Off-the-shelf or custom chatbot (RAG): what is the difference?

An off-the-shelf chatbot answers from the model's knowledge and a few settings — convenient, but it may make things up or stay vague about your specifics. A custom RAG chatbot (Retrieval-Augmented Generation) first looks for the answer in your own documents, then writes from those excerpts. The result: answers grounded in your content, with sources, and less drift. Custom requires a clean knowledge base, but that's what sets an assistant that truly speaks about your company apart from a one-size-fits-all chatbot.

Can an AI chatbot hallucinate (make up answers)?

Yes, it's a real risk with any generative system. The best safeguard is grounding: a RAG chatbot only answers based on excerpts retrieved from your documents, and can say "I don't know" or hand off to a human when information is missing. You reduce the risk further with clear rules, displayed sources and supervision. You never eliminate the risk entirely, but you make it manageable — a point we frame from the design stage.

Is an AI chatbot GDPR-compliant?

It depends on the tool and the configuration. The sensitive points: where conversations are hosted, what is sent to the model, retention duration and user notice. Some solutions host outside the EU. For sensitive data, we favour European hosting, limit the data transmitted and document the processing. This is a point we audit systematically before any deployment.

How much does an AI chatbot for business cost?

There is no single price. Off-the-shelf solutions often charge by usage or per agent, while a custom chatbot involves a design investment plus operating costs (model, hosting). The ballpark depends mostly on conversation volume, the level of grounding expected and the integrations. The most honest approach is to start from your real use cases: we then estimate a realistic range rather than a headline price.

Best AI chatbots for business in 2026

The real question

Do you need an AI chatbot?

Yes, if you receive repetitive questions and your answers are documented somewhere. A well-designed AI chatbot absorbs first-level requests, answers outside business hours and saves your team time. But a poorly grounded chatbot frustrates more than it helps: the real question isn't « do we need a chatbot » but « what will it rely on to answer ».

What an AI chatbot really brings to a business:

Answering continuously to frequent questions, without tying up a human every time.
Qualifying and routing requests to the right team or the right person.
Relieving support so your team can focus on the complex cases.

Overview

The comparison in one table

Solution	Ideal for	Strength	Watch out for
Custom chatbot (RAG on your knowledge base)	Answering from your real content, without generic replies	Grounded in your documents, tone and in-house rules	Requires a clean, up-to-date knowledge base
Intercom Fin	Customer support on an established SaaS product	Built into the Intercom support suite	Usage-based pricing to model upfront
Zendesk AI	Teams already on Zendesk	Native AI in a mature helpdesk	Value tied to the Zendesk ecosystem
Voiceflow	Designing and prototyping conversational journeys	Visual, collaborative flow editor	Still a builder — data grounding takes work
Botpress	Technical teams wanting control	Open source, deeply customizable	Requires skills to do it well
Crisp / French solutions	French SMBs, support + on-site chat	European vendor, accessible pricing	AI features to compare by plan
Voice chatbot	Phone reception, appointment booking	Natural voice on the phone channel	Latency and human handoffs to frame

Indicative comparison — check up-to-date pricing, AI features and hosting with each vendor. Lumyniq, 2026.

Detail

The solutions, by family

Grounded custom (RAG). A chatbot built on your knowledge base looks for the answer in your own documents before writing — that's what avoids generic replies and makes it possible to display sources. See our guide on RAG for business.

Integrated support modules. Intercom Fin and Zendesk AI are the shortest path if you're already in their ecosystem: the AI plugs into your existing helpdesk.

Conversational builders. Voiceflow (visual flow editor) and Botpress (open source, highly customizable) to design journeys; grounding on your data still has to be worked on.

French and SMB solutions. Crisp and other European vendors cover on-site chat and support well for SMBs, with accessible pricing.

Voice chatbot. For phone reception and appointment booking, a voice agent can answer in natural speech — provided you frame latency and handoffs to a human.

Decision

How to choose your chatbot

Start from your real customer questions — list the 20 most frequent requests before choosing a tool.
Check the source of the answers — off-the-shelf or grounded in your documents? That's what drives accuracy.
Look at integration — the chatbot must plug into your website, your CRM and your support.
Control GDPR — where conversations are hosted, data sent to the model, retention.
Plan the handoff to a human — a good chatbot knows how to say « I'm handing over ».

Our role

Integrating a chatbot that truly speaks about you

The classic trap is the chatbot deployed too quickly, which answers off the mark and disappoints. At Lumyniq, we start from your knowledge base to design a grounded assistant (RAG), plugged into your tools via the Claude engine, with safeguards and a handoff to a human — rather than a generic chatbot dropped onto a page.

Want to go further? Discover our custom AI agents and how we secure these deployments (AI agent security).

FAQ

Frequently asked questions — AI chatbots for business

It depends on your channel and your maturity. If you're already on Zendesk or Intercom, their AI modules (Fin, Zendesk AI) are the shortest path. For a French SMB just getting started, a solution like Crisp is accessible. But as soon as answers need to match your products, pricing and procedures precisely, a custom chatbot grounded in your knowledge base (RAG) avoids generic replies. The right choice is decided by starting from your real customer questions, not from the best-known brand.

Best AI chatbots for business in 2026

Do you need an AI chatbot?

The comparison in one table

The solutions, by family

How to choose your chatbot

Integrating a chatbot that truly speaks about you

Frequently asked questions — AI chatbots for business

Read next

Sources

Let's talk about your project

Contact details