Merlin builds chatbots and AI assistants plugged into your content: procedures, contracts, knowledge base, product documentation. They cite their sources, know their limits and hand over to a human when needed.
An AI chatbot is a conversational assistant that answers in natural language. The difference between a gadget and a work tool comes down to four design choices.
Document retrieval (RAG) grounds every answer in your documents: procedures, contracts, resolved tickets, documentation. No generic answer fished from the internet.
Every answer links back to the document behind it. Your teams or customers can verify, and you spot at a glance which content needs updating.
Below a confidence threshold, it does not improvise: it says it does not know and passes the conversation to a human, with the context already summarised.
Combined with AI agents, the chatbot does more than answer: it creates a ticket, updates a file, triggers a follow-up. The conversation becomes an entry point into your processes.
Immediate answers to recurring questions, around the clock, with escalation to your teams as soon as the topic gets sensitive or confidence drops.
Leave, expense reports, procedures, access requests: the questions that come back every week get answered without tying up your support teams.
Your documentation becomes a conversation: ask a question, get the answer and the source document, instead of digging through shared folders.
The chatbot answers visitors' questions, pins down the need and hands your sales team an already-qualified lead, with the conversation history.
An assistant built into your application, grounded in your domain content. That is the principle behind Acuity's chatbot, featured in our case studies.
When the channel calls for it, the assistant extends to voice: phone reception, qualified message taking, answers to simple questions.
The same public ranges as on our general price list, because a provider that hides its prices wastes your time.
| 30-min diagnostic | Free |
|---|---|
| Process audit | from €1,500 |
| Chatbot / AI assistant (RAG) | €4,000 – €20,000 |
| Production support | from €300 / month |
Indicative ranges: every project is quoted after the audit. What moves the price: document volume, deployment channels, compliance requirements, hosting.
Firm quote after the audit: no time-based billing, no surprises.
We frame the use case and assess the corpus: which documents, what quality, which questions actually come back. A chatbot is worth what its sources are worth.
Scope of the answers, tone, channels, escalation rules to humans, personal-data handling. Everything is settled before the first line of code.
Document retrieval is plugged into your content, then the chatbot is tested on real questions, with your teams as judges. A demo every week.
Conversation monitoring, weak answers corrected, corpus kept up to date. A monthly review of questions solved and those left to cover.
The week-by-week method is detailed on the home page. Six to eight weeks depending on the tools to connect.
A conversational assistant that answers questions in natural language. The version we care about answers from your content (document retrieval, RAG): procedures, contracts, knowledge base, product documentation. It cites its sources and hands over to a human when it does not know.
A chatbot with document retrieval (RAG) costs between €4,000 and €20,000 depending on scope: document volume, deployment channels, compliance requirements. Production support starts at €300 per month. Our ranges are published in the Pricing section and every project is quoted as a fixed price after the audit.
A chatbot answers questions in a conversation. An AI agent acts inside your tools: it creates a record, sends a follow-up, updates a file. The two combine well: a chatbot up front, agents doing the work behind it.
A bare model can, yes. That is why our chatbots answer only from your documents, cite their sources and say 'I do not know' rather than improvising: below a confidence threshold, the question goes to a human.
Your website, your application, your internal tools (Slack, Teams), or as a widget in a customer area. A voice version is possible when the use case calls for it.
No. Your documents are only used to answer questions, through document retrieval: they do not train the models. Depending on your constraints, the whole system can be hosted in Europe or on your servers, with GDPR taken into account from the design stage.