It is the first question almost every business asks, and the honest first answer is the one nobody likes: it depends. But it does not depend on magic. It depends on a short list of things you can actually understand, and once you see them, you can put a realistic number on your own case. Here is that list, with real 2026 ranges, in plain euros.

First, two very different things are both called "a chatbot"

The word covers two products with wildly different price tags. The first is a ready-made subscription tool: you sign up, paste it on your site, and it answers from a help page. The second is a custom AI agent: something built for your business that reads your own documents, connects to your systems, and can actually do things (create a ticket, check an order, update a record). Confusing the two is where most price shock comes from.

Option 1: the ready-made subscription tool

If your need is simple (answer the same handful of questions on a website), an off-the-shelf tool can cost roughly between 20 and 300 euros per month, depending on the volume of conversations and the features. This is the cheapest way to start and the fastest to switch on. The trade-off: it knows only what you feed its panel, it does not really connect to your invoicing or your CRM, and it sounds generic because thousands of other businesses use the same engine. For a small shop with simple questions, that can be perfectly fine.

Option 2: a custom build for your business

When the bot needs to answer from your documents, speak in your tone, and connect to the tools you already run, you are now building software, and the price reflects that. As a rough 2026 guide:

  • A focused custom assistant on your own content (a proper retrieval setup over your documents, one or two channels) typically lands in the low five figures.
  • An AI agent that takes actions and integrates with your systems (helpdesk, CRM, e-shop, accounting) is more, because the integrations are the real work. Connecting to your existing tools alone commonly adds twenty to forty percent on top of the base.
  • Heavily regulated or high-accuracy cases (anything touching money, health, or personal data under strict rules) sit at the top of the range, because auditability and safeguards cost engineering time.

The cost nobody puts in the brochure: running it

A chatbot is not a one-time purchase, it is a small piece of live software. Whatever you build, budget for the monthly running cost: the AI model usage (you pay per conversation behind the scenes), hosting, and ongoing maintenance as your content and your tools change. A useful rule of thumb is to set aside roughly fifteen to thirty percent of the build cost per year for upkeep, plus the monthly usage. Anyone who quotes you a build price and goes silent on the running cost is hiding half the picture.

What actually moves the price (so you can lower it)

  • How many integrations. A bot that just talks is cheap. A bot that reaches into three systems and does things is not. Each connection is real work.
  • Volume. More conversations means more model usage every month. It rarely changes the build much, but it changes the running cost.
  • How much it must get right. "Roughly helpful" is cheap. "Never wrong about a price or a deadline" needs guardrails, testing, and a human in the loop, and that is engineering.
  • Languages and channels. Greek and English, website plus Viber plus email: each adds a little.

How to find your number before you spend a euro

You do not have to guess. The honest way to price an AI project is to start from one task, not from a tool. Pick the single most repetitive job (the same questions, the same lookup, the same reply), and have it proven on your real data first. That one proof tells you the value (hours saved per week) and the cost (what it actually takes to build it well) at the same time, so you are deciding with numbers instead of a brochure. We treat that first proof as a free proof of concept precisely so the cost-to-value question gets answered before you commit. If it does not pay for itself, you have lost nothing.

If you are still deciding whether to automate at all, our guide on where to start with AI automation walks through picking that first task. For customer questions specifically, see how we build an AI chatbot for customer support in Greek. And when you want a real figure for your own case, the fastest route is to tell us the one task you would automate first: we will tell you honestly what it would cost to do it well, and prove it free before you spend.