How much does an AI chatbot cost for a business in 2026?

"How much does a chatbot cost?" is a bit like asking how much a car costs. A simple button menu on your website and an AI assistant that answers in two languages on WhatsApp, books appointments and updates your CRM are very different projects. Both get called "chatbots".

This guide breaks the cost down the way a business owner actually pays it: a one-time setup, a monthly platform or service fee, AI usage, and messaging fees if you use WhatsApp. Prices are in US dollars and are the ones we could verify in late October 2026. Vendors change prices often, so treat them as a guide, not a quote.

The three main types of chatbot

1. Rule-based chatbots

These follow a fixed script: buttons, menus and keyword triggers. "Press 1 for prices, 2 for opening hours." They are cheap, predictable and fine for simple questions, but they break as soon as a customer types something unexpected.

2. AI chatbots on your website

An AI assistant uses a large language model (from OpenAI, Anthropic, Google or others) to understand free text and answer from your own content: services, prices, FAQs, policies. A good one stays on topic, admits when it does not know, and passes the visitor to a human or a contact form.

3. AI chatbots on WhatsApp

The same idea, but inside WhatsApp, which customers in much of Europe, Latin America, India, Africa and the Middle East use every day. This needs the WhatsApp Business Platform (the API) through a provider, and Meta charges per message. It is often the most valuable option for service businesses, and the one with the most moving parts. Our WhatsApp Business API guide covers the setup in detail.

Setup costs vs monthly costs

Most chatbot budgets fail because people only look at one of these.

One-time setup

• Mapping the questions customers actually ask, and the answers your team gives today.
• Writing and structuring the knowledge the AI answers from.
• Designing the handover to a human: when, to whom, and how fast.
• Connecting the bot to your website, WhatsApp number, CRM or booking system.
• WhatsApp verification and message templates, if WhatsApp is involved.
• Testing in every language you support, including awkward questions.

Recurring monthly costs

• Platform subscription: the chatbot or inbox software.
• AI model usage: billed by the amount of text processed.
• WhatsApp message fees: paid to Meta, usually through your provider.
• Maintenance: updating prices, offers and answers, and reviewing conversations the bot got wrong.

Pricing models vary a lot. Some platforms charge a flat monthly plan. Others charge per outcome: Intercom's Fin AI agent, for example, charges $0.99 per resolved conversation on top of seat fees. Agencies building fully custom AI agents often quote five figures. Each model makes sense for a different size of business, so compare them on your expected monthly volume, not on the headline price.

What AI model usage really costs

AI models are billed per "token", roughly three quarters of an English word. You pay for the text sent to the model (your instructions, the relevant knowledge and the conversation so far) and the text it writes back.

As one verified example, Anthropic's Claude Haiku 4.5, a fast model suited to customer service, costs $1 per million input tokens and $5 per million output tokens. A hypothetical eight-message conversation that sends about 4,000 tokens of context each time and gets short answers back would cost around $0.04. A thousand such conversations a month would be roughly $40 in AI usage.

So for most small and mid-sized businesses, the AI itself is not the expensive part. Costs climb when you use the largest models for every message, send huge documents with each request, or let the bot run long, unfocused chats. A well-built assistant sends only the knowledge relevant to each question.

WhatsApp fees to add to the budget

If your assistant lives on WhatsApp, Meta's per-message fees apply. They depend on the country of the customer receiving the message. From 1 October 2026:

• Service replies (answering within 24 hours of the customer's message): the first 1,000 per number per month are free, then each is charged at the country's utility rate.
• Utility templates such as booking confirmations and reminders: now charged even inside the 24-hour window. Examples: about $0.0034 in North America, $0.022 in the UK, $0.055 in Germany, $0.0014 in India.
• Marketing templates: from about $0.0118 in India to $0.1365 in Germany, with the UK at $0.0635.
• Click to WhatsApp ads: conversations started from these ads open a free window when you reply in time.

Local taxes apply, and your provider may add a markup. One more 2026 rule: Meta bars general-purpose AI assistants from the WhatsApp Business API, but a business using AI to serve its own customers is allowed.

What drives the price up or down

• Number of channels: website only, WhatsApp only, or both plus Instagram.
• Languages: one language, or several (for example English, French, Spanish or Arabic). Each needs testing by a native speaker.
• Integrations: a bot that only answers is simpler than one that checks availability, books slots and creates CRM deals.
• Volume: more conversations mean more AI usage and more message fees.
• Quality of your existing content: clear prices and FAQs make setup faster. Scattered information takes longer to organise.
• Handover design: routing to different team members by topic, region or language adds work, and adds value.
• Ongoing care: a bot nobody updates starts giving wrong prices within months.

Is it worth it? Two hypothetical ROI examples

These examples are illustrations with made-up numbers, not results from real clients. Plug in your own figures.

Example A: a dental clinic

Imagine a clinic that receives 600 chat enquiries a month, and staff spend about 4 minutes on each: 40 hours a month. If an AI assistant fully handles half of them (prices, location, opening hours, booking), that frees around 20 hours. If faster replies also turn just three more enquiries a month into patients, the extra revenue alone may cover the monthly cost.

Example B: a real estate agency

Imagine an agency getting 300 leads a month from portals and ads, many arriving at night or on weekends. An assistant that replies within a minute, asks budget, area and timeline, and pushes qualified leads into the CRM means agents start the day with sorted leads instead of a cold inbox. If that helps close one extra deal a quarter, the commission usually dwarfs the chatbot cost.

The pattern is the same: the return comes from speed of reply and staff time, not from the bot alone. If nobody follows up on the leads it qualifies, the return disappears.

Data privacy and AI rules

Chatbots collect personal data: names, phone numbers, sometimes health or financial details. Wherever you are, the same basic rules apply.

• GDPR (EU and UK): you need a lawful basis for processing, a clear privacy notice, a data processing agreement with each provider, and safeguards when data leaves the EU or UK. Health data is a special category with stricter rules.
• EU AI Act: from 2 August 2026, people in the EU must be told when they are talking to an AI system, unless it is already obvious. The 2026 changes to the Act delayed the high-risk rules but kept this disclosure date.
• United States: there is no single federal privacy law, but several states have their own. California, for example, has required bots used to sell or influence a purchase to disclose that they are bots since 2019.

A practical checklist for any chatbot:

• Tell users they are talking to an AI assistant, and how to reach a human.
• Link to your privacy policy and explain what you collect and why.
• Collect only what you need to answer or book.
• Avoid asking for sensitive data such as medical history or ID numbers in chat unless there is a clear legal basis and a secure process.
• Know where your platform and AI provider store and process data.
• Choose AI providers that do not train their models on your business data by default for API use, and check their terms.
• Restrict who on your team can see chat transcripts.

This is general information, not legal advice. If you handle sensitive data, check with a lawyer in your country.

How dhowly prices it

We keep this simple. Our Growth Partner plan is AED 2,750 per month (about USD 750) and includes an AI assistant on your website or WhatsApp, monthly updates, monitoring of your automations and a monthly report. If you also need a website or CRM connected first, our Automate package starts from AED 2,450 (about USD 670). Prices exclude VAT where it applies and third-party subscriptions such as WhatsApp message fees, and we tell you upfront what those are likely to be. Details are in our pricing section and on our AI chatbot service page. If you want the assistant to feed leads into a pipeline, see our automation service.

If you are in the UAE

At 3.6725 AED to the dollar, the AI usage example above is about AED 150 a month for 1,000 conversations. Meta's UAE WhatsApp rates from 1 October 2026 are $0.0576 (about AED 0.21) for marketing and $0.0157 (about AED 0.058) for utility, authentication and service replies beyond the free 1,000, plus 5% VAT. As a local reference, SleekFlow lists its Pro AI plan at AED 547 a month and Premium AI at AED 1,098 a month, before message fees. The UAE Personal Data Protection Law (Federal Decree-Law No. 45 of 2021) applies to most mainland businesses, while DIFC and ADGM have their own data protection laws and healthcare has extra sector rules. Cross-border data transfers have conditions, and you should avoid collecting Emirates ID numbers in chat. Test Arabic answers with a native speaker, including dialect.

Next step

dhowly works with businesses worldwide, remotely, in English and French. Not sure which type of chatbot you need, or whether you need one yet? Book a free 15-minute call. We will look at how customers contact you today and give you an honest estimate of setup and monthly costs.

Frequently asked questions