The best fit among ai chatbots for customer service depends on how the tool charges: pick per-resolution pricing when ticket volume is unpredictable, bundled seat pricing for a steady bill, and keep a human agent ready for complex cases.
This list sits within technology for business innovation. It is ordered by how each tool charges, not ranked by quality, and no vendor performance claims are used to place anything.
What AI chatbots handle well
Basic chatbots are rule-based systems that follow scripts and predefined flows. They suit narrow questions whose answers rarely change. A scripted bot has no way to recover when a customer words a question differently from the flow it was built for.
AI text agents work differently. The same source describes them as using natural language processing (NLP) and machine learning to comprehend, respond to and learn from conversations. Conversational agents built on NLP and deep learning go further and understand context, intent and even sentiment. Large language model assistants belong to this second group. They suit customers who describe a problem in their own words, and the learning side of this technology is covered in machine learning for business.
Availability is the other clear strength. Human agents cannot be online 24/7, while a chatbot can answer outside staffed hours.
When a bot resolves a question without a person stepping in, the ticket never reaches the queue in the help desk software. That outcome is called ticket deflection, and it is the unit most vendors now bill against.
Chatbot tools and their public pricing
The market prices an AI-resolved ticket anywhere from $0.49 to $2.00, according to Drag’s July 2026 review of AI support pricing. The items below are grouped from pay-per-outcome to bundled pricing. Check each vendor’s own pricing page before buying, because rates change.
Fin: the per-resolution reference price
Fin’s price of $0.99 per resolution, with a 50-per-month minimum, is the figure the market keeps quoting each other, per the Drag review. It suits teams that want a benchmark to test other quotes against. The minimum makes it a weaker match for very low-volume support.
Featurebase: low per-resolution rate on top of a plan
Featurebase charges $0.49 per AI resolution, billed on top of the plan price. It suits small teams that already need the plan. The plan fee is a fixed cost, so the low per-resolution rate is not the whole bill.
Kayako: pay only for resolved tickets
Kayako states that you only pay if tickets are resolved, at $1 per resolved ticket plus per-seat pricing by tier. It also integrates with Zendesk, Freshdesk and Salesforce. That suits teams that want to keep an existing help desk. The seat fees remain, so the outcome-based charge does not remove fixed costs.
Zendesk: highest metered rate, with a seat add-on
Zendesk charges $1.50 to $2.00 per automated resolution, plus Copilot at $50 per agent per month. Kayako’s guide notes that on the Zendesk bot agents can step in anytime for complex cases. It suits teams already running Zendesk as their help desk. It is the most expensive per resolution in this list.
Freshdesk: per-session billing
Freshdesk bills about $0.49 per Freddy session, plus Copilot at $29 per agent per month. A session is not the same unit as a resolution, so a session that ends without a solution may still be charged. It suits existing Freshdesk users who want a modest seat add-on.
Trengo: per-conversation surcharge
Trengo adds €0.25 (annual) to €0.30 (monthly) per AI-handled conversation on top of the plan bundle. It suits teams that prefer paying per conversation handled. The charge applies whether or not the conversation ends in a solution.
Crisp: monthly AI credits
Crisp sells AI through credits, and unused credits expire monthly. That suits steady, predictable volumes. A quiet month wastes the credits, so it is a poor fit for seasonal demand.
Drag: AI bundled into the seat
Drag, the publisher of the pricing review, lists its AI as included in the seat from $18, with monthly AI credits. That gives the most predictable bill in this list. It suits small teams that want one price per agent, and the monthly credits cap how much AI work the seat covers.
Intercom: AI bots gated by plan
Kayako’s guide says Intercom’s AI bots are only available on Grow and above plans, at higher cost. It suits teams already planning to run on those plans. Smaller teams need to price the plan upgrade before counting the bot as an add-on.
Pricing at a glance
| Tool | Pricing basis | Reported figure | Planning use |
|---|---|---|---|
| Fin | Per resolution | $0.99, 50-per-month minimum | Benchmark for other quotes |
| Featurebase | Per resolution, on top of plan | $0.49 per AI resolution | Add the plan fee to the usage cost |
| Kayako | Per resolved ticket plus seats | $1 per resolved ticket | Cost follows outcomes; seats stay fixed |
| Trengo | Per AI-handled conversation | €0.25 (annual) to €0.30 (monthly) | Charged on handling, not on outcome |
| Crisp | Monthly credits | Unused credits expire monthly | Fits steady volume only |
| Drag | Bundled in seat | Seat from $18, with monthly AI credits | Most predictable bill |
| Intercom | Plan-gated | AI bots on Grow and above plans | Price the plan, not just the bot |
Figures are as reported in Drag’s AI support pricing review and Kayako’s chatbot software guide.
What a business chatbot costs to set up and run
Running costs depend mainly on which of five pricing models a vendor uses. Drag’s review lists flat-included, per-resolution, per-conversation, hybrid seat-plus-meter and expiring credits as models that now coexist.
- Flat-included: the AI comes with the seat, as with Drag. This gives a steady bill and suits teams that dislike variable invoices.
- Per-resolution: the bill follows solved questions. It suits unpredictable volume, but the per-unit rate is higher than a bundled seat would imply.
- Per-conversation: the bill follows conversations handled, as with Trengo. It rewards activity rather than results, so check what happens when the bot cannot answer.
- Hybrid seat-plus-meter: seats and usage both count, as with Zendesk and Kayako. Add the seat fees to the metered charges before comparing totals.
- Expiring credits: a monthly allowance, as with Crisp. Size the allowance to a normal month rather than a peak one.
The rates cannot be compared directly. The review points to a four-fold spread in per-resolution rates, which signals that the unit is not standardized. Ask each vendor what counts as a resolution before comparing totals.
Chatbot vs live agent
A chatbot and a live agent cover different parts of the same queue. The comparison below follows what the evidence supports.
- Coverage: a chatbot can be available around the clock, which human agents cannot match.
- Complexity: scripted bots stop at the edge of their predefined flows. Conversational agents read context, intent and sentiment, but complex cases still call for a person.
- Cost shape: bots are billed per outcome, session, conversation or credit, while agents are billed per seat. A mixed team pays both.
The handoff rule
Editorial judgement based on these capabilities: let the bot keep a conversation only while it is resolving it. Transfer to a human agent when the question falls outside what the bot can answer, when the case is complex, or when the customer’s sentiment turns negative.
Choose tools that let agents step in at any point, as Kayako’s guide describes for the Zendesk bot. Where billing is per resolution, a failed bot answer is not charged, but the customer still needs a person. Where billing is per conversation or session, a failed answer can still cost money. In both cases, a quick handoff protects the customer’s experience.
Choosing a tool
Work through these points in order before comparing prices.
- Fix the pricing unit. Ask whether the vendor bills per resolution, session, conversation, seat or credit, and what each one means.
- Check help desk fit. Kayako lists integrations with Zendesk, Freshdesk and Salesforce, so a bot can sit alongside an existing desk.
- Read the plan gates. Intercom’s AI bots start at the Grow plan, so the real cost is the plan.
- Match volume to billing. Unpredictable demand favours per-resolution billing, and steady demand suits credits or a bundled seat.
- Test the handoff. Confirm that agents can take over a live chat without the customer repeating the problem.
Frequently asked questions
What is ticket deflection?
Ticket deflection means a customer’s question is resolved by the bot, so no ticket reaches a human agent. Several vendors bill on this outcome.
Can a chatbot work with the help desk software a team already uses?
Yes, in some cases. Kayako’s AI integrates with Zendesk, Freshdesk and Salesforce, so it can run alongside an existing help desk rather than replacing it.
Is a chatbot the same as live chat?
No. Live chat is a real-time conversation with a person, while a chatbot answers automatically. The two can share a channel, and the handoff rule decides when the conversation moves from bot to agent.
Is a scripted bot still worth using?
It can be, for narrow questions with fixed answers. Its limit is that it follows predefined flows, so anything outside them needs a handoff or a more conversational agent.
