An AI chatbot on FAQs and documents starts from €2,000, an agent connected to your ERP from €5,000. Here is what each tier includes, what model and server cost each month and what pushes the price up.
An AI chatbot answering from FAQs and documents costs from €2,000, a WhatsApp Business chatbot on Meta’s API from €3,000, an AI agent connected to your ERP from €5,000 and a project spanning several processes and integrations from €15,000. Running costs for the AI model, server and maintenance come on top and, for a low-volume chatbot, start at a few tens of euros a month.
| Solution | Starting price | What it includes |
|---|---|---|
| Website AI chatbot on FAQs and documents | from €2,000 | Knowledge base, AI Act notice, handover to a person by email or chat, log review in the first month |
| WhatsApp Business chatbot (Meta API) | from €3,000 | As above, plus Meta business verification, number, message templates and a shared inbox for staff |
| AI agent connected to your ERP | from €5,000 | One process (for example orders from email), data extraction, draft entries in the ERP, human review, logging |
| Multi-process agent or private AI project | from €15,000 | Several integrations, per-user permissions, model on a dedicated or on-premises server, evaluation before release |
| n8n automation, ChatGPT or Copilot rollout, training | quoted after the review | Depends on the number of workflows, users and sessions |
Prices exclude VAT and apply as of November 2026. They are starting amounts: the final quote follows the free review, once the process is clear.
After launch an agent or chatbot has three monthly costs. We always list them separately in the quote, so you know what you pay us and what you pay others.
Cloud models (OpenAI, Anthropic, Google) are pay-per-use, based on how much text is processed. For a chatbot answering a few hundred questions a month the spend stays low; it grows with high volumes, long documents and more powerful models. A private model has no per-use cost, but the dedicated server has a fixed fee.
The assistant, its knowledge base and logs run on a managed server in Europe or on yours. On WhatsApp you add the messages Meta charges for: per message since 1 July 2025, and from 1 October 2026 free-form replies too above 1,000 free per month per number (source: Meta for Developers, WhatsApp Business Platform pricing, checked on 9 November 2026).
Content updates, log reviews, fixes when your ERP or procedures change, security updates. Agreed as a monthly fee or a block of hours.
As a rough guide, for a custom chatbot with real traffic, model and hosting usually run between €200 and €500 a month.
The questions we ask during the review.
A chatbot that only reads documents costs less than one querying ERP, CRM and calendar.
Up-to-date procedures in one folder load quickly. Scattered files and duplicate versions need clean-up work.
Cloud is cheaper to start. Private on-premises AI costs more up front, but data stays in-house.
Answering is simpler than writing into the ERP. Every action needs permissions, checks and a log.
Website, WhatsApp, email or Teams: each extra channel adds setup and testing.
More conversations mean more model usage. We estimate it from your real numbers before quoting.
The maths is done on the process, not the technology. During the review we measure with you how often the task happens, how long it takes and what an error costs. If an office spends several hours a week retyping orders from email, you can compare that time with a year of the agent, running costs included.
If there is no saving, we say so. It happens more often than you would think: sometimes a simple n8n automation beats an agent, or tidying the documents comes before adding AI. Italian companies may be able to include the spend in the 2026 super-depreciation scheme, which needs checking with their accountant.
The starting price covers the process review, building the knowledge base from the documents you provide, channel setup, the AI Act notice, testing with your team and the first month of log review. You get documentation of what the assistant does and which data it uses.
Third-party costs are not included: AI model usage, software licences such as ChatGPT or Copilot, messages Meta charges on WhatsApp and, if you choose on-site AI, the hardware. We list them separately with the price-list source and the date we checked it, so you can compare them or buy them directly.
Writing new content, such as FAQs that do not exist yet, can be left to us or done in-house: in that case we give you a template to follow.
With no surprises halfway through.
Thirty minutes to understand process, volumes, systems and data constraints.
Set-up cost and running costs kept apart, with third-party items listed separately.
Start small: one process, one channel, clear metrics.
After a few weeks we review real times and costs together.
Only what worked gets extended.
The solutions these prices refer to.
Describe the process or channel and rough volumes. We reply within one working day with a first estimate.
We will email you within one working day to book the 30 minutes. If it is urgent, message us on WhatsApp.
Prefer to talk right away?
Mon–Fri 9:00–12:30 and 14:00–20:00 (CET).