping. Pricing pages change, and an account manager's email is the record you point to if the invoice differs.
How do you read the first Zendesk invoice against your forecast?
Compare the first Zendesk invoice with your budget sheet line by line instead of looking at the total. Three lines matter: seats, included resolutions and resolutions billed above them. In the garage example the forecast was 60 resolutions in month one, so 45 billable. If the invoice shows 70 billable, the AI agent is resolving more than planned, which is either good news (more conversations closed) or a sign the inactivity window is counting abandoned chats. Open the sample of counted conversations to find out which. If the invoice shows 20 billable, the agent is being used less than planned, perhaps because the chat widget is hard to find. Either difference is information, and the review turns it into a decision rather than a surprise.
If you want help working out whether the AI agent suits your support queue, I offer one-to-one AI consultations for small businesses.
Cutting the bill without hurting customers
- Turn the AI agent on for one channel first. Website chat before email, so you see the volume before you scale it.
- Improve the help articles. The AI answers from your content. A clear, current article raises the share of resolved conversations; a vague one produces hand-offs and complaints.
- Set the overage rule deliberately. Allow it only if you accept a bill above budget.
- Review the sample monthly. Count wasted resolutions, as above, and fix their cause.
- Compare against the alternative. Another tool with a different pricing model may suit a small team better; the tutorial on Tidio vs Zendesk for a Three-Agent Team sets two side by side.
When the Copilot add-on enters the sum
The $50 Copilot add-on is a different cost line: it helps your human agents draft and summarise rather than answering customers. It is available on Professional and above, so a Suite Team customer cannot add it. For three agents on Professional that is $150 a month extra. It does not replace resolution spend and does not reduce it directly, so decide which product addresses your problem. If your queue is full of repetitive questions, the AI agent comes first. If your agents are slow at complex replies, Copilot does. Zendesk Copilot Pricing: Is $50 per Agent Worth It? looks at that choice from the other side.
A first-month budget you can copy
Team: 3 agents, Suite Team, annual billing
Seats: 3 x $55 = $165
Included resolutions: 3 x 5 = 15
Month-1 estimate of resolutions: 60 (AI on website chat only)
Over allowance: 45 x $2.00 = $90
Month-1 total: $255
Overages: ALLOWED up to $150, then pause (set by the account admin)
Review: read every 5th counted resolution in week 4
Decision at day 45: commit to a volume, stay PAYG or switch off
The sheet is short on purpose. It records the seat cost, the estimate, the cap and the review date. A cap you wrote down in advance prevents the argument about the invoice later.
What to confirm with Zendesk before you sign
- Your included allowance in writing: count per agent or a currency pool, and its value.
- The default rate when you exceed the allowance and nothing has been committed.
- How a commitment is billed in a month you use less than the committed volume.
- The inactivity window that ends a conversation for counting, per channel.
- What the overage setting does on the day the allowance runs out.
The figures in this tutorial were read from Zendesk's own pages in November 2026. Seat prices load dynamically on Zendesk's site and rates change, so check the page you will buy from and use these sums as a method, not a quote. For the first week after you switch the AI agent on, How to Use Zendesk AI: Your First Week With a Two-Person Team is the practical follow-up.
Zendesk AI agent billing questions
How much does the Zendesk AI agent cost for a small team?
In the page's worked example, a three-agent Suite Team pays $405 a month: $165 for seats plus $240 for 120 resolutions above the 15 included, at $2.00 pay-as-you-go. Your total depends on how many conversations the AI agent resolves, so a quiet month costs far less ($185 in the example) and a busy one more ($735).
What do 5 included resolutions per agent cover?
Very little. Five included resolutions per agent a month is 15 for a three-agent team, about 3% of 450 conversations. It is a trial allowance, enough to see the AI agent working but not to run a channel on, and a larger team gets proportionally more because the allowance scales with agents.
Do unused included resolutions roll over?
No. Zendesk's help documentation says automated resolutions do not roll over to the next billing period. If your team resolves fewer than the included number in a billing period, the spare allowance is lost rather than banked for the next one.
Can I stop overage charges entirely?
Yes, in the sense that Zendesk's help article says admins can allow or disallow overages on the Overage page. When the allowance runs out, you choose between pausing the features that need automated resolutions or paying for extra ones as you go. Decide which you want before launch, and have the setting reviewed by whoever receives the invoice.
When does a conversation count as a resolution?
It counts after a quiet period and a check that the answer was relevant. Zendesk's earlier published rules said two hours of messaging inactivity by default or 72 hours for email and web forms, plus an AI relevance check. Those windows and the tier rules have been revised, so read the current help page before you forecast volume.
Is $2 the same on every plan?
Zendesk's pricing page lists the same $2.00 pay-as-you-go rate and $1.50 committed rate across Support Team, Suite Team and Suite Professional. What differs by plan is the included allowance per agent. Enterprise-level plans are quoted through sales, so their allowance and rates should be confirmed directly.
Further reads
- How to Draft Zendesk Macros With ChatGPT Without Exposing Data — Draft macros and replies with ChatGPT, safely.
- Is Zendesk Too Much for a Three-Person Support Team? — Whether Zendesk suits a three-person team at all.
- How Much Does Zendesk Implementation Cost? Hours, Partners, DIY — Setup hours beyond the licence price.
- Freshdesk vs Zendesk for a Small Support Team Using AI — A rival help desk costed for a small team.
- Should a Small Team Buy Zendesk AI Agent or Copilot First? — Which Zendesk AI product to buy first.
- Zendesk to Freshdesk Migration: What to Move, Rebuild or Retire — A move, rebuild or retire checklist for Zendesk to Freshdesk: tickets, macros, help centre, AI retraining and the parallel run.
Browse all AI-for-business tutorials
Sources: Zendesk pricing page and its automated resolution section (zendesk.com/pricing), and Zendesk's help article on automated resolution tiers, read in November 2026.
