An AI support agent costs between roughly $29 a month and several thousand, and the spread has almost nothing to do with quality. It is the billing basis. Four are in use and they behave very differently as you grow.
TL;DR
- Four bases are in play: per agent, per resolution, per user, and flat regardless of seats.
- Per-agent charges for the size of your IT team, which is unrelated to how much the thing gets used.
- Per-resolution bills you more as adoption improves, which is defensible externally and awkward internally.
- Flat is indifferent to both headcount and volume, which makes it the only basis you can budget two years out.
- At 12 seats the gap between the cheapest and dearest basis is already several hundred dollars a month.
- Run the sum at your current numbers and again at your projected ones before accepting any ranking.
The four ways an AI support agent is priced
Every figure below was read from the vendor’s own page on 15 October 2026.
| Basis | Example pricing | Scales with |
|---|---|---|
| Per agent | Freshservice Starter $19, Growth $49, Pro $99 per agent per month, billed annually, plus AI at $29 per agent per month | Size of your IT team |
| Per agent, suite | Zendesk Support Team $19, Suite Team $55, Suite Professional $115 per agent per month, paid yearly, plus Copilot at $50 per agent per month paid yearly | Size of your IT team |
| Per resolution | Intercom Fin at $0.99 per resolution | How much it is used |
| Per user | Chatbot.com from $19 per user per month, Essential $29.50, Growth $79 | Number of licensed users |
| Flat | Matram $29 Basic, $69 Standard, $199 Pro per month, unlimited seats | Nothing |
| Flat per workspace | Crisp $0, $45, $95, $295 per month, extra agents charged on top | Agent count above the included set |
Note that two entries describe themselves as flat and only one is flat all the way down. Crisp charges for additional agents above its included allowance, so the published figure is a floor rather than a total.
The same team, priced four ways
Take a team of 12 people who would need seats, fielding 600 questions a month. Then the same team grown to 20 people and 1,200 questions. Every sum is written out so you can check it rather than trust it.
On a per-agent basis with AI included, Freshservice Growth plus the Freddy add-on is $78 an agent, so 12 agents is $936 a month and 20 agents is $1,560 a month.
On a suite per-agent basis, Zendesk Suite Team at $55 plus Copilot at $50 paid yearly is $105 an agent, so 12 is $1,260 a month and 20 is $2,100 a month.
On a per-resolution basis, Intercom Fin at $0.99 each means 600 resolutions is $594 a month and 1,200 is $1,188 a month. Note that the seat count is irrelevant and the volume is everything.
On a flat basis, Matram Pro is $199 a month in both scenarios, which is $2,388 over a year either way. Disclosure that should shape how you read this: Matram is owned by the same people who publish this site.
Where that option struggles, stated plainly: there is no free tier; it answers questions but does not action tickets, so nothing gets provisioned or closed; and it is not an ITSM platform, so there is no asset register or change management. If your workload is request fulfilment rather than questions, the per-agent products are doing a job it does not do.
What each basis is really charging for
Per agent charges for your team size. An internal AI support agent serves the whole company, yet a per-agent basis prices it as though the IT team were the users. That mismatch is why support-shaped products get expensive internally for reasons unconnected to usage.
Per resolution charges for success. Externally a resolved conversation has commercial value, so the model aligns. Internally an answered question is purely a cost, so your bill rises precisely when adoption works, which creates a quiet incentive not to promote the tool.
Per user charges for licences. Fine for a support team of eight and poorly shaped for a whole-company audience, because the natural user count is your headcount.
Flat charges for nothing variable. Which is predictable and sometimes more expensive in year one at small scale. The value is forecastability rather than cheapness.
How to pick the basis before the vendor
What to do: write down three numbers before looking at any pricing page. How many people would need a seat. How many questions arrive a month. And what both of those will be in two years.
Checklist:
- Seats means people who resolve, not people who ask
- Count questions from a real fortnight, not from memory
- Project both numbers to your two-year headcount plan
- Check whether the AI capability is a tier feature or a paid add-on
- Check whether a flat price includes all seats or just an allowance
- Confirm whether the figures are monthly or annual-only
That last item catches people out. Several products in this market publish monthly-looking figures that are annual commitments, so a price that reads as a subscription is a year’s spend.
Verify everything against the source before committing, because pricing here changes within weeks: Freshservice, Crisp, Chatbot.com and Matram all publish openly.
The add-on that breaks most budgets
On the per-agent products, the AI capability is frequently priced separately from the tier, and that single line moves a budget more than the choice of tier does.
Freshservice publishes its Freddy AI Copilot at $29 per agent per month on top of whichever plan you choose, so a Growth seat with AI is $78 rather than $49.
Zendesk publishes its Copilot at $50 per agent per month paid yearly, which on a Suite Team seat takes the per-agent figure from $55 to $105.
In both cases the add-on is a larger proportional increase than stepping up a tier, and in both cases it is the capability you came for. Budget the combined figure from the start rather than the tier price, because a plan modelled on the tier alone will be wrong by roughly half.
Final thoughts
The published prices for an AI support agent sort the market in one order and your own arithmetic sorts it in another. At 12 seats and 600 questions a month, the four bases above produce monthly totals of $199, $594, $936 and $1,260 for products that do broadly similar things.
So the useful work is not comparing headline figures. It is deciding which variable you are comfortable having your bill attached to, and then checking what that choice costs when the variable grows.
How much does an AI support agent cost?
Between roughly $29 a month and a few thousand, depending almost entirely on the billing basis rather than on capability. Flat products start at $29 a month with seats included. Per-agent products run from $19 to $115 per agent per month before any AI add-on. Per-resolution pricing is about $0.99 per resolution. The right comparison is your own numbers against each basis, not the headline figures against each other.
Which pricing basis is cheapest?
It depends on two numbers: how many people need seats and how many questions arrive. At 12 seats and 600 questions a month, a flat product at $199 is the cheapest of the four bases by a wide margin. A small team with very low volume can find per-resolution cheaper. The crossover moves with both variables, which is why the sum has to be done with your own figures.
Why does per-resolution pricing worry people internally?
Because your bill rises exactly when the tool works. Externally that is reasonable, since a resolved customer conversation carries commercial value. Internally an answered employee question is purely a cost, so good adoption produces a bigger invoice and creates an incentive not to promote the thing you bought. Model it at the volume you would have if it succeeded, not your current volume.
Is flat pricing always better?
No, and it can be more expensive in year one at small scale. What it buys is forecastability: a figure that does not move when headcount or question volume moves, which is unusual in this market and genuinely valuable if you are growing. For a stable team of five with low volume, a per-agent or per-resolution basis may well cost less.
Is the AI capability included or extra?
It varies and it is the most common reason a budget comes in low. Several per-agent ITSM products price their AI assistant as an add-on that stacks on top of the plan, so a seat with AI can cost substantially more than the tier figure suggests. Ask explicitly whether the capability is a tier feature or a separate line, and get the per-agent figure for both.
What does flat with unlimited seats actually mean?
Check whether it means all seats or an included allowance, because both get described as flat. One product in our comparison publishes a flat workspace rate and then charges for agents above its allowance, which makes the published figure a floor. Another publishes a flat monthly price with seats genuinely unlimited. The difference matters most for a whole-company internal audience.
