How Much Does AI Customer Support Cost? Flat Plans vs. Per-Resolution Pricing

If you have ever opened a customer-support software pricing page, scrolled past three tiers of seat fees, and closed the tab feeling more confused than before — you are not alone. AI support pricing is genuinely confusing right now, because vendors charge in completely different ways. Some sell you a flat monthly plan. Others charge you every time the AI actually solves something.

Both models can be fair. Both can quietly drain your budget. The difference is math — and this article does the math for you.

Strip away the marketing, and nearly every AI support product uses one of two models.

Flat plans. You pay a fixed amount per month and get a bundle of AI answers included — for example, Starter at $9 per month with 3,000 AI answers included. If you stay inside the bundle, your bill never moves. If you outgrow it, you move to the next plan or buy an extra block of answers at a published price.

Per-resolution billing. You pay for each conversation the AI resolves — one solved customer problem, one charge. Intercom's Fin agent, the best-known example, publishes $0.99 per resolved conversation, on top of seat fees for the helpdesk itself (from $29 per seat per month on annual billing). Nothing is bundled. Ten resolutions cost ten times one resolution.

The pitch for per-resolution pricing is appealing: you only pay when you get value. The pitch for flat plans is predictability: you know the bill before the month starts. To decide which is cheaper for you, ignore the pitches and run your own numbers.

Take a small business whose website chat handles 500 customer conversations a month. Assume the AI resolves most of them without human help — that is the optimistic case vendors show in demos.

At $0.99 per resolution, 500 resolved conversations cost about $495 per month — before seat fees. Add two support seats at $29 each and you are at roughly $553 per month. For 500 conversations.

Now scale it. At 3,000 conversations a month, the same math gives roughly $2,970 plus seats — over $3,000 per month. At 10,000 conversations, you are looking at roughly $9,900 plus seats every single month. The cost grows in a perfectly straight line with your success: the more customers talk to you, the more you pay, forever. There is no volume discount built into the unit price, only the privilege of paying it more times.

To be fair, per-resolution billing has one honest advantage: months when nobody writes to you cost almost nothing in usage fees. If your demand is extremely spiky — dead quiet for weeks, then a flood — you are not paying for capacity you do not use. And large companies with existing support teams sometimes prefer it because the AI line item maps neatly onto the tickets it replaced.

But notice what the model quietly assumes: that every conversation ends in a paid resolution. In real life, some conversations escalate to a human, some visitors abandon the chat, and some questions need a follow-up. You still paid your team to handle those — the per-resolution fee just happens to be the part of the bill you can see clearly.

Now the same volumes on flat plans. Pravia's published plans are a good reference because every number below is on the pricing page: Free at $0 with 50 AI answers a month, Starter at $9 with 3,000, Pro at $19 with 12,000, and Business at $49 with 50,000. Extra answers, if you ever need them, are $5 per 1,000 — also published, no surprises.

The pattern is hard to miss. At 500 conversations, flat costs less than two percent of per-resolution billing. At 10,000 conversations, the gap is roughly $9,900 every month — more than a full-time employee's salary in many countries — for the same answered questions.

The reason is structural, not promotional. Answering one more chat question costs the vendor fractions of a cent in computing power. A flat plan shares that efficiency with you. Per-resolution pricing does not: the unit price stays fixed while your volume grows, so the vendor captures all of the efficiency gain.

Unit prices and plan fees are only the visible part. Before you sign anything, check these four lines of the contract.

Seat minimums and platform fees. Some vendors require a helpdesk subscription on top of the AI fee — every human agent needs a paid seat whether or not they touch the AI conversations. A $0.99 headline can easily double once seats, add-ons, and messaging-channel fees join the bill.

Monthly minimums. Per-resolution products often include a minimum commitment — for example, paying for at least 50 resolutions a month even in a quiet month. Ask what the floor is, not just the unit price.

What counts as billable. Is a resolution counted when the customer confirms the answer helped, or merely when they stop replying? Definitions differ between vendors, and an "assumed resolution" after a customer silently leaves can inflate your bill. Read the definition, not the headline.

Overage behavior. The scariest bill is the one you did not approve. Some platforms keep serving answers past your limit and charge overage automatically; others pause and ask. A flat plan with explicit, opt-in extra blocks — $5 per 1,000 answers, bought deliberately — is the version of overage that never produces a horror story.

Here is the honest rule of thumb. Per-resolution billing fits large support operations with steady, high-value conversations, an existing helpdesk team, and a finance department that reconciles AI spend against deflected tickets every quarter. They have the volume to negotiate and the staff to audit definitions.

Flat plans fit everyone else — and that is most small and medium businesses. If you handle hundreds to tens of thousands of conversations a month, want one predictable number on the invoice, and would rather spend your time serving customers than auditing resolution definitions, flat wins by a margin so large it is barely a contest. Start on a free tier with real limits (50 answers is enough to prove value), upgrade to $9 when customers actually use the chat, and move to $19 or $49 when growth demands it. Each step is a deliberate decision, never a surprise.

One more test worth running: ask every vendor for the total bill at your current conversation volume, seats included, in writing. The vendors with fair pricing answer in one sentence. The rest send you a calculator link.

AI customer support costs either a flat monthly fee or roughly $1 per resolved conversation plus seats — and at small-business volumes, those two numbers live on different planets. Five hundred conversations can cost $9 or $553 for the same answered questions. Run the table above with your own volume before you commit, check minimums and seat fees, and prefer pricing you can explain to your accountant in one sentence. If you want to see what the flat side looks like in practice, compare the plans on the pricing page or read the detailed breakdown in Pravia vs. Intercom.