Why 500 and 5000 tickets are two different pricing questions
At 500 tickets a month, almost every vendor's bill is dominated by its floor: the minimum seat count, the entry tier, the small bundle of AI answers included to get you started. A load that size is usually absorbed by one or two people who also do other work, so the seat line is short and the included AI allowance often covers the month. List prices look close together, and the differences that do appear feel like rounding.
At 5000, the picture inverts. Volume becomes the variable that drives everything else, and the billing unit you agreed to decides whether your cost rises in a straight line, in steps, or not at all. Two vendors whose entry prices look nearly identical can end up far apart, and the gap widens every month you grow. This is why comparing sticker prices is the wrong exercise. What you are really comparing is the slope.
What are you actually paying for: seats, resolutions, contacts, or capacity?
Most AI support pricing reduces to one of five units, and each one responds differently to growth.
Per seat. You pay for every teammate with a login, usually monthly, usually tiered by feature set. The cost tracks your headcount, not your ticket count, so it moves in steps rather than smoothly.
Per resolution or per AI answer. You pay when the AI closes a conversation, or sometimes when it simply replies. This is the unit that tracks volume most directly.
Per contact or per person reached. You pay by the size of the audience you can message, which tracks your customer list rather than your support load.
Per ticket tier. You buy a bucket of tickets and pay an overage rate above it. Costs look flat inside a tier and jump at the boundary.
Flat capacity. One price covers the whole workspace regardless of seats or conversation count.
Vendors also define their units differently. One counts a ticket, another counts a conversation that may contain several questions, another counts only resolutions the AI handled end to end. Comparing two published prices without first normalizing the unit produces a number that means very little.
How does each model behave when volume multiplies?
Per seat does not multiply with tickets, but it rarely stays still either. Ten times the volume usually means more people, more shifts, and more channels to cover outside business hours, and every one of those additions is another paid login. Many teams also add occasional users (a finance person checking a billing thread, a developer looking at a bug report) and pay full price for logins used a few times a week.
Per resolution is the model that most closely tracks volume, which is both its fairness and its catch. Your bill at high volume depends on how often the AI succeeds, so a better-performing bot produces a larger invoice. That is a defensible way to price, and some teams prefer paying only for work actually completed, but it means your cost and your success move in the same direction. Budgeting becomes a forecast rather than a line item.
Per contact is largely indifferent to ticket volume. If your support load grows while your customer list stays flat, this model can be the kindest one on the list. If you are growing the list fast, it behaves like a volume charge even when your ticket count is steady.
Per ticket tier is predictable inside a bucket and unpleasant at the edges. Seasonal businesses feel this most: a promotion or an outage pushes you into overage for one month, and you either eat the overage rate or upgrade to a tier you do not need for the rest of the year.
Flat capacity does not move. Your cost per resolved ticket falls as volume rises, which is the opposite of every other model here, and it makes the budget conversation short.
Which costs only appear after you grow?
The published plan price is rarely the whole bill at higher volume. Before you compare anything, check the following.
Channel gating. Voice, WhatsApp, SMS, and social inboxes are often reserved for higher tiers or sold as add-ons. A team at 500 tickets may run on web chat and email alone and never notice. At 5000, customers arrive everywhere, and the channel you need is frequently the one that forces a tier upgrade.
Carrier and platform pass-through. WhatsApp and SMS involve fees charged by Meta and by telecom carriers, separate from whatever your support vendor charges. Those are not in anyone's plan price and they do scale with volume.
Integration tiers. CRM sync, advanced routing, custom roles, and API rate limits are common upsell boundaries. The integration you can live without at low volume often becomes the thing holding the operation together at high volume.
Human time. Setup, prompt tuning, article writing, and ongoing maintenance all cost something even when they do not appear on an invoice. A tool that needs developer time to configure has a real price attached to it.
Commitment terms. Annual contracts, minimum seat counts, and mid-term upgrade rules all affect what you actually pay when your volume changes faster than your contract does.
How do the main vendors structure this?
Treat the following as structural descriptions rather than quotes, because every vendor revises pricing and you should read their current page before deciding.
Zendesk sells per-agent tiers with AI capabilities positioned as add-ons priced around automated resolutions. It is a deep, mature suite built for larger operations with complicated workflows, and the cost structure reflects that ambition. Teams weighing it against lighter options often start with a side-by-side like our Zendesk alternatives comparison.
Intercom combines a per-seat charge with a per-resolution charge for its AI agent. The product is well regarded and the resolution unit is transparent, but it is the clearest example of a bill that grows as the AI gets better at its job.
Freshdesk uses per-agent tiers with AI usage sold in bundles. It tends to be gentler at the low end, with the usual tier boundaries once you need advanced routing or analytics.
Gorgias is built around ecommerce and bills in ticket volume tiers with overage above them. For a store with predictable monthly volume that is easy to plan; for a store with spiky seasons it needs watching.
Chatwoot is open source. Self-hosting removes the license cost and replaces it with hosting, upgrades, and engineering attention, which is a genuine trade rather than a free option. Its cloud version is seat priced.
Help Scout has moved toward pricing tied to contacts handled rather than purely to seats, which decouples cost from team size and suits teams who want everyone in the inbox.
Tidio prices its AI in conversation packs, so the unit is the conversation rather than the resolution. At small volume this is simple and cheap; at larger volume it behaves like any other metered plan.
None of these is a bad choice. They are optimized for different buyers, and the right question is which unit matches the way your support load actually grows.
How do you model your own cost before you sign?
Build the comparison yourself, in a spreadsheet, in about half an hour.
Start by counting last month's volume by channel, separating repeat questions from genuine one-offs. Estimate what share of the repeat questions could be answered from your existing documentation, because that is the part an AI agent realistically takes on. Then, for each vendor, write down the billing unit and convert your volume into that unit rather than into tickets.
Now run the numbers twice: once at today's volume and once at ten times it. Add the hidden items from the section above to both. Finally, divide each total by the number of conversations you expect to resolve, and look at how that cost per resolution changes between the two scenarios. A plan that looks cheap today and triples its per-resolution cost at scale is a different product from one whose per-resolution cost falls. Run a third column at double your worst peak month, because the month that breaks a pricing model is never the average one.
It is also worth checking how much of the work depends on content you already have. A tool that reads your website and existing help articles starts producing value immediately, which is why a well-maintained knowledge base tends to change the economics more than any pricing negotiation.
What changes if the price does not move with volume?
A flat plan removes the volume variable from the decision entirely. You stop rationing seats, you stop treating a busy month as a budget event, and you stop doing arithmetic before switching on another channel. Corebee is priced this way: $99 a month, flat, with unlimited teammates and unlimited AI conversations, no per-seat fee and no per-resolution fee. Web chat, email, voice (an AI receptionist that tells callers it is an AI), WhatsApp, Slack, Facebook, Instagram, SMS, and Telegram are all live on that one price, so adding WhatsApp is a configuration step rather than an upgrade conversation.
The operational effect matters as much as the arithmetic. When logins are free, the finance lead and the developer can both sit in the shared inbox instead of having answers relayed to them. When conversations are unmetered, you can point every channel at the AI and let it hand off to a human whenever a customer asks or the answer is not clear from your material. Setup takes around ten minutes and does not need a developer, and the agent can book meetings through Cal.com or Calendly and sync records to HubSpot, Salesforce, Pipedrive, monday, or Zoho. The full breakdown sits on the pricing page.
Flat pricing is not automatically the cheapest answer for everyone. A team sitting at a few hundred tickets a month with one person answering them may pay less on a metered entry plan, and that is a reasonable choice. The distinction is what happens next: metered plans ask you to re-forecast every time you grow, while a flat plan answers the 5000-ticket question at the same moment it answers the 500-ticket one.