Chatbot pricing in 2026: SaaS, custom builds and hidden costs

How much does a chatbot cost? The question looks simple, and most of the published answers are too simple. You find ranges dropped in a single sentence, tables of software compared on a gut feeling, or the opposite: a dossier so dense that you never come out of it with a figure you can put in a budget. What is missing, in between, is what actually moves the bill: what you buy, and who builds it.

This article breaks the price of a chatbot down line by line, with dated figures: Botnation’s public grid as read on 18 September 2026, the AI credit equivalences published by the same page, and order-of-magnitude figures from third-party providers to situate custom builds. It ends with a comparator that adds up both routes over three years, because the real question is not the displayed price, but the cumulative cost of both routes, and the moment when the route that starts the most expensive becomes the cheapest.

The topic deliberately differs from our more specialised dossiers: if your challenge is multi-team governance and compliance for an enterprise AI chatbot platform, or choosing a provider (see our dossier on the AI chatbot agency), read them as complements. Here, we talk budget.

The price of a chatbot reads across three cost lines

Whatever the vendor, a chatbot bill always breaks down into three lines. Mixing them up is the first cause of bad surprises, because each line follows a different logic: one is a predictable subscription, the second varies with your traffic, the third depends on an organisational decision.

1. The licence is the subscription to the platform: the scenario editor, the channels (website, WhatsApp, Messenger, Instagram), hosting, updates. It is the most readable line, displayed on pricing pages, and the easiest to cap: you pick a tier, it shows up on the invoice every month.
2. Usage covers the variable consumption: AI requests, conversation volume, extra users. It is the line that surprises, because it depends on data few companies measure before launching: what their visitors actually type to the chatbot.
3. Construction, finally: who designs the scenario, wires the integrations, connects the knowledge base, tests and fixes. Done by your team in a no-code editor, this line costs time. Delegated to experts, it is quoted.

These three orders of magnitude read together, not separately. The SaaS grid says nothing about construction cost if you delegate; the custom range says nothing about the licence you will have to pay anyway to run the tool. A complete budget adds them up, which is exactly what the comparator below does.

Four factors move the bill

For an identical subscription scope, two chatbots from the same vendor can end up costing vastly different amounts. The gap almost always comes from one of these four factors:

• The number of channels. A website-only chatbot is the cheapest case. Each added channel (WhatsApp, Messenger, Instagram) widens coverage, and sometimes volume: a popular channel can multiply conversations, and therefore the “usage” line.
• The AI engine you pick. On credit-based platforms, not all answers are equal: a cheap generic engine, an AI personalised to your brand or a premium model do not consume the same number of credits per request. On Botnation, the pricing page details these equivalences: for the same credit budget, you do not get the same number of answers depending on the engine.
• Integrations with your information system. Connecting the chatbot to a CRM, a helpdesk or an internal base is rarely included in a standard build. It is the craft of the chatbot integrator, and it is often that line, more than the scenario itself, that swings a quote.
• Who builds and who maintains. Building in-house costs internal time (training, writing scenarios, testing); delegation costs a quote. In both cases, someone has to maintain: scenarios to enrich, knowledge base to update, statistics to read. This recurring cost is the most often forgotten in the initial budget.

Route 1: self-service SaaS and its public grid

This is the most readable route: you subscribe, you build in the no-code editor, you publish. The displayed price covers the licence, and construction is an investment in internal time rather than money. At Botnation, the public grid as read on 18 September 2026 reads as follows:

Two points of this grid deserve a close read. First, the free AI credits are granted only once at sign-up: they help you start, not run the whole year. Second, the ENTREPRISE plan is not a price tier but another mode of relationship: it literally includes “chatbot creation management”, that is, the delegated construction discussed in route 2. The public grid also ends with the mention that all prices are quoted excluding taxes, VAT applying to legal entities located in the European Economic Area.

AI credits: the usage line, in public figures

Beyond the included credits, AI usage is bought in packs, displayed on the same page with volume discounts:

The usage line has another component, displayed right below the packs: a user outside of a plan costs €0.05 per user per month. On a deployment that exceeds the 500 or 1,000 users included depending on the tier, this line counts: 2,000 users beyond the PRO plan means €100 more per month, to be written into the budget as seriously as the licence itself.

Route 2: delegated construction and its quote

Delegating construction means paying once to get a chatbot built by specialists, then running it on the route 1 subscription. The price depends on scope: number of scenarios, channels, integrations, knowledge base, service level. Two families of sources help situate it.

First source: Botnation’s own pricing page, which answers the question “How much does it cost to create an agent?” in its FAQ, distinguishing the two routes. After presenting the SaaS plan, it writes: “Our team of bot developers will create a custom AI chatbot for you, based on a quote.” And to situate that quote, the same FAQ gives a range: “between €5,000 and €30,000, or even more, depending on the features you need”, immediately adding: “These price ranges are indicative, as it is difficult to estimate the cost of developing your chatbot.” Read on 18 September 2026, this range has the merit of being published by the vendor itself: it is an indicative order of magnitude, not a price list, and even less a commitment.

Second source: third-party providers who publish their own orders of magnitude in French. They cross-check remarkably well:

• Radiank, consulted on 18 September 2026, opens its guide with (published in French only): « La réponse franche : entre 0 € et plus de 50 000 €. » (in plain English, from €0 to more than €50,000). It prices, from €0 to more than €50,000; it prices an enterprise chatbot configured on a platform at « 50 à 300 €/mois d’abonnement + 500 à 2 000 € de configuration initiale » (€50 to €300 per month of subscription plus €500 to €2,000 of initial setup) when you go through a provider; for a chatbot connected to internal documents (RAG), it announces a setup of €3,000 to €15,000 and running costs of €200 to €1,500 per month.
• Digital Unicorn, consulted the same day (published in French only), situates provider-led setups « au l’alentour de 2000 à 5000 euros » (around €2,000 to €5,000) for a first level, a simple chatbot between €1,000 and €5,000, and a hosting and maintenance infrastructure of €50 to €200 per month at basic level, €300 to €1,000 per month at scalable level.
• Agenda Formation, specialised in the public market (published in French only), prices a phone-answering robot at « à 700 € HT par mois sur le marché public » (€700 excluding taxes per month on the public market) and a customisation at « à 1000 € par jour » (€1,000 per day), with market amounts of « 0,50 € HT à 0,70 € HT par minute d’usage » (€0.50 to €0.70 excluding taxes per minute of usage) for voice agents.

These third-party figures obviously do not bind Botnation, and they should be read as market milestones published in French, not as a price list. They still draw a stable hierarchy: a first chatbot built with a provider’s help usually sits between €2,000 and €5,000, an integrated enterprise chatbot between €5,000 and €15,000, and deeply integrated projects exceed €30,000, up to well beyond €50,000 in the complex cases quoted by Radiank.

At Botnation, this delegated route goes through the ENTREPRISE plan and its “bot developers”: the team that builds the custom chatbot is the one that edits the platform. Formulated as a comparison, this model has a concrete budget consequence: the scenario is built in the no-code editor your team keeps in hand after delivery, on your own account, which reduces the maintenance-dependency risk encountered when the build was made in the provider’s proprietary tools. This does not rule out the other paths: an agency or chatbot consultant remains legitimate for a very specific trade, an on-site arrangement or an ecosystem of tools already in place at your company; what matters is comparing quotes on the same scope, not on the same figure.

The right reflex to frame this route: request a quote describing the use case, the channels and the systems to connect, and weigh it against the three-year cumulative cost of the SaaS route, which the comparator below computes for you. The team is also available online, as shown by the “Contact sales team” button of the ENTREPRISE plan.

The hidden costs to write into the budget

Beyond the three cost lines, a realistic budget plans for the expenses that appear on no price line but almost always end up happening:

• AI credits beyond the pack. Packs renew: a successful chatbot consumes, and a request on a premium engine consumes ten times more than a GPT-3.5 request according to the equivalences published by the pricing page. This is the first line that moves during the year.
• Users outside of a plan. €0.05 per user per month beyond the tier: invisible on a mock-up, significant on an internal deployment of several thousand people.
• Maintenance and evolutions. In self-service, it is internal time: enriching scenarios, updating the knowledge base, reading statistics, fixing unanswered questions. Delegated, it is a contract: the third-party providers quoted above price hosting and maintenance at €200 to €1,500 per month depending on complexity.
• Partial rebuilds. A chatbot ages: new products, new organisation, new channels. Planning an annual enrichment budget avoids discovering in month thirteen that the tool has not followed the company.
• Team time. The most underestimated cost: training, writing scenarios, qualifying answers. Even on a no-code platform, someone writes, tests and fixes. That time exists whatever route you choose; only its owner changes.

SaaS or custom: compare the 36 months, not month 1

The question that really settles it: over three years, which route costs less? The comparator below adds up, using the public prices read on 18 September 2026: the monthly licence, the credit pack you renew, then the delegated construction and its yearly running cost if you choose it. It displays the cumulative cost of both routes at 12, 24 and 36 months, and the month when, if any, the custom build stops being more expensive.

Comparator assumptions: the licences (€39 and €59 per month) and the packs (€100 and €250) come from Botnation’s pricing page as read on 18 September 2026; the delegated construction amounts (€2,000, €5,000, €15,000 and €30,000) combine the low end published by a third-party provider and the indicative range of the same page’s FAQ; the FOR FREE tier is not compared (real usage assumed). The yearly running amounts (€1,500 and €4,000) are reading assumptions specific to this article, sitting below the maintenance ranges published by the third-party providers quoted above; the running cost is smoothed over 12 months. The credits offered at sign-up (only once) are not deducted, and users outside of a plan (€0.05 per user per month) are not counted.

Getting a quote that holds the comparison, in five steps

Whether you compare two providers or a provider and the SaaS route, a quote only has value if it is built on the same scope. Five steps make it comparable:

This last step joins the broader organisational choice covered by our dossiers on the price of established vendors and on advanced chatbot training: upskilling your team is also a budget route, sometimes the most profitable of all.

Frequently asked questions about the price of a chatbot

The questions that systematically come up when budgeting, with the dated answers of this article.

The right price is the one of your volume, not the one of the shop window

The price of a chatbot is not a figure, it is an equation with three unknowns: a public and predictable licence, a usage that is measured before being bought, a construction worth either a zero-euro quote or five figures depending on who does it. Market ranges, including the one Botnation itself publishes for custom builds, serve to situate a quote, never to replace it.

The next step comes down to two moves: measure fifteen days of real conversations, then run the comparator with your own figures. If the SaaS route covers the horizon, open the free tier and build; if the custom build catches up, request the quote with the scope written down in black and white. In both cases, you will have a defensible budget, not a hunch.