If you run a side-hustle store — Etsy, Shopify, or both — you have probably seen the pitch: AI customer service for under a dollar a ticket. That number is real, and it is also misleading. Fin charges $0.99 per resolution, Zendesk charges $1.50 per verified resolution on a committed plan, and HubSpot charges $0.50 per resolved conversation. But none of those prices is the whole bill. Seats, plan minimums, ticket-counting quirks, and the conversations the AI fails to resolve all stack on top, and a bundle that looks cheap per unit can cost more per resolved ticket than a pricier tool if half its credits go unused.
This article converts every major tool’s pricing into the one number a side-store owner actually budgets: cost per resolved ticket at real volumes. All prices below are list prices as reported in October 2026 comparisons and vendor coverage; check the vendor’s current pricing page before you buy, because these rates move and bundle sizes change. Where sources conflict, I say so rather than picking the friendlier number.
The rate card: nine tools, one unit (almost)
Here is the per-resolution pricing for the nine tools in the October 24 comparison, updated against October 2026 coverage:
| Tool | AI price | Base plan it sits on | Catch to watch |
|---|---|---|---|
| Intercom Fin | $0.99 per outcome | Seats from $29 per seat per month | “Outcome” includes workflow completions and handoffs, not only full resolutions |
| Zendesk AI | $1.50 per verified resolution (committed), $2.00 pay-as-you-go | Suite seats, plus Copilot at $50 per agent per month | Only verified resolutions bill; Copilot is a separate add-on |
| HubSpot Breeze (Customer Agent) | $0.50 per resolved conversation | HubSpot plan, with included credits before overage | Cheapest per unit, but the plan it requires matters |
| Gorgias AI | $0.90 per resolution (annual), $1.00 (monthly) | Ticket-based plans from $10 per month | Each AI resolution also counts as a ticket, so it is billed twice in effect |
| Help Scout Envoy | $0.75 per resolution | Seats from $25 per seat per month | Seat cost stacks on top of resolutions |
| Tidio Lyro | $32.50 per month for 50 conversations | Tidio plan from $29 per month | Bundle pricing: unused conversations are wasted money |
| Freshdesk Freddy | $49 per 100 sessions, plus Copilot at $29 per agent | Seats from $29 per agent per month | Sessions are not resolutions — a session can end without solving anything |
| Crisp | $45 per month Mini plan, including $5 of AI credits | Crisp plan | Credit allowance is small; overage terms are unclear |
| eesel AI | $299 per month for 500 credits | Standalone | Credit pricing: unused credits raise your effective per-ticket cost fast |
Two tools outside the original nine are worth adding because side-store owners keep asking about them. Salesforce Agentforce runs about $2.00 per conversation — note the unit is a conversation, not a verified resolution, so it is not directly comparable to the $0.99 and $1.50 figures. And several comparison sites list Richpanel at roughly $0.20 per conversation on average with a $200 monthly minimum, but that average is the vendor’s own claim and depends heavily on conversation length, so treat it as unconfirmed until you test it on your own traffic.
The first lesson is already visible in the table: the cheapest per-unit price is not the cheapest tool. HubSpot at $0.50 looks like half the price of Fin, but if you are not already on HubSpot, the plan you must buy to use it changes the total. Gorgias at $0.90 looks cheaper than Fin at $0.99, but the double-counting quirk — the resolution consumes a ticket from your plan allowance as well — can erase the gap entirely at higher volumes. Always price the stack, not the sticker.
What counts as a “resolution” (and why vendors disagree)
This is the most important section in the article, because it decides whether the table above means anything at all. Vendors do not share a definition of a resolution, and the differences are worth real money:
- Intercom Fin ($0.99): bills per “outcome,” which covers a resolution, a workflow completion, or a handoff. A handoff that sends the customer to a human can therefore count as a billed outcome even though no AI resolution happened. Fin does not charge for failed resolutions in the sense of conversations it could not handle at all, but the handoff definition means you must read your own usage report, not just the invoice total.
- Zendesk ($1.50/$2.00): bills only for “verified resolutions,” where an AI evaluation model confirms a genuine resolution rather than a deflection. That verification is genuinely buyer-friendly: a chatbot that merely shows a help article and watches the customer give up does not bill. It is also why Zendesk can charge more per unit and still be competitive — fewer billed units for the same traffic.
- Gorgias ($0.90/$1.00): bills per AI resolution, and each one also counts against your ticket allowance. If your plan includes 300 tickets and the AI resolves 200 of them, you have 100 tickets of headroom left, not 300. At scale, that second meter is the real price.
- Salesforce Agentforce ($2.00): bills per conversation, including conversations that escalate without resolution. You pay for attempts, not outcomes. That is a different bargain, and it only makes sense if your conversations are high-value or if you are already deep in Salesforce Service Cloud.
- Freshdesk Freddy ($49 per 100 sessions): bills per session, and a session is a stretch of interaction, not a solved problem. A customer who chats three times about one order can generate more sessions than resolutions.
Before you sign anything, get the definition in writing from the vendor: what event triggers a bill, who verifies it, and how you dispute a billed unit that did not resolve anything. If a vendor cannot answer that plainly, that is your answer.
The volume math: 50, 500, and 5,000 resolutions a month
Strip out seats and plans for a moment and look at pure AI cost at three volumes. This is simple multiplication from list rates, ignoring included allowances:
| Tool (per-unit rate) | 50 resolutions | 500 resolutions | 5,000 resolutions |
|---|---|---|---|
| HubSpot Breeze ($0.50) | $25 | $250 | $2,500 |
| Help Scout ($0.75) | $37.50 | $375 | $3,750 |
| Gorgias annual ($0.90) | $45 | $450 | $4,500 |
| Intercom Fin ($0.99) | $49.50 | $495 | $4,950 |
| Gorgias monthly ($1.00) | $50 | $500 | $5,000 |
| Zendesk committed ($1.50) | $75 | $750 | $7,500 |
| Zendesk pay-as-you-go ($2.00) | $100 | $1,000 | $10,000 |
At 50 resolutions, the spread between the cheapest and priciest option is $75 a month — noticeable for a side hustle, but not decisive. At 5,000 resolutions, it is $7,500 a month, which is a hire-or-not decision. That scaling is the whole story of per-resolution pricing: it feels trivial at side-hustle volume and becomes your second-biggest support cost at growth volume. If your store is growing, negotiate or switch before you reach the volume where the per-unit gap hurts, not after.
Now add the seat and plan stack back in, because that is where small stores get surprised. A one-person store on Intercom pays seats from $29 a month plus $0.99 per outcome. A store with two support helpers on Zendesk pays Suite seats plus $1.50 per resolution plus, if it wants the agent-assist features, Copilot at $50 per agent — that is $100 a month in Copilot alone for two agents before a single resolution bills. A Freshdesk shop pays seats from $29 per agent plus $29 per agent for Copilot plus $49 per 100 Freddy sessions. The seat stack can easily exceed the AI line at low volumes, which flips the ranking: at 50 resolutions a month, the tool with the cheapest per-unit rate but the heaviest seat requirement can be the most expensive tool in the table.
The number that matters: effective cost per resolved ticket
Here is the formula to use, and it is the original value of this article — the calculation most comparison pages skip:
Effective cost per resolved ticket = (base plan + seat fees + AI bill) ÷ tickets actually resolved by AI
Note the denominator. It is tickets resolved, not tickets touched. If the AI touches 1,000 contacts and resolves 500, you divide by 500, not 1,000. The 500 it failed on still cost you human time, and that human time belongs in your thinking even if it does not appear on the AI invoice.
Worked example for a side store doing 500 contacts a month, where the AI resolves 60% (300 tickets) and 200 go to a human:
| Tool | AI bill (300 resolutions) | Seat/plan note | Effective AI cost per resolved ticket |
|---|---|---|---|
| HubSpot Breeze | $150 | Plus your HubSpot plan | $0.50, before plan cost |
| Help Scout | $225 | Plus seats from $25 each | $0.75, before seats |
| Gorgias annual | $270 | Plus ticket plan; resolutions consume tickets too | $0.90, before plan cost |
| Intercom Fin | $297 | Plus seats from $29 each | $0.99, before seats |
| Zendesk committed | $450 | Plus Suite seats and optional $50 per agent Copilot | $1.50, before seats |
Now change one variable: the resolution rate. If your ticket mix is mostly “where is my order,” returns, and address changes — the repetitive bucket that usually makes up 60–70% of a small store’s contacts — a well-trained agent can resolve the bulk of it. If your mix is mostly damaged-item claims, custom orders, and angry chargeback threats, the AI might resolve 30%, not 60%. At a 30% resolution rate on the same 500 contacts, only 150 tickets resolve, and the AI bill buys far less relief: Fin still bills $148.50, but you still handle 350 contacts yourself. The tool did not get more expensive per unit. Your ticket mix made it less valuable. Measure your mix before you buy, not after.
The bundle trap: when cheap per unit costs more
Bundle and credit pricing deserves its own warning, because it is where side-store owners lose money quietly:
- Tidio Lyro at $32.50 a month for 50 conversations works out to $0.65 per conversation if you use all 50. Use only 30, and your effective price is $32.50 ÷ 30 = $1.08 per conversation — more expensive than Fin’s $0.99, for a narrower tool. Use 50 every month without fail, and it is a bargain. The question is never the bundle price. It is your usage consistency.
- eesel AI at $299 a month for 500 credits is $0.60 per credit at full use. Use 300 credits, and it is $299 ÷ 300 = $1.00 per credit. A side store with seasonal volume — quiet in February, slammed in December — will waste credits in slow months and blow through them in peak months. Credit bundles punish uneven volume.
- Crisp’s $45 Mini plan with $5 of AI credits is too small an allowance to judge; the real cost is whatever overage or upgrade you hit in your second month. Treat any plan whose AI allowance covers less than a week of your volume as a trial, not a price.
- Freshdesk’s $49 per 100 sessions has the same trap in a different shape: sessions are not resolutions, so a store with chatty customers (three short chats per order) burns sessions faster than a store with one long chat per order, even at the same ticket count.
The rule is simple: if your monthly volume varies by more than about 30% between your quietest and busiest months, prefer pure per-resolution pricing over bundles and credits. You pay slightly more per unit in a good month and far less in wasted allowances across the year. If your volume is steady — same ticket count month after month — a bundle can win, but only if you track usage monthly and downgrade the moment you start leaving credits unused.
When per-resolution beats per-seat (and when it does not)
Per-resolution pricing wins in three situations. First, when you are a one-person side hustle: you pay for what the AI actually does, and you do not pay a seat for yourself twice. Second, when your volume is seasonal or spiky: the bill falls in quiet months instead of staying flat like a seat subscription. Third, when your resolution rate is high: every billed unit replaces human work, so the cost is easy to justify.
Per-seat or flat pricing wins in three different situations. First, when your volume is high and steady: a flat plan’s effective per-ticket cost falls as volume rises, while per-resolution cost rises linearly forever. At 5,000 resolutions, even a $500 flat plan beats a $0.50 per-resolution tool. Second, when your ticket mix is complex: if the AI resolves only 30% of contacts, you are paying per-resolution prices for a tool that mostly hands off, and a flat draft-and-assist tool (where AI drafts replies for your humans) may deliver more value per dollar. Third, when you need voice: phone support bills per conversation or per minute, not per resolution, and a $15–$25 human phone call is the cost you are actually replacing — a comparison this article’s chat-first table does not capture.
The decision rule, stated plainly: if your AI resolution rate is above 50% and your monthly volume is under about 1,000 contacts, per-resolution pricing is usually the best deal. If your volume is above 1,000 contacts a month or your resolution rate is below 40%, price out a flat or seat-based plan before you commit. Between those bands, run a two-week trial on both models and compare effective cost per resolved ticket, not list price.
What the vendors report (and what to discount)
Vendor-reported results are useful directionally and unreliable as promises. Gorgias reports that stores using conversational AI see about 45% fewer support tickets, and cites a jewelry brand whose live-chat volume fell 17% while on-site conversion rose 6%. Treat both figures as vendor-reported case data, not independent findings: the brand, the period, and the baseline are the vendor’s choice. What is independently useful is the pattern across vendors — repetitive “where is my order” traffic is the first thing AI clears, and clearing it is what frees a one-person store’s evenings.
The honest benchmark for your own store is not a vendor’s percentage. It is your own two-week trial: route half your chat traffic to the AI, keep half human-handled, and compare resolution rate and repeat-contact rate (customers who come back about the same order). A tool that resolves 70% of tickets but generates repeat contacts on a quarter of them is worse than one that resolves 50% cleanly. Watch repeat contacts above all — they are the hidden cost that never appears on an AI invoice.
A 30-day pilot plan for a side-hustle store
Do not buy an annual plan on day one. Here is a pilot that costs little and answers the only question that matters — what does this cost per ticket I actually stop handling:
- Week 1: count your tickets. Export last month’s contacts by topic. Tag the top five topics and their share of volume. If “where is my order,” returns, and shipping questions are not at least half your volume, AI support will help less than you hope.
- Week 2: pick two tools, not five. Choose the per-resolution tool that fits your existing helpdesk (Gorgias if you are on Shopify, Fin if you are already on Intercom, Zendesk AI if you are already on Zendesk) and one flat or bundle alternative for comparison. Switching helpdesks to chase a $0.20 per-unit gap is almost never worth the migration.
- Weeks 3–4: run the trial on real traffic. Give the AI your top five topics only, with a visible “talk to a person” escape on every chat. Log billed units, resolved tickets, and repeat contacts separately. Compute effective cost per resolved ticket with the formula above, including seats.
- Decide on three numbers, not one. Effective cost per resolved ticket, resolution rate on your top five topics, and repeat-contact rate. If all three beat your current human-only baseline, buy. If only the list price looks good, walk away.
Two guardrails for the whole pilot. First, never let a chatbot trap a customer in a loop: an angry customer who cannot reach a human costs more in refunds and chargebacks than any AI saving. Second, keep the AI away from refunds above your threshold, address changes on shipped orders, and anything involving a payment dispute until you have seen it handle the easy topics cleanly for a month. The AI earns harder jobs; it does not start with them.
The bottom line for side-hustle stores
AI customer service in 2026 is genuinely cheap per ticket — $0.50 to $1.50 per resolution across the major tools — and genuinely easy to overpay for. The overpayment never comes from the per-unit rate. It comes from seats stacked on resolutions, bundles bought for volume you do not use, resolutions counted twice, and conversations billed whether or not anything got solved. Price the whole stack at your actual volume, divide by tickets actually resolved, and compare that number across tools. Do that, and a side-hustle store can buy back its evenings for roughly the cost of a few hundred dollars a month at real side-hustle volume — and, just as importantly, will know exactly when growth makes it time to renegotiate.
Prices in this article are October 2026 list prices from vendor coverage and an October 24, 2026 multi-tool comparison. That comparison is a single-author source, so spot-check 2–3 prices against the vendors’ own pricing pages before purchase, and re-check bundle sizes and seat add-ons — those change most often. This article is general business information, not professional advice for your specific store.
Frequently Asked Questions
What is the cheapest AI customer service tool per ticket?
On list price alone, HubSpot Breeze at $0.50 per resolved conversation is cheapest, followed by Help Scout at $0.75 and Gorgias at $0.90 on annual billing. But the cheapest per-unit tool is not always the cheapest total: seat fees, required plans, and ticket double-counting can reverse the ranking at your volume, so compare effective cost per resolved ticket including the whole stack.
How is cost per resolved ticket different from cost per conversation?
Cost per conversation divides the bill by every contact the AI touched, including failures and handoffs. Cost per resolved ticket divides by contacts the AI actually solved. The second number is the one that matters, because a failed contact still costs you human time. Always ask a vendor which denominator its headline rate uses.
Why does Gorgias cost more than its $0.90 rate suggests?
Because each AI resolution also counts as a ticket against your plan allowance. AI resolutions consume that allowance alongside human tickets, so at higher volumes you hit plan limits — and overage — sooner than the per-resolution rate alone implies.
Is per-resolution pricing good for a seasonal side hustle?
Usually yes, compared with bundles and credits. Per-resolution billing falls in quiet months, while a $32.50 bundle for 50 conversations or a $299 block of 500 credits charges the same whether you use the allowance or not. If your volume swings more than about 30% between months, prefer pure per-resolution pricing.
Should a one-person store buy AI customer service at all?
If repetitive questions — order status, returns, shipping times — make up at least half your contacts, yes, on a monthly plan you can cancel. The value is not just money: it is after-hours coverage without hiring. If your contacts are mostly complex or emotional (damaged goods, custom orders, disputes), start with an AI drafting tool for your own replies instead of a fully automated agent.
What should I check before signing an annual AI support plan?
Get the billing trigger in writing (what counts as a resolution, who verifies it, how disputes work), confirm seat and add-on fees for every agent, check whether resolutions consume ticket or credit allowances, and run a two-week trial measuring resolution rate and repeat contacts. Never sign an annual deal on list price alone — and re-check the vendor’s pricing page, since bundle sizes and Copilot-style add-ons change frequently. Sources for the rates in this article include the Zipchat 2026 customer-service comparison and Rankite’s 2026 support-agent comparison, plus Intercom’s official Fin pricing for the $0.99-per-outcome rate, plus internal guides on this site: AI agent costs for US freelancers, Zelle vs Cash App vs Venmo fees for side hustles, and Etsy Share & Save holiday boost math.