One assistant that advises, sells, takes appointment requests and controls your tools

A simple chatbot works with rigid menu trees and canned answers. XICBOT is trained on your business, reads your website and your shop and answers from your own knowledge base: sourced answers name their source, and an answer without a used excerpt is recorded as a gap. On top of that it acts: it shows product cards, puts together a cart in the chat, takes appointment requests and captures leads — every writing step only after the visitor clicks a proposal they have read in plain text beforehand. When it cannot answer something, it hands over to a human and passes the request on.

For selling, three functions matter most: the shop assistant for advice and availability, and, with your inventory system connected in the Custom package, also for order status, the product cards for product cards in the chat and the cart assistant for putting together and changing a cart in the chat and the way into the shop. They are part of the Pro package and work with a shop based on Shopware Community Edition. The product search knows four forms — search, compare, bundle, also bought — and returns at most six hits, exclusively from your own catalogue. The cart sits with XICBOT, not in your shop: it collects items and calculates with the current price from your shop, or, if the shop cannot be reached, with the last synced price and its date, plus the rules you deposited for shipping, threshold and voucher. “Continue” then opens your shop's stored address with a one-time pickup token that lets your shop fetch the cart from XICBOT, provided it integrates that pickup itself. Checkout and payment stay in your shop.

Both. Seventeen tool types are available: seven at the core — lead, appointment, handoff, product search, cart, an outbound call of your own and pre-filling a form on your site once the visitor confirms — plus ten adapters for merchandise management, calendar, catalogue, customer management and ticket systems; the outbound call of your own and the ten adapters come in the Custom package. The model is offered only what you have entered, switched to active and allowed for the role of the conversation; if the list is empty, it receives no tool field at all. At most five calls run per answer, the sixth is logged as capped. Anything that writes never runs unconfirmed: the proposal stands in plain text in the chat, only the click executes it, and every call leaves a log row carrying one of seven states. More under tool control and custom functions.

The assistant is bound to the deposited excerpts: sourced statements carry the number of the excerpt they came from, and figures, prices and deadlines that are not in your sources are replaced with a notice. If the search finds nothing, it usually says that the deposited knowledge base holds nothing on the subject. Like an answer without a used excerpt, the case is recorded as a knowledge gap and reported outwards, so you find it again in the analytics from the Pro package and can add the missing source. On request the assistant hands over to a human, and when a topic is delicate it is meant to, in each case with transcript and responsible place. How this binding to your own content works is described under knowledge base.

In the language it is written to in. That language is detected in our own code through function words and characteristic characters, with no external service and no extra model call; word lists exist for nine languages: German, English, French, Spanish, Italian, Dutch, Polish, Portuguese and Turkish. If a sentence is not clear-cut, the field stays empty instead of guessing German. Detection happens per sentence, which is why a switch mid-conversation is carried too. You do not have to maintain your content several times over: the knowledge base is not filtered by language, language acts as a ranking bonus. More under multilingual assistant.

There are three packages, each consisting of a one-time setup and monthly operation. Start begins at 990 € net setup and 49 € per month net and covers website and knowledge base, answers with a citation, lead capture, handover to a human and multilingual answers. Pro from 2,900 € net setup and 149 € per month net adds shop integration, product cards and cart, appointment and lead functions and the conversation analytics. Custom begins at 4,900 € net setup; operation there follows the scope and therefore has no list price. The packages also differ in seats, assistants, knowledge sources and the monthly token allowance. All figures are listed under pricing. All services can be cancelled monthly; there is no minimum term.

The frame around it is tight and anchored in the code. The assistant calls only tools you have entered and switched to active; there is no open access to your systems. Anything that writes is forced to confirmation when created or changed, no matter what the request says — without confirmation only something that reads and is explicitly set to automatic will run. Every call leaves a log row that is written before execution and knows seven states, from proposed to deadline exceeded. A one-time key made of conversation, tool and sorted input prevents double execution, and a missing required field prevents execution altogether. At most five calls run per answer. More under tool control.

Permissions are the heart of tool control. The assistant is offered only what is active and what the role of the conversation is allowed to call; a muted conversation gets no list at all. Every tool carries a kind — read, write or irreversible — and an approval mode: automatic or only after confirmation, with confirmation as the preset. This way an availability query runs without a prompt, while a cancellation runs only after explicit consent. Kind and schema come from the register and not from the request, and one answer carries at most five tool calls. Hosting is in Germany and the answers are generated by a specialised model provider, also outside the EU on the basis of standard contractual clauses; details under privacy and hosting.