The AI Assistant for Law Firms: Sort Enquiries, Not Advise

No. Legal advice is hard-blocked in the assistant's rule set, alongside medical information, the assessment of allergies and a binding financing offer. It gives organisational information from your own content — areas of law, responsibilities, process, required documents, deadlines as far as you have stored them — and sorts the enquiry. A legal assessment, a view on prospects of success or a recommendation to act are not part of it. Sourced answers name their source; if nothing is in the sections found, it says exactly that instead of composing something. For the next step the lead tool type puts the enquiry in front of a qualified professional. More under assistant for law firms.

It asks follow-up questions tailored to your firm: area of law, a short factual description of the matter, urgency and, for the conflict check, the parties involved. Which fields it collects you define on the tool; a call with missing required fields is not executed at all. What results lands as an enquiry in the mailbox of the business, with name, address, phone and matter, and files uploaded beforehand are attached. A one-time key made from conversation, tool and input prevents a second click from creating the same enquiry twice. The result is an ordered initial enquiry, similar to the lead assistant but tailored to a law firm.

Both in the same conversation, with the appointment starting out as a request. The appointment tool type takes appointment requests, reschedules and cancels — all three acts live in that one tool, and the kinds of appointment come from your own list. Because that writes, the proposal stands in plain text in the chat beforehand, and only the visitor's click creates the request; it becomes an appointment once you confirm it in your view, and if the visitor gave an email address, they then receive a confirmation email. On request the assistant names free times from your stored kinds of appointment, minus the appointments already requested or confirmed in XICBOT; a time it names remains a suggestion, and a binding entry straight into your calendar only happens with a calendar connection set up in the Custom package. The document checklist comes from your own templates: held as a maintained question-and-answer pair in the knowledge base, it goes out verbatim, without a model call — what it says is what is sent. An hourly run sends the reminder for the confirmed appointment by email, if an address is on file. More at the booking assistant.

XICBOT is built to be GDPR-compliant: hosting in Germany, a data processing agreement, data sovereignty and a deletion concept; the answers are generated by a specialised model provider, also outside the EU on the basis of standard contractual clauses. Data is only captured when it is provided in the conversation and only handed to the destinations you have defined. Inside your account a fixed ladder of roles applies: whoever invites cannot grant a higher role than their own. When our support accesses your account, it runs through a separate session that expires after four hours and leaves a log entry, without your password and without bypassing your second factor. Details under privacy and hosting.

Yes. As a responsibility signpost, the assistant routes by area of law, department or location to the right person. From your content it knows who is responsible for what and hands the enquiry over accordingly — and it is instructed to answer exclusively from your knowledge base, so that no wrong responsibility is invented; an answer without a used citation is recorded as a gap. For the handover the conversation is marked as handed over, the request becomes an enquiry, and your team is notified. For recurring organisational questions the assistant can additionally be set up as a support assistant.

A law firm assistant with initial-enquiry qualification, appointment function, document checklist and conversation analytics is part of the Pro package from 149 € per month net, plus a one-time setup from 2,900 € net. The conversation history is shown to you masked; the plain text is seen only by someone who holds the role for it, states a reason and accepts the line in the audit log. A plain signpost with FAQ and document hints is often already possible with the Start package from 49 € per month net and 990 € net setup. All details are listed under pricing. All services can be cancelled monthly; there is no minimum term.

The room for that is drawn tight. For every question at most five search hits from your knowledge base, plus your text blocks, go to the model, numbered, and sourced answers name the excerpts they rely on. Figures, prices and deadlines that are not in your sources are replaced with a notice. If the search finds nothing or the model declines, the assistant says that nothing on this is held in the knowledge base — and the case is logged as a knowledge gap, as is an answer without a used excerpt. Its rules forbid it to invent contact details, file numbers or appointments, or to pose as a human being. An error cannot be ruled out entirely with any language model; hence the binding to sources, the visible citation and the handover to a person.

For each answer the conversation run fetches five hits from your knowledge base and builds from them the excerpts the assistant may rely on. If none of the hits carries the question, it is instructed not to guess but to say that the deposited knowledge base holds nothing on the subject. In parallel the question is recorded as a knowledge gap; from the Pro package the list of knowledge gaps shows up to 50 cases and is the working stock for your next round of maintenance. On request the assistant hands over to a human: state change, mailbox entry and notification happen in one go, with the transcript attached. An open question therefore stays visible instead of disappearing into the conversation. More at the support assistant.

The assistant spots urgency in two places: the mood of the conversation and its topic. Eight word lists run behind this — among them emergency, health, allergies and complaint — deliberately without a language model, so that the handover still works when the model does not answer. Three points trigger it; asking for a human reaches the threshold on its own. If anything points to a danger to life or health, the assistant is instructed to refer to the emergency services before anything else, and that rule cannot be switched off; if the message contains an emergency word such as unconscious or heart attack, the service appends a fixed emergency notice to the answer, in German conversations with 112 for the emergency services and 116117 for the medical on-call service. The assistant is not meant to judge the case itself: the conversation moves into your mailbox with its transcript and the responsible place already assigned.

Yes. We conclude a data processing agreement with you under Art. 28 GDPR. You remain the controller under the GDPR, and we process the data solely on your instructions for the agreed purpose. It governs purpose limitation, technical and organisational measures as well as deletion and data minimisation, so you can cleanly add the assistant to your record of processing activities. Those measures are not a declaration of intent: tenant separation is enforced in every database query, not in the individual controller, the conversation history is displayed masked, and a plain-text retrieval requires the matching role, a stated reason and leaves a log line behind.