All AI guides
Buyer's guide13 min read

Best AI Customer Support Software in 2026: What Per-Resolution Pricing Actually Costs

Per-resolution pricing charges an AI customer support vendor only when its software closes a conversation, unlike seat-based helpdesk software billed per agent. In October 2026 Intercom Fin charges $0.99 per outcome, Open $0.75 to $0.90, and Zendesk $1.50 to $2.00, while eesel AI bills every ticket it handles, resolved or not.

The short version

AI customer support software now bills per resolution instead of per seat, and vendors count different things. Published rates in October 2026 run from $0.50 at CoSupport AI to $2.00 at Zendesk, while Ada, Sierra and Decagon publish nothing. Model your own ticket volume against each rate before any demo.

Not sure which one fits you? Describe your situation and Smart Match names three tools.Find my AI tools

In this guide13 sections

Intercom charges $0.99 every time its Fin AI Agent closes a support conversation. A rival called CoSupport AI charges $0.50.

Ada reportedly charges up to $3.50. All three numbers describe the same promise: you pay only when the software solves a ticket. Across AI customer support software, the dominant pricing model in 2026 is no longer a monthly seat fee. It is a price per ticket the vendor says it closed.

For a team handling ten thousand tickets a month, the gap between a $0.99 definition of "resolved" and a $2.00 one is real money. Almost nothing written about this category explains that the definition changes by vendor, and so does the unit being counted.

This guide does the comparison the listicles skip: what eight vendors actually charge, how to model your own bill before a demo, which players will not publish a number, and where each pick stops making sense.

What changed since the 12 August review

We re-read every source on 27 October 2026. Four things moved. Two names joined.

  • eesel AI rebuilt its pricing. The old $0.40-per-task, pay-as-you-go plan is gone. It now sells monthly credit bundles, and one credit is any ticket or chat it handles, resolved or not.
  • Intercom added spending limits and refunds reopened outcomes. Fin now has spend alerts, and a conversation that reopens later is not billed. That weakens one of the arguments this guide made in August.
  • Decagon is younger than we said. It was founded in August 2023, not five years ago. The funding facts stand.
  • Two new per-resolution vendors are on Hokai. CoSupport AI and Open both publish rates well under Fin's, which changes the bottom of the price table.

The rest of the lineup holds: Decagon, Fin, Zendesk, Ada, Sierra, Chatbase and echowin are still selling what we described, at the prices below.

The category runs on resolutions now, not seats

For most of the last decade, support software was sold the way most SaaS is sold. You paid a monthly fee per human agent, whether that agent lived in the product or barely opened it. Zendesk still runs on that model at its core, with Suite Professional at $115 per agent per month.

The AI layer stacked on top bills differently. Zendesk's automated resolutions cost $2.00 each pay-as-you-go, or $1.50 with committed volume, after a small free allowance of ten per agent each month.

Intercom went further and built Fin's whole commercial model around the outcome. According to Intercom's own pricing page, Fin bills at most once per conversation, at $0.99, and needs no seat at all if you run it on another helpdesk. There is a minimum monthly commitment, given as roughly 50 outcomes.

Decagon raised $250 million in a Series D round led by Coatue Management and Index Ventures on 27 January 2026, tripling its valuation to $4.5 billion in under six months. It sells enterprise buyers such as Avis Budget Group, Block and Deutsche Telekom on the same logic: revenue tied to conversations closed, not seats provisioned.

That is the shift. The seat still exists, and most of these vendors will happily sell you one, but on any team where the AI closes a meaningful share of the queue it has stopped being the line on the invoice that grows when the business does.

What eight vendors charge per resolution

Put the published and reported numbers side by side and "per-resolution pricing" stops meaning one thing. Some vendors bill the resolution, some bill every ticket the AI touches, and two will not say.

VendorWhat you pay forRate (October 2026)Published?
CoSupport AIAI resolution$0.50 at any volumeYes
OpenResolved ticket$0.75 to $0.90Yes
Intercom FinOutcome, max one per conversation$0.99Yes
Zendesk AIAutomated resolution$1.50 committed, $2.00 pay-as-you-goYes
SierraResolution, inside a custom contractAbout $1.50 (estimate)No
AdaAI resolutionAbout $1.00 to $3.50 (reported)No
eesel AIEvery ticket or chat handled$299 for 500 credits, falling to $0.35 a credit at 5,000Yes
DecagonConversation or resolutionNegotiatedNo

Read the second column first. Three things in that table are easy to miss.

First, the lowest rate does not move with volume. CoSupport charges $0.50 per AI resolution whatever the queue size, and unresolved tickets are free, so a small team pays the same unit price as a large one. Open's rate, by contrast, falls as volume grows.

Second, eesel is the odd one out. It bills handling, not success. A ticket its agent fails on still costs a credit. In exchange, overage is off by default, and if you switch it on you set the monthly cap.

Third, Open prices voice at cost on top, with carrier minutes it puts under $0.10 per resolution. A phone-heavy queue should add that line before comparing.

The vendors that will not tell you a number

Ada, Sierra and Decagon share something beyond outcome billing: none publishes a rate card. Ada requires a demo, a volume conversation and a custom quote. A Salesforce AppExchange listing puts its entry point at $30,000 a year.

Procurement trackers report six-figure annual deals once a queue passes 100,000 tickets a month, which is where Ada does most of its business.

Sierra, co-founded by Bret Taylor, runs the same process. Third-party estimates put contracts from about $150,000 a year. Its annualized revenue doubled to $200 million between November 2025 and May 2026, according to Taylor.

That is not automatically a problem. Enterprise software has priced this way for decades, and large buyers with procurement teams, security reviews and volume commitments to trade are used to negotiating a number rather than reading one. But it does mean a buyer cannot shortlist these three by cost the way they can shortlist Fin, Zendesk, Open or CoSupport.

Any comparison that lists all eight on one price axis is smoothing over a real difference in how much you know before signing.

How to choose an AI customer support tool

Five questions separate a good fit from an expensive mistake, and none of them needs a demo call.

  1. Do you know your monthly ticket volume within about 20 percent? If not, an uncapped per-resolution contract is the wrong first purchase, because you cannot forecast the bill. Start with a capped or flat option until you have three months of data.
  2. What exactly counts as a resolution, in writing? Get the contract definition, not the marketing summary. Intercom counts an outcome when a customer confirms the fix, stops asking after Fin answers, or completes a handoff workflow.
  3. Is there a cap, a flat plan or a committed discount? eesel and Chatbase cap spend by design. CoSupport publishes one per-resolution rate for every volume, so there is no quote to wait for. Zendesk discounts committed volume.
  4. Does it sit on your helpdesk, or replace it? Fin, eesel and CoSupport can run on top of Zendesk or Freshdesk. Decagon, Ada and Sierra are usually the whole front line, which raises switching cost.
  5. What does a human-handled ticket really cost you? Include benefits, tooling and management time, not just salary divided by tickets. Without that number, no per-resolution price has a baseline to beat.

If the answers still leave three or four candidates, Hokai's Smart Match will answer a few questions and give you a shortlist matched to your queue, which is quicker than booking four demos.

A worked example, before you take a demo

Run the numbers first. Take a team fielding 8,000 tickets a month, where a demo claims 60 percent can be closed by AI. That is 4,800 resolutions a month, a plausible mid-range figure for well-documented product questions.

VendorRate usedAI cost per monthWatch for
CoSupport AI$0.50$2,400Unresolved tickets are free
Open$0.75 to $0.90$3,600 to $4,320Voice minutes billed on top
Intercom Fin$0.99$4,752Seats extra if you want Intercom's inbox
Zendesk AI$1.50 committed$7,200Suite seats and Copilot are separate
Ada$1.00 to $3.50$4,800 to $16,800Annual minimum around $30,000

eesel does not fit the same row, because it would bill all 8,000 handled tickets, not the 4,800 closed ones. Its 5,000-credit bundle is $1,749 a month, and the rest needs overage switched on or a custom quote.

The spread is roughly sevenfold. None of these numbers is wrong, because they describe different products with different definitions of done and different depths of automation. A vendor with a looser definition of "resolved" will often show a higher resolution rate in a demo, simply because it counts more conversations as closed.

The only honest comparison is your own ticket volume run through each vendor's published or quoted rate. Do not trust a resolution rate on a slide built from someone else's queue.

The 2026 shortlist

One line per pick, ordered from largest buyer to smallest.

  • 0 is the fastest-growing enterprise choice. Sacra estimates it reached $100 million in annualized revenue by July 2026. It is built for large volumes, not a first deployment.
  • 0 fits Fortune 50-scale contracts with six-figure annual minimums. It is not a mid-market first purchase.
  • 0 targets high-volume enterprise queues. Its lack of public pricing makes it a poor fit for a team that has not run a procurement process before.
  • 0 suits teams already on Suite seats who want automated resolutions without changing platforms, accepting that the AI fee stacks on an existing bill.
  • 0 is the most transparent published price, and it can run seat-free on other helpdesks as well as inside Intercom.
  • 0 is a newer per-resolution vendor priced below Fin, with the rate falling as volume grows.
  • 0 is the lowest published per-resolution rate on this list, $0.50 at any volume, with unresolved tickets free.
  • 0 is the most predictable entry point: a fixed monthly bundle and a cap you control, paid per ticket handled.
  • 0 fits a website-only chatbot at $32 to $400 a month on yearly billing, closer to a bot builder than a support platform.
  • 0 is narrower still, a phone-first AI receptionist for a small business fielding calls rather than tickets.

Other names worth a look if none of these fits: Parahelp and Giga sell enterprise agents on quoted contracts, kapa.ai answers technical questions from developer docs, and Pylon is a B2B helpdesk with AI built in. The full set lives in the customer support section of the directory.

Where CoSupport AI fits, and where its numbers need a second look

CoSupport AI is the newest addition here and the cheapest on paper. So it gets a closer read. It sells two products on one platform: CoSupport AI Agent at $0.50 per AI resolution, and CoSupport AI Assistant at $0.04 per AI response for teams that want AI drafting replies for human agents. CoSupport AI is a SupportYourApp company; the two merged in September 2026.

The AI Agent price is the same at any volume, tickets the AI does not resolve cost nothing, and no seats are needed to run it.

The first 1,000 AI responses are free on your real tickets, with no long-term contract. CoSupport says teams typically see 60 to 70 percent AI resolution in the first 60 days.

Two things deserve a second look before you sign:

  • The headline resolution figures are self-reported. CoSupport reports a 74 percent average AI resolution across clients on its 1.x architecture and up to 85 percent on the 2.0 architecture. Neither figure has been independently audited.
  • There is no self-serve signup. A pilot starts from a demo request. Connecting a helpdesk then takes about 10 minutes, and a full launch typically takes about 3 days.

Decagon versus Intercom: the comparison most buyers are actually making

Most teams in this category are not choosing between ten tools. They are choosing between the fastest-growing enterprise platform and the transparently priced agent they can switch on next week.

Decagon wins on scale and depth: multi-step account changes, returns and refunds across large volumes, the work that brought it more than 100 new enterprise customers in its last fiscal year. Fin wins on predictability. Turning it on does not need a new vendor relationship or a new procurement cycle if the team already runs Intercom.

For a team under fifty agents that has not standardized on a separate platform, Fin is very often the cheaper and faster path to a working AI agent. For queues in the hundreds of thousands of conversations a month, Decagon's purpose-built infrastructure is the one designed for that load.

A third option now sits between them. A team that wants Fin's predictability at a lower unit price can pilot Open or CoSupport on top of its existing helpdesk, then hold either to its published rate.

Data handling: where your tickets go

Every tool on this list reads your support history. That means names, emails, order numbers, delivery addresses and, in a fair share of e-commerce and subscription queues, the last four digits of a card or a refund amount that a regulator would also like to know about. The certifications differ more than the marketing suggests.

CoSupport's security page lists ISO 27001, GDPR and CCPA, a dedicated server per client, and a statement that one client's data never trains another client's model. It does not list a SOC 2 report or a HIPAA business associate agreement. eesel says its SOC 2 Type II audit is underway and offers a HIPAA BAA on its Enterprise plan.

Before any pilot, ask every vendor for the same four documents:

  • the data processing agreement and its sub-processor list;
  • the training policy, in writing, for your tickets and your customers' messages;
  • the retention period for transcripts, and how deletion requests are handled;
  • the current audit reports, with dates, rather than a list of badges.

Ask anyway. A vendor that cannot produce those within a week of a demo is telling you something about how procurement, onboarding and every renewal after it will go.

Who should skip AI customer support software this year

A team fielding fewer than a few hundred tickets a month should not buy any of these yet. Even the cheapest bundle on this list adds overhead that one part-time agent could clear without automation.

The math favors an AI layer only once volume is high enough, and repetitive enough, that the per-resolution price beats the loaded cost of a person handling the same request. Below that line, a good FAQ page and a shared inbox will beat a metered AI contract on cost per solved ticket.

Teams whose queue is mostly one-off technical investigations should also wait. Give it a year. Every vendor here, CoSupport included, is built for repeat questions about orders, billing and accounts.

The best argument against outcome pricing, and why it does not settle the question

The obvious rebuttal to everything above is that outcome pricing is the buyer-friendly model. You only pay when the software does its job, so the vendor's incentives line up with yours by design. That holds only if "the job" is defined the way the customer would define it.

Intercom's definition includes a conversation where the customer simply stops replying after Fin answers. A customer who gives up, rather than one whose problem was solved, can produce the same line on the invoice.

To Intercom's credit, a reopened conversation is now refunded, which closes part of that gap. It does not close the case of a customer who leaves quietly, decides the answer was not worth arguing with, cancels at the next renewal and never opens another conversation that anyone could reopen.

Get it in writing. Ask for the exact billable definition before the demo. Then ask what share of a recent customer's resolutions were reopened within a week. That single number says more about whether a $0.99 or a $2.00 resolution is the better deal than the sticker price will.

What would change this verdict

This guide ranks on published price, contract transparency and fit. Four developments would reorder it:

  • Sierra or Ada publishing a rate card. Either would move up for mid-market buyers overnight.
  • An independent resolution audit. Every resolution rate here is vendor-reported. A third-party benchmark on shared ticket sets would matter more than any price cut.
  • Open quoting small teams near its floor. If its $0.75 rate reaches a 2,000-ticket queue, it joins CoSupport, which already charges $0.50 at any volume, as a default first pilot.
  • Fin dropping the 50-outcome minimum. That would make it the obvious choice for very small teams, too.

For a wider view of the bots that sit in front of these systems, see Hokai's guide to AI chatbots in 2026.

Frequently asked questions

What does "per resolution" pricing mean for AI customer support software?

The vendor charges each time its AI agent closes a support conversation, rather than a flat fee per human seat. What counts as closed differs by contract: Intercom bills at most one outcome per conversation and refunds it if the conversation reopens, while eesel bills every ticket its agent handles, solved or not. Read the written definition before comparing sticker prices.

How much does Decagon cost compared to Intercom's Fin AI Agent?

Decagon publishes no rate card and negotiates every enterprise contract, billing per conversation or per resolution. Fin is $0.99 per outcome and can run without Intercom seats on another helpdesk, with a minimum monthly commitment of roughly 50 outcomes. For a team under fifty agents, Fin is usually the faster and cheaper start.

Which AI customer support tool is cheapest for a small team?

For a predictable bill, eesel AI's smallest bundle is $299 a month for 500 handled tickets, with overage off unless you turn it on. CoSupport AI publishes the lowest per-resolution rate, $0.50 at any volume, with unresolved tickets free. Chatbase starts at $32 a month for a website-only bot.

Do Ada and Sierra publish their AI customer support pricing?

No. Both require a demo and a custom quote, with no self-serve tier. Reported figures put Ada at $1.00 to $3.50 per resolution with a $30,000 annual entry point, and Sierra near $1.50 per resolution inside contracts estimated from $150,000 a year.

How should a team estimate its AI customer support costs before signing a contract?

Multiply your own monthly ticket volume by a realistic AI resolution rate, then by each vendor's published or quoted price, not the rate shown in a demo. Check whether the vendor bills resolutions or handled tickets, since that changes the base number. Teams that cannot forecast volume within about 20 percent should start with a capped or flat plan.

Covered in this guide

  • Ada: Enterprise AI customer service platform that autonomously resolves support interactions across chat, voice, email, and social channels, built for large Zendesk and Salesforce deployments.
  • CoSupport AI: AI support agent that resolves repetitive tickets in the customer's language across chat, email and voice, billed per AI resolution, with no seats.
  • eesel AI: eesel AI automates customer support on Zendesk, Freshdesk and Intercom, from $299 a month for 500 handled tickets or chats. Claims 81% autonomous resolution for 2,000+ companies.
  • Fin AI Agent: Customer service platform for SaaS teams: the Fin AI agent resolves most incoming queries automatically, and Intercom serves more than 30,000 organizations worldwide.
  • Open: Open is an AI customer support platform that automates 77% of tickets across voice, chat, email, WhatsApp and SMS, billing per ticket resolved.
  • Sierra: Enterprise AI agent platform from Bret Taylor valued at $15.8B in 2026, used by 40% of the Fortune 50 for customer service across chat, voice, and email.
  • Zendesk: Omnichannel customer support platform with AI agents, ticket management, and 1000+ integrations for 100,000+ businesses.
  • Chatbase: AI chatbot platform for customer support. Train on docs, PDFs, website, or Notion. Deploy to web, Slack, WhatsApp, Instagram. Free tier. $40/mo Hobby plan.
  • Decagon: Decagon is a $4.5B-valued AI concierge that automates chat, voice, email and SMS support for enterprises, with contracts typically starting near $95K/year.
  • echowin: AI phone receptionist that answers calls 24/7 in 30+ languages with 7,000+ integrations for small businesses and agencies.
  • Giga: AI support agent platform that resolved 90%+ of DoorDash's support tickets across voice, chat, and email in 99 languages.
  • kapa.ai: AI documentation and knowledge-retrieval assistant used by companies like OpenAI and Docker to auto-answer technical support questions from their own docs.
  • Parahelp: AI support agent from an 11-person YC-backed team that resolves complex tickets end-to-end via email, chat, and Slack for Perplexity, Framer, and Replit.
  • Pylon: Pylon is a B2B customer support platform where AI agents pre-investigate every ticket across Slack, email, and chat; trusted by 1,500+ companies.

Sources

Still deciding?

This guide covers a handful of options. Smart Match checks every listing in the directory against how you actually work and what you can spend, then hands you the shortlist and the reason behind each pick.

Find your tools

Keep reading

Find your toolsAll AI guides