AI chatbot prices range from small packages below US$50 a month to platforms costing hundreds or a few thousand euros a month. What you pay depends on the capabilities your business needs, the volume of enquiries and how usage is billed.
You pay for software, AI usage and, where applicable, setup. Against those costs sit less manual service work, faster help for customers and potential additional sales. A useful budget considers both sides.
A pharmacy receiving, for example, 50 monthly questions about opening hours, its address and collection could manage with a simple FAQ chatbot. For a dealership group handling several thousand enquiries about vehicles, locations and appointments, the plan needs to cover more conversations and may need to include integrations. A larger online shop might also need order lookups and returns. Those requirements determine the right package; company size alone does not tell you which features are needed.
How much does an AI chatbot cost per month?
Published pricing shows how differently these services are packaged:
- Small usage packages: One entry-level option starts at around US$33 per month, billed annually. This example includes 50 conversations shared between AI and your team. More enquiries require more usage allowance. Pricing source
- Broader platform packages: Another published price list shows plans with starting prices from €750 to €2,500 per month with annual payment. They include unlimited conversations but differ in channels, integrations and actions the agent can perform. Pricing source
- Usage-based pricing: Some offers start at US$0.99 per billable outcome. A monthly minimum or additional software fees may apply. Pricing source
Prices checked on 19 November 2026. These figures illustrate different offers, rather than market averages. Original currencies and payment terms are retained.
The wide range has a reason: “AI chatbot” covers very different services. A low starting price with limited usage tells you little about the cost of handling several thousand enquiries. Equally, a higher fixed fee can make sense at high volumes.
Why does the scope make such a difference?
Specific customer questions make the differences clearer. What information does the agent need, and should it only provide an answer or also complete a task?
Task · Example · What the chatbot needs
| Task |
Example |
What the chatbot needs |
| Answer questions from existing content |
Example“How long do I have to return something?” |
What the chatbot needsAn up-to-date PDF, help centre content or imported website pages. |
| Retrieve order and shipping status |
Example“Where is my parcel?” |
What the chatbot needsAccess to the correct order and current shipping data, with a suitable way to match the order to the customer. |
| Recommend products in a conversation |
Example“Does this jacket meet my requirements, and is it available in medium?” |
What the chatbot needsPreferences already shared, product information and current shop data for availability. |
| Perform an action |
Example“Can you still change my delivery address?” |
What the chatbot needsPermission to make changes, an order-status check, rules for allowed changes and a reliable confirmation. |
| Support several shops or brands |
Example“Which returns policy applies to my order?” |
What the chatbot needsThe correct shop system, country and brand policies, and handover to the relevant team when needed. |
For an FAQ chatbot, uploading a document or importing selected website pages may be enough. It does not require a custom shop integration. A “Where is my order?” enquiry, however, needs current information about a specific order.
Registering a return, rescheduling an appointment or updating customer details adds another layer. Permissions, business rules and failed actions need to be handled. For example, a refund must not be treated as paid simply because the agent has written a confirmation.
Conversation context alone does not automatically mean an expensive custom solution. Remembering a size mentioned earlier is different from combining order history, availability and account information from several systems. An existing standard integration can substantially reduce the work involved.
Our guide to AI customer service for ecommerce explains these tasks and human handover in more detail.
How is AI usage billed?
Alongside the scope, billing determines how your costs change with usage.
With per-conversation pricing, the provider counts chats according to its definition. Several messages may belong to one conversation. With included usage, what matters is what happens once you reach the allowance: extra charges, a plan change or a usage limit.
With outcome-based pricing, the definition matters. For one provider, a completed workflow that hands over to a person can count as a billable outcome. Cases where no further help is requested after an answer may also count. That matters for the invoice, but it does not always mean the enquiry was resolved without human involvement. Billing definition
Outcome-based or value-based: what is the difference?
Outcome-based pricing charges for a defined result, such as a resolved enquiry or a qualified lead. Value-based pricing aligns the price with the economic benefit to the business. The two can overlap: a qualified prospect for a large order may be worth more than a general answer about delivery times. Some services already charge different amounts for different outcomes. Example of different outcome prices
An agreed share of attributable revenue is another possible arrangement. For it to work for both parties, the contract needs to define the contribution being rewarded and how returns or cancellations are handled. A high order value alone does not establish the additional value the agent created. A well-defined model can align the provider's and the business's interests more closely with results.
A useful comparison therefore needs a price at your expected volume and a clear explanation of what triggers a charge. During seasonal peaks, this can matter more than a small discount on the base fee.
What does setup cost?
Some offers include setup; others charge separately for onboarding or custom integrations. Standard plans without an additional setup fee are also available. Example price list
The work depends above all on what the chatbot needs to do. Some services let you enter a website address and have the agent import the available content. That can be enough for simple questions about opening hours or delivery policies.
In our experience, these solutions soon reach their limits when customers expect individual information or specific actions. A website may state the usual delivery time, but it does not tell the agent where a particular customer's parcel is. That needs current order data. Registering a return requires the appropriate integration and rules for what is allowed.
Current FAQs and existing standard integrations make it easier to get started. Conflicting content, multiple shops or custom connections add work. That is why a single assumed project fee would be misleading.
You do not need to write every possible customer question and answer in advance. The agent can use existing content. Setup still includes choosing reliable sources, setting the tone of voice, testing and making sure handover to your team works. Actions such as returns or address changes also need the relevant rules and permissions.
Which ongoing fees should you budget for?
Alongside software and AI usage, you may be charged for additional team seats, extra features or maintenance of custom integrations. Whether these apply depends on the offer.
If you build your own chatbot, hosting, model usage and software maintenance belong in the budget too. The language model's price alone is therefore not a complete comparison with a ready-made platform.
A WhatsApp chatbot can also incur messaging charges, which depend in part on recipient market and message category. Our guide to WhatsApp Business API costs explains how the channel is billed. The official pricing overview sets out the current terms.
Example: what do 1,500 AI sessions a month cost?
A price per session is different from a price per resolved enquiry. A session may be billable even if your team takes over afterwards. For this example, we therefore use a specific session-based tariff rather than a supposed average price.
The published price list lists €49 per team seat per month with annual billing and €45 per 100 AI sessions. A chat session covers interactions within 24 hours of its start.
We assume three team seats and 1,500 sessions per month. Those are choices for our example business, not charles prices or an average enquiry volume.
Cost · Calculation · Amount
| Cost |
Calculation |
Amount |
| Software and team seats |
Calculation3 × €49 |
Amount€147 per month |
| AI usage |
Calculation15 packs of 100 sessions × €45 |
Amount€675 per month |
| Recurring fees |
CalculationSoftware plus usage |
Amount€822 per month |
| Setup and custom integrations |
CalculationDepends on the scope required |
AmountVariable |
At constant usage: €9,864 per year plus variable setup costs. Amounts exclude VAT. Additional channel charges, custom integrations and one-off trial allowances are excluded. The software is billed annually under this tariff; its monthly equivalent is shown for comparison.
Our PURELEI success story shows that several thousand monthly enquiries can be a real operating volume: its agent resolves more than 2,000 FAQ enquiries a month without human handling. That is an individual customer example, not an industry average or evidence for the costs in this calculation.
When is an AI chatbot worth the cost?
An AI chatbot can make a substantial difference when your team handles many recurring enquiries. What matters is how many it resolves fully without human involvement. This is its automated resolution rate.
Suppose an agent resolves 60 to 70 per cent of the straightforward enquiries it receives on its own. Your team no longer needs to handle that share of those enquiries. These could include common questions about delivery policies or “Where is my order?” when the agent can retrieve the current order status. The percentage is an illustrative assumption; the achievable resolution rate depends on the enquiries and the agent's capabilities.
As more of these enquiries come in, the agent can take more manual work off your team's hands. At a similar resolution rate, a higher enquiry volume means a greater reduction in workload. Your team has more time for personal advice and cases that need human support. At high volumes, the time saved can substantially outweigh the effort of keeping information current and checking quality. The financial benefit also depends on recurring fees.
Customers benefit too. For enquiries the agent can resolve itself, they get immediate help without waiting for an available team member or the next service day. This can improve the customer experience and make buying decisions easier. When human support is needed, passing on the conversation context helps the team take over.
Looking ahead: when customers send their own AI agents
Personal agents such as Meta Muse and dots in ChatGPT already perform tasks on their users' behalf. That makes the question of how businesses serve these agents more relevant.
Consider a possible example: a customer asks their agent to arrange a return. With a suitable integration, the business's agent could check authorisation and return the reference number, shipping label and status directly. This could reduce follow-up questions and unnecessary messages between the two agents.
Agent detection (recognising a personal agent) would only be part of it. Who the agent represents and what it is allowed to do also matter. Simply claiming to be the customer's agent is not enough. Approaches such as the Personal Agent Protocol therefore provide for controlled access to business systems.
We see this as a development that could become relevant when choosing a provider. Its effect on your enquiry volume or costs remains uncertain. For today's budget, start with the functions actually supported and the volume you receive.
Help customers and encourage their next purchase with charles
A provider that combines customer service and marketing can offer value beyond answering an individual question. You can help customers and reach them again later when an offer matches their interests.
For example, a customer gets help with an order and then chooses to receive your brand's WhatsApp offers through the chat. Later, you can tell them about a relevant accessory or a new product. An unresolved complaint or an item they have already bought should be taken into account.
A support conversation alone is not a marketing opt-in. Consent must cover the subsequent marketing messages; proactive WhatsApp messages also need to follow the rules on approved templates. WhatsApp policy
With charles, we combine AI agents for customer enquiries with campaigns and relevant audience segments. Alongside service efficiency, that creates an opportunity for further purchases. The integrations and marketing features included belong in the same pricing comparison.
In a demo with our team, we look at your service enquiries, systems and marketing goals together. From there, we develop a useful scope and a clear budget. You do not need a finished business case to get started.