AI Support Pricing Models Compared: Per Resolution vs Per Conversation vs Flat
Four billing models dominate AI customer support, and the same 1,000 conversations a month can produce a $25 invoice or a $594 one depending purely on which one you signed. Here is what each model meters, what behaviour it rewards, and the arithmetic to redo with your own numbers.
The 30-second answer
Per resolution pricing wins at low volume and loses badly at high volume. Per conversation pricing charges you for demand whether or not the AI helped. Per seat pricing taxes headcount, not usage. Flat rate wins once volume is high or unpredictable, and is genuinely the wrong purchase at very low volume, where a free tier costs less. The crossover is one division: flat monthly price divided by the per-resolution price. At Intercom's published $0.99 per outcome, a $95 flat plan breaks even at about 96 resolutions a month.
The four billing models that actually exist
Every AI support vendor uses one of four meters, or stacks several of them. Marketing language varies wildly. The meter does not. Once you know which one you are buying, the pricing page stops being confusing.
1. Per seat. You pay for each agent login per month. Zendesk publishes Support Team at US$19, Suite Team at US$55 and Suite Professional at US$115 per agent per month paid yearly. This meters your headcount, not your customers. It rewards a lean team and punishes the staffing patterns support actually needs: weekend cover, seasonal temps, an engineer who dips into the queue twice a week. The usual consequence is login sharing, which quietly ruins your per-agent reporting.
2. Per conversation. You pay for inbound volume regardless of outcome. Tidio sizes plans in billable conversations: the free plan allows 50 a month, Starter is $24.17 and Growth starts at $49.17. This meters demand, not quality. A conversation the bot failed costs the same as one it solved, and a successful campaign gets expensive. Teams on this model start burying the chat widget on high-traffic pages.
3. Per resolution, or per outcome. You pay only when the AI is judged to have solved something. Intercom prices Fin at $0.99 per outcome as of July 2026, and defines an outcome as a customer confirming resolution, or not asking for more help after Fin responds, or Fin completing a workflow, which Intercom notes includes handoffs. Zendesk bills per automated resolution but publishes no dollar figure, so you need a quote. On paper this is the most aligned model in the category. In practice your bill grows in direct proportion to the metric you are trying to improve.
4. Flat rate. You buy a tier and volume stops mattering. Crisp charges per workspace with unlimited conversations on every plan: Free, Mini at $45 a month for 4 seats, Essentials at $95 for 10 seats and Plus at $295 for 20 or more, with extra seats at $10. EzyConn, our own product, is also flat: Basic free forever with 2 seats and 100 AI conversations a month, Pro at $25 and Unlimited at $95. Flat meters nothing per chat, so it wins on spiky traffic and loses on tiny traffic. Watch for the asterisk: some flat plans still meter AI separately.
The four models side by side
Most real invoices combine two or three of these. A per-seat platform with a per-resolution AI add-on is the most common stack in 2026.
The same traffic, three volumes, wildly different bills
The assumptions are identical across all three scenarios: a three-agent team and a 60 percent deflection rate, meaning the AI closes 60 percent of conversations without a human. All figures were checked against public vendor pricing pages in July 2026.
Scenario A: 200 conversations a month
200 x 0.60 = 120 AI resolutions, with 80 reaching a human.
Per resolution: 120 x $0.99 = $118.80 in AI charges before seats. Intercom does not print a numeric seat price on its public pricing page, it routes you to a calculator, so the real total is $118.80 plus an unpublished seat line.
Per conversation: Tidio Starter at $24.17 covers the plan. Lyro AI is metered separately from $32.50 a month for 50 AI conversations, so the entry quota covers 50 of your 120. Add $16.67 to remove Tidio branding on that tier. Floor: $24.17 + $32.50 + $16.67 = $73.34, with the AI quota still short.
Flat: Crisp Mini at $45 includes 4 seats and $5 of AI credit, which Crisp states is roughly 90 automated conversations. That implies about $0.056 each ($5 divided by 90), so 120 costs about $6.67, or $1.67 over the bundle. Call it $46.67. EzyConn Basic covers 100 AI conversations free, so 120 pushes you to Pro at $25.
Spread: $0 to about $119 plus seats on identical traffic. Tawk.to sits at the floor with a free core and unlimited agents, though its AI is an assist layer rather than a deflection engine.
Scenario B: 1,000 conversations a month
1,000 x 0.60 = 600 AI resolutions, 400 to humans.
Per resolution: 600 x $0.99 = $594 a month, or $7,128 a year, plus seats. Your traffic multiplied by five and so did the AI line.
Per conversation: Tidio Growth from $49.17, plus a Lyro quota sized for 600, plus $16.67 for branding. Tidio publishes quota steps up to 1,000 or more, but only the $32.50 entry price is public, so the realistic floor is $115.34 with the AI line still to be quoted.
Flat: Crisp Essentials at $95 includes 10 seats and $25 of AI credit, stated as about 450 automated conversations. 600 costs roughly $33.33 at the same derived rate, about $8.33 over, landing near $103. EzyConn stays at $25 on Pro, or $95 on Unlimited.
Spread: about $25 to $594. The per-resolution bill is now more than twenty times the cheapest flat option, on exactly the same conversations.
Scenario C: 5,000 conversations a month
5,000 x 0.60 = 3,000 AI resolutions, 2,000 to humans.
Per resolution: 3,000 x $0.99 = $2,970 a month, or $35,640 a year, plus seats.
Per conversation: Tidio Plus starts at $300 a month plus monthly usage, and Tidio does not publish that usage rate, so it needs a quote.
Flat: Crisp Plus at $295 includes 20 or more seats and $75 of AI credit, stated as about 1,350 automated conversations. 3,000 costs roughly $166.67 at the derived rate, about $91.67 over, so near $387. EzyConn Unlimited is $95 with unlimited seats.
Spread: roughly 30x between cheapest and most expensive. The per-resolution vendor is not overcharging. It is doing exactly what it said, charging for outcomes. The model simply has no ceiling.
The hidden meters buyers miss
The headline model is rarely the whole invoice. Five smaller meters explain most of the gap between quoted price and actual charge.
Branding removal. A clean widget is often an add-on rather than a feature. Tidio charges $16.67 a month to remove its branding on Starter and Growth, bundling custom branding only from Plus, which starts at $300. Tawk.to confirms a fee exists but does not publish the price. Budget for it, or accept someone else's logo.
AI credit bundles inside flat plans. Flat does not always mean unmetered. Crisp offers unlimited conversations but funds AI through credit bundles of $5, $25 and $75, and its free tier includes no AI credits at all. Because those bundles are denominated in dollars rather than conversations, the conversation equivalent moves whenever model costs move.
One-off versus recurring free allowances. One word changes everything here. Tidio's free plan includes 50 Lyro AI conversations described as one-off, so free-tier AI stops permanently once spent. EzyConn's free Basic plan includes 100 AI conversations every month. One is a sample, the other is a running capability, and a comparison table renders both as a number.
Seat minimums and bundles. Tiers that bundle seats look generous until you need one more. Crisp Plus starts at 20 or more seats, and extra seats cost $10 each. If you are a team of six on a plan sized for four, that is a recurring cost nobody quoted you.
Annual lock-in. Headline rates are often the paid-yearly rates. Zendesk's $19, $55 and $115 per agent are all quoted as paid yearly, and monthly billing costs more. The saving is real, but so is committing twelve months to a model you have not stress-tested. Our guide to AI chatbot pricing and the cost per conversation breakdown go deeper on these mechanics.
How to forecast your own bill in ten minutes
You need four numbers. Do this before any demo.
Step 1, pull real volume. Export inbound conversation counts for the last three months. Use your busiest month rather than the average, because you are buying capacity for peaks.
Step 2, apply a deflection rate. Most teams land between 40 and 70 percent once their content is decent. Use 50 percent if you have no data yet, and read our deflection rate guide to work out a defensible figure for your own question mix rather than a vendor's marketing number.
Step 3, multiply. Monthly bill = base plan + (peak conversations x deflection rate x AI unit price) + (seats x seat price) + add-ons. For 1,400 peak conversations at 55 percent against a $0.99 unit price: 1,400 x 0.55 = 770 resolutions, and 770 x $0.99 = $762.30 of AI before a single seat is counted.
Step 4, double it. Re-run the same maths at twice your volume. Under flat pricing the number does not move. Under metered pricing it doubles. That comparison tells you whether the model fits your growth curve, not just your budget. For the human-cost side of the equation, use our chatbot ROI guide and the ROI calculator.
Which model fits your situation
There is no best model, only a best fit for your volume shape. Find the row that describes you honestly, including the first one, which most vendors skip.
We sell a flat-rate product, so it is worth saying plainly where flat rate is wrong. If you handle forty conversations a month, a flat plan is money set on fire, and Crisp, Tidio, Tawk.to and EzyConn all publish a free tier you should take instead. If resolution quality genuinely matters more than cost, per-resolution pricing buys the strongest AI in the category and you should pay it. Flat earns its keep at volume, on spiky traffic, and when the bill needs to stop being a variable, which our EzyConn vs Intercom and EzyConn vs Crisp comparisons work through vendor by vendor.
Frequently Asked Questions
Which is cheaper, per resolution or flat rate?
Divide the flat price by the per-resolution price to find the crossover. At $0.99 per outcome, a $25 flat plan breaks even near 25 resolutions a month and a $95 plan near 96. Below that, per resolution genuinely wins. Above it, the gap compounds.
What actually counts as a resolution?
Whatever the vendor defines. Intercom counts an outcome when a customer confirms resolution, or does not ask for more help, or when Fin completes a workflow including handoffs, and charges once per conversation. Zendesk bills per automated resolution without publishing a rate.
Is flat rate ever the wrong choice?
Yes. At very low volume you buy headroom you will never use, and a free tier costs less than any flat plan. Flat only starts winning past the crossover point.
Why is this so hard to compare?
Partly because the models genuinely differ, and partly because stacked meters allow a low headline price. A $24.17 plan plus a $32.50 AI quota plus a $16.67 branding add-on is a $73.34 invoice, and only the first number appears in the ad.
Run your numbers against a flat rate
EzyConn is flat: Basic free forever with 100 AI conversations a month and no vendor branding, Pro $25, Unlimited $95. No per-conversation or per-resolution metering. If your volume is small, take the free plan and keep your money.
Start FreeLast updated . Pricing was checked against public vendor pricing pages in July 2026. Intercom publishes no numeric seat price, Zendesk publishes no per-resolution figure, Tidio publishes only its entry Lyro quota price, and Tawk.to does not publish its branding removal price. Verify current figures before purchasing. See also our Tidio alternatives guide or more guides.