Firm-wide AI or a service for one matter: how to document the classification
Is an AI tool part of the firm's general infrastructure, or does it serve one specific matter? How law firms can document that classification for counsel, with examples and a template outline. Not legal advice.
The German rules on attorney service providers (BRAO §43e) draw a line that matters for every AI tool: is the service used for the firm as a whole, or does it serve one specific matter? Where it serves a single matter directly, the rules contain an additional requirement involving the client's consent.
Many firms never write down which side of that line a tool sits on. This piece shows how to document the classification so that counsel can decide it.
Not legal advice. Which side of the line a given use falls on, and what follows from it, is for counsel to decide. This method only prepares the facts. For the overall decision chain, start with how German attorney service-provider rules apply to AI tools in a law firm.
Why the classification belongs on paper
The same tool can sit on either side. A drafting assistant available to everyone in the firm looks like infrastructure. The same assistant set up for one client's due diligence, with that client's documents loaded, looks like a service for one matter. The difference lies in how the tool is used, not in the product name.
If the classification lives only in someone's head, it cannot be reviewed, and it will drift as usage changes.
Step 1: Describe the use, not the tool
Write one entry per use. "General research assistant for all lawyers" and "document review for one transaction" are two entries, even if the software is the same.
Step 2: Record who set it up and for whom
- Was it procured or configured for the firm generally, or for a particular client or matter?
- Who has access: everyone, a practice group, or only the matter team?
- Is the client named in the setup, the workspace, the contract or the invoice?
Step 3: Record what data it touches
Note whether the use works with general material (templates, public sources, internal know-how) or with documents and facts from a particular matter. If it receives matter data, note whether client and matter details are removed first. The method is in how to remove client and matter details from AI prompts.
Step 4: Record how the tool is connected
A tool connected to the firm's software may reach many matters at once, or be limited to one. Note the scope. Day 1 of this series covers the setup: how to connect AI to your law firm software through an interface.
Step 5: Record where the data goes
Attach or reference the map of providers, locations and logs from how to map where an AI tool's prompts and logs actually go.
Step 6: Leave the classification field for counsel
The entry should end with an empty field: "Classification (decided by counsel)", plus the name, date and any conditions. The firm describes; counsel classifies.
Illustrations, not answers
These are generic illustrations of how facts differ. Each one could be classified differently depending on the details, and none of them is a conclusion.
- Firm-wide knowledge search over the firm's own templates and know-how, open to all lawyers, no client documents loaded. Facts that point toward general use.
- Review workspace for one transaction, set up for a single client, loaded with that client's data room, access limited to the deal team. Facts that point toward a service for one matter.
- Firm-wide drafting assistant used on one file: general setup, but a lawyer pastes facts from a specific matter. Facts on both sides; this is where documentation matters most.
Template outline
- Use (one sentence)
- Tool and contracting provider (internal record only)
- Set up for: firm / practice group / named matter
- Access: who
- Data: general material / matter data, anonymised or not
- Connection scope to firm software
- Provider chain map (reference)
- Settings in use
- Open questions
- Classification, decided by counsel: name, date, conditions
- Next review date and trigger (for example, a change in use)
FAQ
Is a firm-wide tool automatically outside the consent requirement? Not automatically. How it is actually used can change the picture, which is why each use gets its own entry. Counsel decides. Not legal advice.
Who should fill in the template? The person responsible for the use, usually with IT. The classification field is left for counsel.
When should the entry be revisited? Whenever the use changes: a new matter, new data, wider access, a new provider in the chain.
How we build this at AETOP
We build agents on Spectacle on the same principle: a model gets only the objects a task needs, numbers are calculated in code, the model handles language, every request is logged, and a human releases before anything changes. Scope is part of the setup, not a later discovery. The aim is time back for the people who still do this coordination by hand, without taking the decision away from them. Approval-gated actions are on our roadmap and not yet bookable; we do not describe them as live before they are. If you want to talk through your own setup: see Spectacle.
Not legal advice. This piece describes operational practice and does not replace a legal review.