Visitors to your site tend to ask the same questions, often at hours when nobody is there to reply, and some of them leave without an answer. An AI chatbot for business handles that first exchange: it informs, points people in the right direction, qualifies the need and offers an appointment. Its job is not to know everything, but to deal well with what comes up most.
A demonstration on your own questions, free
A useful chatbot does more than talk: it answers from your own documents, recognises what it does not know, and hands over to your team at the right moment. Here is what that looks like on an online shop.
Simulated conversation, for illustration.
Try a question:
Demo transcript
The clearest signal is repetition: the same enquiries by phone, email or contact form, eating sales time for answers that already exist in writing. The second is losing contacts outside business hours.
We start from your actual enquiries: email history, call transcripts, questions submitted through the contact form. That material defines the response perimeter: the topics the chatbot handles, those where it redirects, and those it explicitly declines.
The chatbot draws on a knowledge base built with you, then connected to a language model, for example through a mainstream provider’s API. Answers are generated from your approved content rather than the model’s general knowledge, which sharply reduces the risk of an invented reply. Outside its perimeter, the chatbot says so and offers a handover: a form, a slot in your calendar, a phone number or a messaging channel.
It can be installed on your site, on a campaign landing page or on an instant messaging service your customers already use, for example WhatsApp. A dashboard lists conversations, unresolved requests and the appointments generated.
The word “chatbot” covers very different tools. The right choice depends on what you expect from it: guiding, answering, answering from your sources, or acting.
| How it works | Best for | Limits | Budget | |
|---|---|---|---|---|
| Scripted chatbot | A tree of questions and answers written in advance. | Steering visitors to a few simple paths. | Does not understand unexpected questions. | € |
| Generative AI chatbot alone | A language model answers freely. | Conversing, rephrasing, drafting. | Can make things up: avoid in front of customers without sources. | € |
| RAG chatbot on your documents | The model answers only from the passages found in your sources, which it cites. | Answering accurately about your offer, terms and procedures. | Depends on the quality and freshness of the documents. | €€ |
| AI agent | The model plans and acts in your tools, with rules and approvals. | Handling a request end to end. | Requires precise scoping of rights and controls. | €€€ |
The customer-facing chatbots Kopus deploys are based on RAG: answers limited to your sources, cited, with human hand-over.
We group your incoming enquiries by theme and frequency, and you choose which to cover first.
Each theme gets a reference answer reviewed by your team, including the cases where the chatbot should hand over.
Phrasing, qualification questions, the moment to offer an appointment, the refusal rule outside the perimeter.
Your staff and a few willing customers talk to the chatbot, and we fix the weak answers.
Installation on the chosen channels and monitoring over the first weeks, with adjustments based on unresolved requests.
A chatbot speaks on behalf of your company. It must be as reliable and as discreet as a well-trained employee.
The chatbot answers only from the documents you approve and shows its source. Without reliable information, it says so.
Conversations are used to answer and improve your base, not to train a model. The provider’s terms are checked with you.
Hosting in Europe where required, a defined retention period for conversations, restricted access.
Information for visitors, minimal data collection, rights of access and deletion, an up-to-date record.
Sensitive or out-of-scope requests are passed to your team with the conversation history.
Conversation review at launch, tracking of unanswered questions and reports, regular adjustments.
At Kopus, an AI chatbot that answers from your content usually costs €3,000 to €7,000 excl. VAT to build, then €100 to €300 excl. VAT per month for hosting, model usage and supervision. A chatbot connected to your CRM or calendar, on several channels or in several languages, sits between €8,000 and €20,000 excl. VAT, with running costs of €250 to €800 excl. VAT per month. The monthly cost depends mostly on conversation volume.
€3,000–€7,000 excl. VAT
Running cost: €100 to €300 excl. VAT / month
Answering your visitors’ frequent questions, on your website.
Typical timing: 3 to 5 weeks
Request a quote€8,000–€20,000 excl. VAT
Running cost: €250 to €800 excl. VAT / month
Qualifying requests, booking appointments, answering on several channels.
Typical timing: 6 to 10 weeks
Request a quoteFrom €20,000 excl. VAT
Running cost: depending on volume
An assistant that acts in your tools: opening a case, tracking an order.
Typical timing: schedule set during scoping
Request a quoteIndicative ranges, excluding VAT, based on our projects. The running cost is estimated on your conversation volume before any commitment.
Your project deserves a precise quote
Describe it in a few lines, even without a specification: we will come back to you with a reasoned estimate and the possible options.
Answers to the questions we are asked most often about this service. Another question? Ask us directly.
Ask a questionAnswering your visitors’ frequent questions, guiding them through your offer and qualifying their need before passing serious prospects to your sales team.
No. Conversations are used only to reply and to improve your knowledge base. We check the chosen provider’s terms with you, along with where the exchanges are stored.
A business chatbot works before the sale, to inform and qualify. A support chatbot works after it, on existing accounts, with the follow-up and escalation mechanisms that support requires.
Yes. It can answer in the visitor’s language from content written in one language, with a quality we check for each language enabled.
At Kopus, allow €3,000 to €7,000 excl. VAT to build a chatbot that answers from your content, then €100 to €300 excl. VAT per month to run it. A chatbot connected to your CRM or calendar, on several channels, costs €8,000 to €20,000 excl. VAT, with €250 to €800 excl. VAT per month.
Yes, provided it is connected to your calendar or your CRM. It collects the useful information, suggests slots and passes a complete file to your team.
By restricting it strictly to your sources, configuring it to say “I don’t know” rather than guess, and reviewing conversations in the first weeks. That is the method Kopus applies to every deployment.
A generic AI chatbot answers from its general knowledge and can make things up. A RAG chatbot first retrieves the relevant passages from your documents, then answers only from them and cites the source. That is the approach to choose for answering customers.
It can be, provided visitors are informed, the data collected is kept to a minimum, the provider does not train its models on your conversations and a retention period is set. Kopus handles these points from scoping onwards.
Let’s design a chatbot that genuinely answers your visitors and sends you the right contacts.
A demonstration on your own questions, free
