What does an AI chatbot cost in 2026?

The three ways to get a chatbot (and why they cost so differently)

Almost every price you'll see online falls into one of three buckets. Knowing which bucket you're in explains a 100x difference in price more than any feature list does.

• Do it yourself with no-code SaaS. You sign up, paste your website URL or upload a FAQ, and the tool trains a chatbot on your own content. This is the cheapest and fastest route: think tens of euros per month, live the same afternoon. Best for small and mid-sized businesses and webshops.
• Have an agency build it. A partner scopes, designs and integrates a bot for you, often on top of a SaaS platform or a framework. You pay a one-off project fee plus a monthly license. Budget from a few thousand euros up. Best when you need deep custom integrations or you simply don't want to touch it yourself.
• Enterprise / custom platform. A large contact-center or conversational-AI suite (with SSO, audit logs, dedicated support, SLAs). Priced per year, usually five figures and up, quoted after a sales call. Best for large organisations with compliance and volume requirements.

Typical price ranges in 2026

These are the ballpark figures we see across the market in 2026. Treat them as orders of magnitude, not quotes; every vendor packages features differently.

For do-it-yourself SaaS, a usable plan usually sits somewhere between €25 and €150 per month, with cheaper entry tiers that cap the number of conversations and pricier tiers that unlock e-commerce actions, more channels and analytics. Free plans exist but are almost always demo-grade (a hard message cap, no removal of vendor branding).

An agency build typically starts around €2,000 to €10,000 as a one-off for scoping, content and integrations, on top of the monthly platform license. A full enterprise conversational-AI contract runs from roughly €15,000 per year upward, often much higher once you add seats, volume and professional services.

The pricing models, decoded

Two tools at "€79 a month" can cost wildly different amounts once you're live, because they meter different things. Here are the models you'll run into and what each one rewards or punishes.

• Per workspace / flat plan. One price for the whole account, whatever the team size. Predictable, and the cheapest to reason about. ovellan uses this model.
• Per seat (per agent). You pay for each human team member with a login. Fine for small teams, expensive to scale a support department, and it punishes you for adding colleagues who only handle the occasional handover.
• Per resolution / per conversation. You pay for each conversation the bot handles (or resolves). Fair-sounding, but your bill now moves with your traffic, so a busy month or a spike can be a nasty surprise. Watch for how a "resolution" is defined.
• Per message / per token. The rawest model, usually only at the API level. Cheapest per unit, but hardest to predict and it needs engineering to run safely.

What's actually included (and what should be)

Price only means something next to what you get. When you compare plans, check that these are included rather than paid add-ons, because this is where cheap tools quietly get expensive.

• Training on your own content (website, PDFs, FAQ) so the bot answers from your business, not the open internet.
• The channels you need: website widget, and often WhatsApp, Instagram, Facebook and email.
• A shared inbox and human handover, so a person can take over when the bot can't help.
• Removal of vendor branding on paid plans (a "powered by" badge on a paying customer's site looks cheap).
• GDPR-ready hosting in the EU, a signable data processing agreement (DPA), and export/delete controls.
• Analytics and a view of unanswered questions so you can improve the bot over time.

The hidden costs nobody puts on the pricing page

The sticker price is rarely the full price. Budget for these so you're not caught out in month two.

• Overage. Going over your conversation, message or seat limit often costs more per unit than the plan implies. Read the overage rate, not just the base price.
• Setup and content. Someone has to gather the FAQ, tidy the website content and test the answers. If you don't do it, an agency will, and that's billable time.
• Extra channels and integrations. WhatsApp, a CRM sync or a webshop connector can sit behind a higher tier or a paid add-on.
• Annual lock-in. The headline monthly price is often the annual-billing price; paying monthly can be noticeably higher.
• Your team's time. A poorly grounded bot that hallucinates creates support tickets instead of removing them. A bot that answers honestly from your content saves time; one that guesses costs it.

How ovellan prices it

We price ovellan per workspace, not per seat, so adding a colleague never raises your bill. Plans are flat and public: Starter at €29 per month, Pro at €79, Business at €149, and Enterprise from €199 (contact sales). Every self-serve plan starts with a 7-day free trial, and Starter needs no card up front.

What's included from the start: training on your own content, the website widget, a shared inbox with human handover, EU hosting in Germany, a signable DPA, and no ovellan branding on paid plans. Higher tiers add e-commerce actions (product cards, live stock, checkout links), more channels like WhatsApp, and richer analytics. You can try the real engine on a sample knowledge base before you pay a cent.