Customers
AI customer service for your business: a chatbot that knows when to hand over
AI customer service that answers from what your company has written, admits what it does not know and passes the case to your team with the context. Obinova builds it for your website or for internal use, connected to your tools.
What AI customer service can handle, and what it cannot
When a chatbot frustrates a customer, it is usually for one of two reasons. Either it answers confidently about something nobody gave it information on, or it will not let them reach a person. Done well, AI customer service answers only from what your company has written, says when it has no answer and passes the case to your team with the conversation summarised.
There are two main uses. A website chatbot serves customers and prospects: services, terms, opening hours, lead times, the status of a request, booking an appointment or asking for a quote. An internal chatbot answers your staff: how to claim an expense, what the holiday policy says, where a product sheet lives or how to use a company tool.
In both cases, it should not make decisions for you. Unpublished discounts, complaints, legal or tax advice and anything sensitive go to a person. Obinova builds these chatbots with n8n, connected to your documents, your CRM and your help desk.
What the service includes
An organised knowledge base
We gather your FAQs, terms, product sheets and policies, remove duplicate versions and write what is missing.
A chatbot on your website
Built into your pages, in your brand's tone and in your customers' languages. It introduces itself as an automated assistant from the first message.
Handover to a person
When it is unsure, when the topic is sensitive or when the user asks, it creates an alert or a ticket with a summary, so your team can reply without asking again.
Lead capture
If a visitor wants a quote, the chatbot asks the necessary questions and creates the contact in your CRM with the conversation summarised.
Internal chatbot
In the chat tool your team already uses, with access to each document based on the asker's group.
Review and improvement
A conversation log and a weekly list of unanswered questions, used to fill the gaps in the knowledge base.
Before building: the source sheet
Every answer the chatbot gives should come from a document someone keeps up to date. So we start with a five-column spreadsheet, which your team keeps afterwards.
Topic
Each subject the chatbot may answer on: delivery times, warranty, payment methods, opening hours, expenses policy.
Source document
The exact text the answer comes from. If it does not exist, it is written before the chatbot goes live, or the topic stays out.
Owner
The person who confirms the information is right and flags it when a price, a lead time or a policy changes.
Review
How often the document is checked. Price lists may need a monthly check and an internal policy a yearly one.
What it never answers
Topics that always go to a person, such as off-list discounts, complaints, other customers' data or any legal advice.
How we build your AI chatbot
- 01
Collect the real questions
We go through your recent emails, forms and messages and group them by topic, so the chatbot starts with what people actually ask.
- 02
Fill in the source sheet
We give each topic a document and an owner, and write the missing texts with you.
- 03
Build and set the tone
We build the chatbot, its handover rules and the link to your CRM or help desk, and tune the style of its answers.
- 04
Test with real questions
We run a batch of real questions, including awkward ones and ones with no answer. Each result is marked correct, incomplete, wrong or should-have-handed-over, and we fix things until errors are rare.
- 05
Open in a limited way
We start on some pages or at certain times of day, and widen it once real conversations confirm the tests.
- 06
Review every week
We read the unanswered questions and the handovers, and add what is missing to the knowledge base.
Personal data, the GDPR and the AI notice
Article 50 of the EU AI Act has applied since 2 August 2026. It requires that people know they are interacting with an AI system, unless that is obvious, at the latest at the first interaction. The chatbot introduces itself as an automated assistant in its first message and shows it clearly in the chat window.
A conversation with a customer contains personal data. The chat window links to your privacy policy, the chatbot asks only for the data the task needs and conversations are deleted after the period you agree with us. You also need data processing agreements with the model and platform providers.
With an internal chatbot, the main risk lies in permissions. Someone in the warehouse should not get answers based on payroll or appraisals. So we separate documents by group, and the chatbot only consults the ones that match the person asking.
Who it is for, and what the price depends on
A website chatbot suits companies that get many similar questions by form, email or chat: professional firms, training providers, estate agencies, service businesses, distributors or software companies whose customers ask for support. An internal chatbot suits teams where the same people get the same questions several times a week.
Switching on an off-the-shelf chat widget takes an afternoon. What a chatbot agency adds is the work around it: the source sheet, the handover rules, the tests with real questions and the connection to your CRM or help desk, all documented for your team.
The price depends on the number of topics and documents, the tools to connect, the languages, the channels and the volume of conversations, which drives model usage. We send a quote after the first meeting. To prepare it, gather twenty or thirty real questions from customers or staff and the documents the answers come from today.
Frequently asked questions
How much does an AI customer service chatbot cost?
It depends on how many topics it covers, the tools it connects to, such as your CRM or help desk, the languages and the volume of conversations. There is a build cost and a running cost, which includes AI model usage and the platform. At Obinova we give you a specific quote after the first meeting.
How is an AI chatbot different from a scripted one?
A scripted chatbot follows a tree of buttons and set phrases and gets lost as soon as a question does not fit. An AI chatbot understands questions however they are worded and writes the answer from your documents. In return it needs a well-kept knowledge base, testing and clear rules for handing over to a person.
Can the chatbot make up answers?
An AI model can state something false as fact if it is not constrained. That is why the chatbot answers only from the documents on the source sheet, says it does not know when the information is missing and hands the case to a person. The pre-launch tests and the weekly review catch the errors that slip through.
Can the chatbot hand the conversation to a person?
Yes, and it is central to the design. When it is unsure, when the topic is sensitive or when the user asks, the chatbot creates an alert, an email or a ticket with a summary of the conversation. During office hours it can also offer direct contact with your team, depending on the tools you use.
What does a chatbot agency do that a ready-made tool does not?
A ready-made tool gives you the chat window. An agency does the work around it: collecting the real questions, organising the documents, setting handover rules, connecting the CRM or help desk, testing with real questions and reviewing conversations after launch. We hand all of it over documented so your team can maintain it.