Seven new inquiries by morning — and the lawyer is already behind

You open your inbox. Seven inquiries came in overnight. Two are about tenancy law — not your focus. One wants a “quick, free” opinion on whether she has a case. Three contain nothing you can work with. And one — the one that actually fits — has been waiting for a reply since last night.

By the time you’ve sorted them, politely declined the mismatches, and called back the promising ones, the morning is gone. That’s not an exception. That’s daily life in many solo and small law firms.

At a glance (TL;DR)

An AI assistant on the firm’s website checks, in a conversation, whether the matter fits the firm’s practice area, captures the issue and its urgency in a structured way, and answers organizational questions around the clock. The firm only receives pre-qualified inquiries. Legal advice stays exclusively with the lawyer — that’s not optional, it’s required by professional rules.

What an unstructured first inquiry really costs

The answer isn’t a fee — it’s your time. And in a solo or small firm, lawyer time is the scarcest resource there is.

A significant share of first inquiries simply doesn’t fit the firm’s practice area. Another share expects a free assessment — not an engagement. A third contains so little information that a callback is needed before the basics are even clear.

The problem isn’t the inquiry itself. The problem is that you only find this out after you’ve read, phoned, and replied. Every mismatched inquiry you weed out is, in hindsight, lost billable time.

How an AI assistant solves this problem

The assistant sits on your website and guides visitors through a structured conversation — before you even know about it.

The crucial part: the assistant makes no legal assessment of anything. It captures and sorts. Whether a matter falls within the firm’s practice area is a categorization question — not a legal one. Whether someone needs help urgently or has time is an organizational question. What documents to bring is in your FAQ.

The assistant doesn’t decide whether someone has a case. It decides whether the inquiry fits the firm — and paves the way for the first consultation.

What reaches the firm is no longer a raw incoming email. It’s a qualified summary: practice area, the matter in bullet points, urgency, contact details. The first consultation doesn’t start from zero.

What the AI assistant actually does — and what it doesn’t

This isn’t an academic distinction. It protects you from a professional-conduct problem and gives the assistant its proper place.

Task	Who handles it?
Check the practice area: does this fit the firm?	AI assistant
Capture the matter and its urgency in a structured way	AI assistant
Answer FAQ about process, appointments, required documents	AI assistant
Politely refer mismatched inquiries elsewhere	AI assistant
Hand over a qualified summary to the firm	AI assistant
Legal assessment of the individual case	Lawyer only
Evaluating the chances of success	Lawyer only
Checking for conflicts of interest	Lawyer only
Accepting an engagement	Lawyer only
Judging deadlines and statutes of limitations	Lawyer only

This separation isn’t a convenience — it’s a legal necessity.

Which regulatory requirements apply

Four sets of rules matter here — under German and EU law, since this article is about firms operating in that market. All four have practical consequences for how the assistant is built.

Section 2(1) RDG (Germany’s Legal Services Act) — legal advice is reserved for lawyers

Any legal analysis of an individual case counts as a legal service — and that’s reserved for lawyers. The chat assistant on your website is permissible as long as it doesn’t assess, only captures and sorts. That sounds simple, but in the conversation design it’s the critical point: any wording that sounds like an assessment must be avoided. This gets hard-coded during setup. When in doubt, the assistant refers to you — instead of answering itself.

Section 43a(2) BRAO (Germany’s Federal Lawyers’ Act) and Section 203 of the German Criminal Code — confidentiality

Confidential information from visitors stays confidential — even when it’s collected in a chat. As soon as inputs are transmitted to an external AI provider, Section 43e BRAO also applies: the provider must be bound to confidentiality in writing, a no-training clause must be agreed contractually, and a data processing agreement (DPA) under Article 28 of the GDPR (the EU’s data-protection law) is mandatory. A DPA alone doesn’t satisfy the professional rules — Section 43e BRAO sets stricter requirements than the GDPR.

Section 43a(4) BRAO — conflicts of interest

The chat assistant doesn’t accept engagements. It also doesn’t tell the inquirer that there’s no conflict of interest — because it can’t know that. That check remains exclusively with the lawyer, before accepting the engagement. That’s why the assistant collects data sparingly: name, contact details, practice area, a brief description of the matter — no complete account of the facts.

GDPR Art. 13, Art. 28, Art. 9 and EU AI Act Art. 50 — transparency and data protection

When the chat starts, visitors need to know two things: that they’re talking to an AI system (AI Act Art. 50, binding from August 2026, but it should already be standard today) and how their data is processed (GDPR Art. 13 — purpose, legal basis, retention period, recipients, data-subject rights). A notice at the start of the chat and a link to the privacy policy are the minimum. Special attention goes to GDPR Art. 9: information about health, criminal proceedings, or family circumstances is sensitive data — storing and transmitting it requires an explicitly established legal basis.

You’ll find more on the general GDPR requirements for AI tools in the post Data privacy with AI tools: what small businesses need to know.

What matters in the build — and where it goes wrong

An AI assistant on a law firm’s website isn’t an off-the-shelf product you just switch on. Four points decide whether the setup runs reliably or creates problems:

Separate capturing from advising clearly, from the start. The conversation design must explicitly define which phrasings the assistant uses — and which it doesn’t. Any answer that sounds like a legal assessment is a problem: under professional rules and in terms of liability. Whoever thinks this through rigorously during setup has peace of mind afterward.
Calibrate the practice-area filter realistically. Too strict, and suitable clients get turned away. Too broad, and the inbox fills up again with fringe topics. Edge cases — say, the employment lawyer who occasionally touches social-security law — get defined together before go-live.
Collect minimal data until the conflict check is possible. Before an engagement is accepted, you don’t know whether a conflict of interest exists. So the assistant captures only what’s necessary: name, contact details, practice area, a brief description of the matter. No complete account of the facts — that protects everyone.
Keep the callback option visible at all times. Many clients prefer talking to a person right away — especially in emotionally difficult situations like divorce, termination, or criminal proceedings. The assistant must never be an obstacle. The phone number and a callback option have to be reachable in the chat at any point.

The setup takes several weeks: defining practice areas and boundary questions, conversation flow and escalation rules, privacy-compliant configuration including the DPA under Section 43e BRAO, clarifying third-country data transfers and retention periods. The knowledge base — how the initial consultation works, required documents, FAQ — gets built cleanly once. After that, the assistant runs on its own.

The effort pays off when lawyer time no longer flows into inquiries that will never become an engagement.

Frequently asked questions

Is a chat assistant on a law firm’s website even respectable?

What looks unprofessional is an inquiry that sits unanswered for three days. An assistant that takes structured intake around the clock, says honestly when a matter doesn’t fit the practice area, and immediately names the next step comes across as professional — provided it’s built cleanly and consistently stays out of legal assessments.

What happens to an inquiry that doesn’t fit the firm’s practice area?

The assistant points out, in a friendly way, that the matter isn’t part of the firm’s focus and — where possible — refers to suitable places to turn. That’s not brushing people off; it’s honest service. The inquirer doesn’t lose time on a first consultation that would never lead to an engagement anyway.

Can the assistant give no legal information at all?

Organizational information — how the initial consultation works, which documents to bring, how to reach the firm — is no problem. It becomes critical the moment the assistant assesses the individual case: whether someone has a claim, what the chances of success are, or whether a deadline is running — it must not answer those. This boundary has to be hard-built into the conversation design — not as a note in the documentation, but as a technical barrier in the conversation flow.

How do visitors learn they’re talking to an AI?

From August 2026, this is a binding obligation under EU AI Act Art. 50 — but it should already be standard today. A clear notice at the start of the chat — something like “You’re talking to the firm’s AI assistant” — meets the requirement. At the same time, a path to a real conversation must be visible at all times.

Can the assistant work outside office hours too?

Yes — that’s one of the practical advantages. Someone who wants to describe their situation in the evening after a workday finds a point of contact. The qualified summary is then waiting in the firm’s inbox the next morning — structured, complete, ready to act on.

If you have more fundamental questions about whether chatbots suit your context, the post Chatbots in customer support: when are they really worth it? gives an honest assessment.

My takeaway

The benefit is clear: lawyer time only flows into inquiries that fit the practice area and are pre-qualified. That’s not a technology promise — it’s a structural outcome when the assistant is built cleanly.

The regulation isn’t an obstacle here; it’s a frame that defines the field of use precisely. The RDG, Sections 43a and 43e BRAO, the GDPR, and the AI Act together say: capturing and sorting is allowed — assessing and advising is not. Whoever implements that rigorously during setup is on safe ground.

Who is this worth it for? Firms with a clearly defined practice area, a visible online presence, and a regular stream of unstructured first inquiries. For generalists without a focus, or for firms that receive hardly any online inquiries, the setup effort probably isn’t justified.

If this sounds like your firm

If you run a solo or small firm and regularly receive first inquiries that don’t fit your practice area or come in unstructured — and you’d like only pre-qualified inquiries to reach you — then drop me a short note:

What practice area do you work in?
How do first inquiries currently reach you — phone, contact form, email?
What costs you the most time in the process?

I’ll get back to you with a concrete assessment of whether and how an AI assistant can be put to sensible use in your firm.

Email: marketing@gudrun-ponta.de Phone: +49 176 21110218