AI Customer Support for Small SaaS, Priced at 500 and 2,000 Tickets

The 60-second verdict

• At 500 tickets a month: Chatwoot's free Hacker tier covers it at $0 (2 agents, 500 conversations), and airys Solo covers it at $29, right at its cap. tiny.help is also $29, with no cap stated.
• At 2,000 tickets a month: Chatwoot's free cap is long gone. airys Team and Corebee both cost $99, with airys at its 2,000 cap and Corebee with no cap at all. Resolvly needs its $149 Pro tier.
• Most of the nine don't meter tickets at all. They meter conversations, resolutions or credits. Plain does not publish what one credit buys in tickets, and Chatwoot states its ratio only in its user guide.
• Corebee is the only one with a stated price that holds at any volume. Pylon is the opposite case: no number until you've sat through a demo.
• Per-resolution pricing inverts the usual incentive. Help Scout's AI Answers at $0.75 costs you more as it gets better at its job, so its ceiling figure in the table below is a worst case.
• Rates below were read from each vendor's own page between 15 and 17 September 2026, with eesel, airys and Plain re-read on 2 November 2026.

Nine tools built or positioned for a one-to-ten-person SaaS team, priced at 500 tickets a month and again at 2,000. The cheapest option at the lower volume is usually not the cheapest option at the higher one. Look at the meter to see why. Most of these vendors do not meter tickets at all.

Corebee and tiny.help charge one flat number regardless of volume. eesel sells monthly batches of credits, one per ticket or chat. Resolvly, airys, Plain and Chatwoot sell a monthly allowance of something (conversations or credits) that is not the same unit as a support ticket, and none of the four states the exchange rate on its pricing page.

Read the allowance before the price. A $99 plan with a 2,000-conversation cap and a $99 plan with no cap are not the same product at 2,000 tickets, and the page selling you either one is unlikely to point that out.

What each tool actually meters

Rates read from each vendor's own pricing page between 15 and 17 September 2026, with eesel, airys and Plain re-read on 2 November 2026.

That split matters more than any single price. A tool billing per successful resolution gets cheaper the worse it performs (a bot that resolves nothing costs nothing), and a tool billing per credit can run out mid-month in a way a ticket count never shows you coming.

The shortlist, priced at both volumes

Assumptions stated up front, because every total below depends on them. Team size is 2 seats where a tool charges per seat. Help Scout bills per successful resolution, so its total is the cost if every ticket in the volume were resolved automatically. That figure is a ceiling, and the row labels it as one. eesel's totals assume its agent works on every ticket.

One row says "ceiling" and one says "unclear." A per-resolution vendor cannot be modelled to an exact monthly bill without knowing how many of your tickets it actually resolves, and Help Scout publishes no average resolution rate you could plug in. The only tool here with a stated price that holds at any volume is Corebee. Pylon is the opposite case, and will not give you a number until you talk to sales.

The one public number that turns that ceiling into a likely bill, and who published it

There is exactly one public figure we found that lets you convert a per-resolution ceiling into a realistic estimate for B2B SaaS, and it comes with an interest worth stating before the number: AI resolves B2B customer support requests end to end about 15% of the time, against roughly 35% in B2C. Published 28 May 2026.

Apply that to the ceiling above and the picture changes materially. At 2,000 tickets a month, Help Scout's $1,500 AI ceiling implies roughly $225 in practice, plus $50 for two seats. It also cuts against eesel's model, which charges a credit for every ticket its agent works on, resolved or not.

Three more figures from the same source, each of which cuts against the marketing. Two-thirds of the time in human-AI hybrid setups, the AI looks at the question first and then triages it to a human without saying anything. About a third of all AI-assigned inbound is unwanted noise. And requests take 5.3 messages to resolve on average against 3.9 at baseline, so the AI adds messages to the conversation even when it helps. When it does actively engage a ticket, it cuts the human's workload by about a third.

The interest: that data is Pylon's own, published in collaboration with a16z, which led Pylon's Series A and co-led its Series B. Pylon is one of the nine tools here. No sample size or methodology is stated. Treat the 15% as directional. It is not a rate you can hold a vendor to.

Corebee: the flat-rate case

• One plan: $99/month, unlimited teammates, conversations and AI resolutions across chat, email and WhatsApp. 14-day trial, no card required, 30-day money-back guarantee.
• It documents a governance control most of this shortlist doesn't: refunds over $100 route to a human for approval rather than letting the agent act alone.
• The risk here is durability. A single-plan vendor at $99/month has a thinner margin cushion than a venture-funded per-seat platform, and a support tool disappearing is a worse outage than an expensive one. Supp, a small vendor in this same bracket, posted on 3 October 2026 that it had shut down its product because nobody was paying for it. Nothing here says Corebee will follow.

tiny.help and the unpublished ceiling

• Freelancer: $29/month, three sites, three people, 14-day trial, billed via Stripe from day 15. Its own copy: "$29 a month after. Your agent does the setup. No per-answer fees, ever". That line targets Intercom's per-seat and per-resolution metering explicitly.
• No conversation or ticket cap is stated anywhere on the page. "Unlimited AI-powered responses" is listed as a feature on every plan, and a tool advertising zero metering at this price is either subsidising heavy usage or hiding a fair-use limit. Both are common in this bracket.
• Treat "no cap found" as a question to ask before you commit a real support queue to it. It is not a confirmed unlimited plan.

airys and Resolvly: the conversation-cap pair

airys positions itself as "AI support for indie SaaS" specifically, answering from a help-centre knowledge base with unlimited seats on every tier.

• Free: 50 conversations/mo. Solo $29: 500 conversations. Team $99: 2,000 conversations. Unlimited articles on every plan (read 2 November 2026).
• At exactly 500 tickets/mo, Solo is at its cap, and one conversation over means Team at $99, a 3.4x step. At exactly 2,000, Team is at its cap, and the page lists no plan above it.

Resolvly serves a different buyer than its evidence trail suggests.

• Its own site names six target verticals: property management, legal, healthcare and dental, restaurants, e-commerce, home services. SaaS is not one of them. That doesn't disqualify it, though a SaaS team should notice they aren't the built-for audience.
• Pricing: Starter $49/300 conversations, Pro $149/2,000, Business $399/10,000.
• Marketing claims, Resolvly's own: "$0.05 per AI-handled conversation — 120x cheaper than human agents," plus a 70% autonomous resolution rate, both with no methodology attached.
• The $0.05 figure doesn't obviously reconcile with the $49-for-300 Starter tier, which works out closer to $0.16 a conversation. The published page does not explain the gap.

eesel: a credit for every ticket it works on

eesel is the one tool here built explicitly to sit on top of a help desk you already run rather than replace it.

• $299/mo for 500 credits, $499 for 1,000, $849 for 2,000, $1,749 for 5,000, read 2 November 2026. A ticket or chat is one credit, however long. No platform fee, no seat fee, 100 trial credits.
• Credits do not roll over, and optional overage is $0.80 each. The rate falls with size, from about $0.60 a ticket at 500 to about $0.35 at 5,000.
• You pay for tickets the agent works on, whatever the outcome. Spam and categories you tell it to skip do not count.

Where seats come back: Help Scout, Chatwoot, Plain

• Help Scout is priced the way most of this site's directory is: per seat, with AI bolted on separately. Free: 5 users, 100 contacts/mo. Standard $25/user, Plus $45, Pro $75 (10-seat minimum). AI Answers add-on: $0.75/resolution, with a 3-month unlimited-resolutions trial before the meter starts.
• It is the only per-resolution vendor here that publishes what the unit actually is. Its documentation defines a resolution as "a single AI chat session that is resolved without help from a human," and states "you'll only be charged once per session, even if AI Answers replies to multiple questions during that session." Billed in arrears.
• That definition changes the arithmetic. A customer who asks four questions in one session costs $0.75 in total, where four separate sessions would cost $3.00. So the $375 and $1,500 ceilings in the big table only hold if every ticket is a separate session. For a chat-led product where customers batch their questions, the real figure is lower.
• One wrinkle worth knowing: Help Scout's reporting labels ("Contact helped," "Contact not helped," "Human escalation") live on a different documentation page from its billing definition, and they are not the same thing. The outcome labels are analytics; the session is the meter.

• Chatwoot's free Hacker tier: 2 agents, 500 conversations/mo, live chat only. That covers 500 tickets a month at $0. Startups $19/seat (unlimited conversations, 300 Captain AI credits); Business $39/seat (500 credits); Enterprise $99/seat (800 credits). Extra credits: $20/1,000.
• The credit ratio is published, though not on the pricing page. Chatwoot's user guide states "all actions consume 1 credit per message, since a fixed model configuration only is supported". So 300 bundled credits buys roughly 300 messages. It does not buy 300 conversations. A Startups-tier team at 2,000 tickets is well past the bundle and into $20-per-1,000 top-ups.
• Credits are consumed by more than customer replies: Captain assistant responses, Copilot lookups, editor actions (rephrase, summarise, suggest a reply), label suggestions, audio transcription, and "any workflow or automation that triggers a model call." Usage does not reset when you top up, change plan or change seat count.

• Plain's Foundation: $45/user/mo billed annually ($55 otherwise), up to 5 member seats, 1,000 credits/mo. Horizon: $99/user/mo, minimum 3 seats, 2,500 credits. Frontier: custom. Extra credits are $10 per 1,000 (read 2 November 2026).
• Plain's pricing page never states what one credit costs in tickets, so the 1,000-credit figure can't be checked against your volume. Ask directly before committing a queue to it.

Pylon: the one that will not price itself at all

Pylon's own pricing page, read on 16 September 2026, carries no plan names, no per-seat rate and no usage rate. There is only a form to "Schedule a personalized 30-minute demo." For a one-to-ten-person SaaS team trying to compare nine options in an afternoon, that is a real cost: every other row in this table took minutes to read, and this one takes a booked call and a follow-up email before you have a number to compare. For Pylon, the sales-call requirement is the finding itself.

The seat trap, run in reverse

Every prior comparison this site has published in this category (Gorgias, Zendesk, the enterprise agentic vendors) found the same shape: price rises with team size, and a growing headcount pushes the bill into the next bracket. This shortlist mostly doesn't work that way.

• Charge nothing extra for seats, or don't mention them at all: Corebee, tiny.help, airys, Resolvly, eesel.
• Meter by seat, with AI billed separately: Help Scout, Chatwoot, Plain.
• For this bracket, ticket volume and credit consumption move your bill. Headcount does not. A small SaaS team hiring a fourth or fifth support person pays the same Corebee or airys bill it paid at two. That inverts the dynamic this site has documented everywhere else in the category.

What founders actually say

Three founder accounts, read as accounts rather than as findings.

A three-person SolidWorks-plugin shop, on Hacker News in February 2026, explained why they built rather than bought: "They're too small for Zendesk, too busy to answer the same questions repeatedly, and too niche for generic solutions," and, on the client itself, "They're a three-person shop — they don't have a support team or an IT team." That is close to a textbook description of the buyer this whole shortlist is aimed at. Worth noting too: the poster's answer was to write a custom Flask bot instead of adopting any commercial platform. The option to "build it yourself" remains a real competitor to every tool in this table.

“If you're launching a new webapp, pretty quickly you'll need a way to manage customer queries, onboarding, collate user feedback and track actions with analytics.”— Show HN: HeyDeacon, 11 August 2026

That post, introducing a drop-in SDK called HeyDeacon aimed at SaaS founders launching a new product, did not publish a price in the thread itself; a commenter described it only as "a much cheaper and friendlier way to handle support" than the alternatives they had tried, without naming a number. Read it for what it is: one founder's account of one build, and no more than that.

A June 2026 Indie Hackers post by a founder writing as EllaReid_24 reframed the whole comparison exercise in a way worth carrying into any shortlist: "Deflection is a vanity metric. Resolution is the business metric." The post's advice for lean teams named YourGPT and Freshdesk, which sit outside this shortlist, but the distinction generalises directly to every row in the table above. A tool that answers a lot of messages and a tool that closes a lot of tickets are not the same purchase, and a conversation-metered vendor has no particular incentive to tell you which one it is.

What breaks at renewal

The failure mode in this bracket is a bill nobody modelled. Four specific ways it happens:

• A conversation-capped plan (airys, Resolvly) gets crossed by a single good marketing month, and the jump to the next tier arrives in full-tier increments.
• A credit-metered add-on (Chatwoot, Plain) runs out mid-cycle in a way a ticket counter never shows coming, because nobody's watching the credit balance the way they'd watch a ticket queue.
• A per-resolution add-on (Help Scout AI Answers) costs more as its automation gets better. That is a genuinely different incentive from every seat-priced tool here, and it is easy to forget once the bill starts climbing.
• A single-plan flat-rate vendor (Corebee, tiny.help) is the easiest bill to forecast right up until it isn't. Fair-use limits and vendor durability are the risks a flat number hides while leaving them in place.

The verdict

At 500 tickets a month with two seats, Chatwoot's free Hacker tier is the cheapest entry with a stated allowance that covers the volume, followed by airys Solo at $29. Both sit exactly at their caps. tiny.help is also $29 with no cap stated, so get its fair-use limit in writing first.

At 2,000 tickets, the free tier is gone and $99 is the number to beat. airys Team and Corebee both charge it, and Resolvly needs its $149 Pro plan. Choose Corebee over airys if you expect to grow past 2,000, because it has no cap to cross and airys lists no plan above that.

If you already run a help desk you like, eesel is the option built to sit on top of it, from $299 for 500 tickets. Use its 100 trial credits to see how much it resolves, because you pay for every ticket it works on.

If you are already committed to a seat-priced help desk, Help Scout's $0.75-per-resolution AI Answers is the smallest possible add-on to test the idea without a platform migration, and its three-month unlimited trial removes the cost of finding out.

Which one to start with, by what you actually look like

Seven small-SaaS profiles against the constraint that decides the answer.

Prices read from each vendor's own pricing page between 15 and 17 September 2026, with eesel, airys and Plain re-read 2 November 2026. All three had changed in those seven weeks. If a figure here has moved, tell us and it gets corrected and re-dated.

WRITTEN BY AR · UPDATED 2026-11-02

I read the fine print. Vendor pricing pages, billing definitions, terms, funding filings and acquisition notices — then I do the arithmetic nobody publishes: what a platform actually costs at your volume, what its headline metric is really counting, and who owns it now. I do not run benchmarks, and no page here pretends otherwise.

Editorial policy