AI chatbot pricing generally follows one of four models: a fee per seat, a fee per conversation, a fee per resolution, or a flat monthly bundle of replies. The model matters more than the headline price, because it determines whether cost rises in step with volume, with success, or not at all. Organizations should model total cost across a realistic range of volumes before committing.
This guide explains each model, shows how to estimate chatbot cost with worked figures, and identifies the charges that are commonly overlooked.
In brief: per-seat and flat pricing are predictable, while per-conversation and per-resolution pricing scale with usage and can become expensive at high volume. A total cost model should include the platform fee, volume-based charges, the staff time needed to maintain the system, and the savings from fewer human-handled conversations. The examples below use Bund AI's published flat plans and one clearly labeled hypothetical per-resolution rate.
What are the main AI chatbot pricing models?
Per-seat pricing charges for each human user of the platform, whether agents, administrators, or supervisors. It is common in help desk products to which AI has been added. Cost grows with team size and not with conversation volume, which suits organizations with stable staffing but can penalize those that add staff to handle peaks.
Per-conversation pricing charges for each conversation the bot handles, whether or not the customer's issue is resolved. It aligns cost with usage, but a rise in traffic, such as a seasonal peak, produces a proportional rise in the bill.
Per-resolution pricing charges only when the bot resolves a conversation, according to the vendor's definition of resolution. It is attractive in principle because the buyer pays for outcomes, but the definition carries the weight. If a conversation is counted as resolved when the customer simply stops replying, the buyer may pay for outcomes that did not occur.
Flat reply bundles charge a fixed monthly amount that includes a defined quantity of AI replies. Cost is known in advance, and the unit cost falls as utilization rises. The trade-off is that the buyer must estimate volume and select a bundle that fits.
Why do buyers distrust per-resolution pricing?
Per-resolution pricing deserves separate attention because it is common among established vendors and because buyer complaints about it are consistent. The concern is not the principle of paying for results but the difficulty of verifying that a result occurred.
Public feedback illustrates the point. One customer wrote in the Zendesk Community on 5 June 2025: "So we are being charged for autoresolutions on tickets that aren't actually resolved." Another reviewer on Trustpilot wrote on 31 July 2026: "At 1.50 USD per AI conversation I cannot justify switching it on." These are individual views and not evidence of a general pattern, but they indicate the questions a buyer should put to any vendor with this model: how is a resolution defined, what happens when a customer reopens an issue, and is there a cap on monthly charges.
Readers who want the current published structures for individual vendors can consult the guides to Intercom pricing, Zendesk pricing, and Gorgias pricing, which record plan details as verified in September 2026. Competitor prices are not repeated here because they change.
How is chatbot cost calculated for a flat plan?
A flat plan is straightforward to model once the unit is clear. For Bund AI, an AI reply is every message the agent sends, and a typical conversation involves two to three replies. A plan of 1,000 replies therefore covers roughly 333 to 500 conversations.
The published plans are as follows. Basic costs $0.99 per month and includes 50 AI replies. Starter costs $49 per month with 1,000 replies and three seats. Growth costs $99 per month with 5,000 replies and ten seats. Scale costs $199 per month with 10,000 replies and 25 seats. Enterprise is priced individually and includes unlimited replies and seats. There are no per-resolution fees. The current details are on the pricing page.
The effective cost per conversation can be calculated by dividing the plan price by the number of conversations it covers, assuming full use. On Starter, $49 across 333 to 500 conversations is about $0.10 to $0.15 per conversation. On Growth, $99 across roughly 1,667 to 2,500 conversations is about $0.04 to $0.06. On Scale, $199 across roughly 3,333 to 5,000 conversations is about $0.04 to $0.06. These figures assume the plan is fully used, and actual cost per conversation rises if utilization is lower.
What does a worked example look like for a small business?
Consider a hypothetical online retailer that receives 1,200 customer conversations a month, of which the agent handles most. At two to three replies per conversation, the volume is 2,400 to 3,600 replies, which falls within the Growth plan at $99 per month, with room for a seasonal increase.
For comparison, the following is an illustration only and does not describe any vendor's actual rate. Suppose a hypothetical provider charged $1.00 per resolution. If the agent resolved 900 of those 1,200 conversations, the monthly charge would be $900, before any platform fee. At that rate, cost would rise by $100 for each additional 100 resolutions, so a peak month with 2,000 conversations and 1,500 resolutions would cost $1,500.
The point of the comparison is structural and not a claim about any specific product. Under a flat bundle the cost is fixed within the plan, so management can budget it. Under a per-resolution model the cost is variable, which may be acceptable for organizations with low volume but becomes significant as automation succeeds. The AI support cost calculator allows organizations to test these scenarios with their own numbers.
What other costs belong in a total cost model?
A complete model extends beyond the subscription. Implementation time is the first item: preparing knowledge sources, configuring actions, and testing handoff. Bund AI accepts crawled web pages, uploaded PDF, DOCX and Markdown files, pasted text, and help articles, so much of this effort is editorial and not technical.
Ongoing maintenance is the second. Knowledge must be kept current, and transcripts reviewed, particularly in the early months. Integration effort is the third: connecting order systems or other tools may require custom HTTP actions, which should be estimated in staff hours.
Offsetting these costs are the savings from conversations that no longer require a person. The cost per ticket calculator provides a way to value this, and the broader economics are discussed in our articles on the ROI of customer support and on how to reduce customer support costs. A cautious approach uses a conservative resolution rate in the model, since results vary by organization. The 86.7% no-touch resolution figure reported for Bund AI comes from early rollouts and should not be assumed in planning.
How should an organization compare quotes?
Comparison is easiest when every quote is converted to the same unit. A common unit is the cost per resolved conversation at three volumes: current, current plus 50 percent, and a peak month. Ask each vendor how a billable event is defined, whether failed or abandoned conversations are charged, how seat requirements will change as the team grows, and what additional modules are required for the features the organization needs.
Seats deserve separate attention. A plan that includes ten seats at a flat price differs materially from one that charges for each agent, especially for organizations with supervisors, part-time staff, or seasonal teams. Our Intercom alternative and Zendesk alternative pages describe the differences in approach, and the wider market is compared in the best AI customer support software roundup.
The AI customer support guide and the support automation overview give the wider context. For an overview of the deployment considerations that sit alongside cost, see the guide to the AI customer service agent, and for the range of automation categories, customer support automation software.
When Bund AI is not the right fit
Flat plans suit organizations whose volume can be estimated and fits within a bundle. A business with highly irregular or very low volume may find that a smaller plan is sufficient, or that a pay-per-use model costs less in a quiet month. Organizations that need telephone, SMS, or WhatsApp support will need a different product, as Bund AI operates through a web widget and email inbox. It is hosted and closed source, and it integrates with commerce platforms through custom actions and not a native app.
The figures in this article are for planning. Confirm current plan limits on the pricing page before purchase, and run a pilot against your own conversations before relying on any estimate.
Frequently asked questions
How much does AI chatbot pricing typically cost? Costs range from tens of dollars per month for flat plans to several hundred or thousands for per-resolution or per-conversation models at volume. The structure matters as much as the headline figure, so model your own volume under each option.
What is the difference between per-conversation and per-resolution pricing? Per-conversation pricing charges for each conversation the bot handles, while per-resolution pricing charges only when the vendor counts a conversation as resolved. The second depends heavily on how resolution is defined and verified.
How do I estimate chatbot cost for my business? Estimate monthly conversations, assume a realistic share the bot will handle, and convert to the pricing unit of each vendor. For reply-based plans, assume two to three replies per conversation and compare the result with plan limits.
What does Bund AI cost? Bund AI offers Basic at $0.99 per month for 50 AI replies, Starter at $49 for 1,000 replies, Growth at $99 for 5,000 replies, and Scale at $199 for 10,000 replies. Enterprise is custom with unlimited replies and seats, and there are no per-resolution fees.
Are there hidden costs in chatbot pricing? Common additional costs include extra seats, add-on modules, implementation and maintenance time, and integration work. Ask each vendor for a written definition of what is billable and test it against a month of your own traffic.
Is a flat-rate AI chatbot better than a per-resolution one? Flat rates offer predictability and tend to be cheaper as volume grows, while per-resolution models can cost less at very low volume. The better choice depends on volume, seasonality, and how much you trust the vendor's definition of resolution.