Cutting Support Costs: At What Ticket Volume Does an AI Chatbot Pay Off?

“A chatbot pays off from 500 tickets a month.” You’ll find this rule of thumb in plenty of guides. It isn’t wrong, but it isn’t right either. Whether an AI chatbot pays off doesn’t depend on a magic ticket count. It depends on four variables that look different in every company: what a ticket costs today, how many tickets the bot actually resolves, how much maintenance it needs, and how the vendor bills.

That last point changed fundamentally in 2026. Vendors such as Fin, Zendesk and HubSpot no longer bill only per seat but per resolved ticket. That shifts the break-even point, in some cases significantly lower. At the same time, it creates a second tipping point above which per-resolution billing becomes more expensive than a flat fee.

This article derives the threshold step by step, compares the three pricing models, and provides a calculator to work out your own number.

What does a support ticket really cost?

Cost per ticket is the single most important number in the whole calculation, and also the one most often guessed. Figures from “€2” to “€25” circulate online, usually without a source. So we derive it ourselves, so you can recalculate every step with your own values.

Step 1: What does an hour of support cost?

• Salary: A customer support employee in Germany earns a median of around €33,100 gross per year (Stepstone salary report 2026).
• Non-wage labor costs: According to the German Federal Statistical Office (Destatis), every €100 of gross pay comes with around €29 in non-wage costs. That puts employer costs at around €42,700 per year.
• Productive hours: After vacation, sick leave, meetings and training, experience shows around 1,500 productive hours per year remain. That’s about €28.50 per productive hour.
• Overhead: We add a flat 30% for workspace, ticketing system, telephony and team lead. Result: around €37 per productive hour.

The values for productive hours and overhead are assumptions. For comparison: according to Destatis, an hour worked in Germany cost €45 on average across all industries in 2025.

Step 2: From hour to ticket

These figures are broadly in line with international surveys. The UK benchmark by ContactBabel (2024) reports an average of £3.05 per chat, £3.55 per email and £5.58 per call, with an average call duration of just over seven minutes. Gartner puts live channels overall at around $8 per contact and self-service at around $0.10.

The break-even formula: at what ticket volume does a chatbot pay off?

A chatbot pays off as soon as the working time it frees up is worth more than it costs. You need five variables:

• N = tickets per month
• r = actual resolution rate, i.e. the share of tickets resolved without a human
• c = cost per ticket when handled by a human
• Fixed costs = everything that is due each month regardless of ticket volume: license or flat fee, knowledge base maintenance, amortized setup costs
• p = cost per ticket caused by the chatbot itself (with per-resolution billing, the price per resolution)

The freed-up working time per month is r × N × c. The break-even point is reached when this saving covers the costs:

Break-even ticket volume = fixed costs ÷ (r × (c − p))

Three things stand out immediately:

1. The resolution rate matters more than the list price. Halve the resolution rate and the required ticket volume doubles.
2. Maintenance is a fixed cost. Ten hours of maintenance a month at €37 per hour cost €370, often more than the license itself.
3. If the price per resolution reaches the cost per ticket, the chatbot never pays off. With very simple requests at €2 and a price of $2 per resolution, that happens quickly.

Three pricing models and how they shift the threshold

Model 1: Per-resolution billing

With outcome-based pricing, you only pay when the chatbot actually resolves an issue. The main vendors at a glance:

Advantage: Hardly any fixed costs, so the threshold is low. At €5 per ticket, a 50% resolution rate and €0.85 per resolution, the chatbot pays off from about 320 tickets a month.

Disadvantage: Costs rise with every resolved ticket. The more successful the chatbot, the higher the bill. And the bill is hard to predict.

Model 2: Flat fee or quota

Vendors such as moinAI charge a fixed monthly fee, for example €750 per month on the Business plan with annual billing, with unlimited conversations and servers in Germany. Others sell quotas, such as Lime Connect (formerly Userlike) with an AI agent package for €99 per month including 100 conversations, or Tidio with Lyro from $32.50 for 50 conversations.

Advantage: Predictable costs, and every additional resolved ticket is free.

Disadvantage: Higher fixed costs, so a higher threshold. With the same assumptions as above, a €750 flat fee pays off from about 570 tickets a month.

Model 3: Self-built

A self-built chatbot, for example with n8n and a language model via an API, costs very little per conversation. According to Anthropic, a typical support conversation with Claude Haiku 4.5 costs less than half a cent. With a more capable model and access to a knowledge base, it’s more like 5 cents. In return, you carry development, hosting and maintenance yourself or buy them in.

Advantage: Full control over data and logic, lowest cost per ticket at high volume.

Disadvantage: The highest fixed costs. Self-building usually only pays off at several thousand tickets a month, or when data protection or deep system integration leave no other choice. Our article on building your own AI covers the options.

The threshold at a glance

The table shows the monthly ticket volume from which a chatbot pays off. Assumptions: around €500 of maintenance per month (just over 13 hours at €37 per hour), €2,000 setup spread over 12 months, €0.85 per resolution or a €750 flat fee.

The second tipping point: when the flat fee becomes cheaper again

With per-resolution billing, the bill grows with every resolved ticket; with a flat fee, it doesn’t. Both models cost the same when:

Tipping point = flat fee ÷ (resolution rate × price per resolution)

With a €750 flat fee, a 50% resolution rate and €0.85 per resolution, this point is around 1,760 tickets a month. Below it, per-resolution billing is cheaper; above it, the flat fee. If you’re growing fast, look not just at today’s volume when signing a contract but at where you expect to be in a year.

Calculate your own threshold

Set your ticket volume, handling time and chatbot assumptions. The calculator shows, for all three pricing models, the volume from which each pays off and what your monthly balance looks like.

What counts as “resolved”? The pitfalls of per-resolution billing

With outcome-based models, the definition of a “resolution” determines your bill. And it’s different for every vendor.

Fin bills a resolution when the customer confirms it or doesn’t follow up after the bot’s answer. If the customer stays inactive for 24 hours, the issue counts as an “assumed resolution”. If they return to the same conversation, the resolution is deducted again. Since March 2026, Fin has also billed in “outcomes”, which include not only resolutions but certain handoffs to humans when they come at the end of a configured procedure.

Zendesk switched to tiered resolutions in May 2026. Only “verified resolutions”, checked by a second language model, are billed; escalations and partial resolutions are free. Zendesk does not publish an official price for this.

Freshworks, by contrast, bills per session, i.e. per interaction, whether resolved or not. The first 500 sessions are included; after that, 100 sessions cost $49.

Before signing, get it in writing: what exactly is billed? How long is the time window? Are repeat requests through other channels detected? And is there a monthly cost cap?

Realistic resolution rates: what vendors promise and what you get

The resolution rate is the biggest lever in the calculation, and the number where marketing and reality are furthest apart.

• Vendor figures: In June 2026, Fin reported an average resolution rate of 76% across more than 12,000 customers. Other vendors advertise “up to 80%”. These figures usually include assumed resolutions.
• Independent field tests report significantly lower values, often between 40 and 50%, and below 30% with a thin knowledge base.
• The launch phase: In the first three months, rates are almost always below their later level because the knowledge base still has gaps.

Our recommendation: base your decision on 30 to 50%, not on the figures in the vendor’s pitch deck. If the chatbot does better later, that’s a bonus.

The best-known example shows how much quality matters: in 2024, Klarna reported that its AI assistant handled two thirds of all chats in its first month, the work of 700 full-time agents. In May 2025, CEO Sebastian Siemiatkowski admitted the company had focused too heavily on cost and quality had suffered. Klarna has since been hiring people for customer service again.

Which tickets are suitable for a chatbot

It’s not the total number of tickets that matters, but the number of similar, recurring tickets. A company with 2,000 tickets a month that are all individual benefits less than one with 400 tickets, 300 of which ask the same five questions.

In e-commerce, delivery status requests typically make up a large share of all tickets. How to automate returns in detail is covered in our guide on automating return requests.

Three worked examples

The following examples use the calculator’s default values. All figures are rounded.

Small team: 200 tickets a month

A trades business with two people in the office handles around 200 requests a month, mostly scheduling questions and status updates at €5 per ticket. At a 50% resolution rate, 4 hours of maintenance and €1,000 setup:

• Freed-up working time: around €500 a month
• Per-resolution billing: costs around €316, balance around +€180
• €750 flat fee: balance around −€490

Result: Only the per-resolution model narrowly pays off. Often a good FAQ page with online booking is the more economical first step here.

Online shop: 3,000 tickets a month

An online shop receives 3,000 requests a month, mainly about delivery status and returns, at €4 per ticket. At a 55% resolution rate, 10 hours of maintenance and €3,000 setup:

• Freed-up working time: around €6,600 a month
• Per-resolution billing: balance around +€4,580
• €750 flat fee: balance around +€5,230
• Self-built at €900 operations: balance around +€4,930

Result: All models pay off clearly. The flat fee is cheapest because the volume is above the tipping point.

B2B service provider: 800 tickets a month

A software company receives 800 support requests a month, which are more complex on average (around €9 per ticket) but only 35% automatable. With 15 hours of maintenance and €3,000 setup:

• Freed-up working time: around €2,520 a month
• Per-resolution billing: balance around +€1,480
• €750 flat fee: balance around +€970

Result: Despite the lower resolution rate, the chatbot pays off because every resolved ticket is expensive.

Hidden costs that aren’t on any price sheet

1. Knowledge base maintenance: A chatbot is only as good as the information it draws on. New products, changed prices and new questions have to be added. Vendors often cite 40 to 80 hours for the initial build and 5 to 15 hours a month for ongoing maintenance. Independent measurements are scarce.
2. Quality control: Someone has to regularly review conversation logs, correct wrong answers and analyze escalations.
3. Cost of wrong answers: Incorrect information can cost a follow-up ticket, a complaint or a lost customer. According to Gartner (September 2026), only 27% of customers would use a chatbot again after a bad experience.
4. Setup and integration: Connecting to your ticketing system, shop, ERP or calendar. HubSpot, for example, charges a one-time onboarding fee of $1,500 or $3,500 depending on the plan.
5. Minimum commitments and contract terms: Monthly minimums or terms of 6 to 12 months push the threshold up when volume fluctuates.

When an AI chatbot doesn’t pay off

• Too few similar requests: If less than half of your tickets fall into the “very good” or “good” categories, the resolution rate stays too low.
• Very cheap tickets: If a ticket costs less than €2, the price per resolution eats up much of the saving.
• No owner: Without someone who maintains the knowledge base and reviews reports, every chatbot gets worse over time.
• Customers who expect a human: According to Bitkom, 62% of German online shoppers want to reach a person quickly in customer service. Only 50% are satisfied with chatbots, compared with almost nine in ten with a human agent. According to Gartner, 87% of customers say access to a human is a must when AI is used in service. A chatbot without an easy handoff to your team costs customer loyalty.

In these cases, a revised FAQ page, canned responses in your ticketing system, or an AI assistant that drafts replies for your team are often the better first step. The different stages of automation are covered in our guide to automating customer support with AI.

Legal: what to consider when deploying a chatbot in the EU

Transparency obligation under the EU AI Act: Since August 2, 2026, AI systems that interact directly with people must be designed so that users are informed they are dealing with an AI (Art. 50(1)). The Digital Omnibus did not postpone this obligation. The notice must be given at the latest at the first interaction, for example with a clear sentence at the start of the chat. Violations can be fined up to €15 million or 3% of worldwide turnover; for SMEs, the lower amount applies. More in our EU AI Act compliance checklist.

Data protection: You need a data processing agreement with the vendor (Art. 28 GDPR) and a privacy notice in the chat. For vendors based in or with servers in the US, the rules on international data transfers also apply.

Works council (Germany): If the chatbot also evaluates data about your employees, for example who handled which escalation, the works council must be involved under Section 87(1) No. 6 of the German Works Constitution Act.

Checklist: ticket analysis before deciding

Frequently asked questions

Conclusion: the threshold is a calculation, not a rule of thumb

The blanket answer “from 500 tickets” is a useful guideline, nothing more. Your actual threshold could be 200 or 1,500 tickets, depending on what your tickets cost, how many of them can really be automated, and how your vendor bills. The new per-resolution models make AI chatbots interesting for smaller volumes, but they demand a close look at what counts as “resolved”.

The most important recommendation: analyze a sample of your tickets before talking to vendors. With a realistic resolution rate, your real cost per ticket and honestly planned maintenance, you’ll make a decision that still holds up a year from now.