It’s getting easier to drop a new inquiry into an AI tool or a marketing workflow and ask for a campaign idea, a case-study outline, or a follow-up sequence. The copy that comes back can sound polished. That isn’t the risk that should worry you first.
The risk is what the drafting step received. Names, matter facts, conflict notes, medical or financial particulars, and a prospective client’s story can move from an intake form into a marketing brief or a prompt without anyone meaning to publish them.
This is a marketing-operations issue, not legal advice. I’m not a lawyer, a security auditor, or a custodian of a firm’s client files.
Intake work and marketing work are not the same job
To respond to an inquiry, your team often needs a name, contact details, conflict information, and a working account of the matter. Keep that in the intake or matter system, limited to the people who need it.
To plan a public campaign, you usually need something else: a practice area, an audience, a goal, an offer, and proof points the firm has already approved for public use. You do not need last week’s consult notes.
Those two jobs often live in the same CRM, spreadsheet, or shared drive. That’s why “just paste the intake notes into the brief” becomes the path of least resistance.
LawFuel’s seven automation tips for law firms already cover useful work on the public side of the funnel: lead capture, drips, reminders, and reviews. This piece looks at a narrower question: what those workflows are allowed to receive.
The confidentiality duty is broader than a privileged memo
Under ABA Model Rule 1.6, a lawyer shall not reveal information relating to the representation of a client unless the client gives informed consent, the disclosure is impliedly authorized in order to carry out the representation, or a Rule 1.6(b) exception applies.
That duty covers all information relating to the representation, whatever its source, not only secrets the client labeled confidential. Rule 1.6(c) also requires reasonable efforts to prevent inadvertent or unauthorized disclosure of, or unauthorized access to, that information.
Intake adds another wrinkle. Many inquiries come from people who are not yet clients. ABA Model Rule 1.18 treats a person who consults about forming a client-lawyer relationship as a prospective client, and it generally bars a lawyer from using or revealing information learned in that consultation even if no engagement follows.
ABA Formal Opinion 512, issued in July 2024, applies those duties to generative AI tools. Before lawyers input information relating to a representation into a GAI tool, they must evaluate the risk that it will be disclosed to or accessed by others inside or outside the firm. For many of today’s self-learning tools, the opinion says a client’s informed consent is required first.
Generic boiler-plate in an engagement letter is not enough.
The Model Rules are models, and other privacy laws may also apply, so get advice for your firm. Treat this as general information. If a fact came from intake, it does not belong in a campaign brief or a prompt unless counsel has approved that use.
Give campaign briefs a small, approved input set
Don’t scrub a live intake record. Start with a template that accepts only approved fields:
- The practice area, described in general terms
- The intended audience
- The campaign goal
- The offer or call to action
- Proof points already approved for public use
- Constraints: claims to avoid, jurisdictions, and tone
Leave out names, contact details, matter narratives, conflict notes, opposing-party names, medical or financial particulars, quotes from an intake form, and “we just signed someone who…” details, even if you plan to anonymize them later.
Build field-level separation into the workflow. Intake fields stay in the intake or matter system. The marketing brief is a different object, filled from the template, not copied from a consult note. If your tools can sync data, sync only the fields on the approved list, or don’t sync at all.
Redaction after the fact is a weak substitute, because a rushed find-and-replace can still leave identifying facts.
Use synthetic examples when you develop prompts
When you are testing a prompt or explaining a process, use invented or composite examples. The comments to Rule 1.6 allow a hypothetical only if there is no reasonable likelihood that the listener can ascertain the identity of the client or the situation involved. Dropping a last name from a real intake note usually fails that test.
Keep a small library of approved sample briefs and prompts, so no one has to improvise with a live file. Asking an AI tool to “turn this into a case study” does not help if the input still contains information relating to a representation or a consultation.
Control access, check vendor settings, and review the brief itself
A clean template still fails if raw intake notes sit in a shared folder, or if the drafting tool trains on prompts. Before you adopt a tool, ask:
- Does the vendor use prompts or uploads to train its models?
- Who can see workspace history, shared chats, or exported logs?
- How long is data retained, and can it be deleted?
- Where is it stored, and which firm users have access?
Limit intake access to the people who need it to evaluate or handle the inquiry. Marketing should work from the approved brief, not the underlying file. The same caution applies when intake fields go to advertising or measurement tools, as LawFuel notes in its piece on auditing enhanced conversions before sending intake data to Google.
Put a human review boundary in two places. First, review the brief before it goes into any AI tool or marketing platform. Second, review the output before anything is published or scheduled.
Keep a record of the source brief: the version used, the date, who approved the inputs, and who approved the output. Formal Opinion 512 also reminds lawyers that supervisory duties apply to staff and to outside vendors.
A short checklist you can apply this week
- Separate intake fields from marketing-brief fields in your systems.
- Publish one approved brief template and require it for campaign and content requests.
- Keep client or prospect identifiers, conflict information, and matter facts out of prompts, briefs, and marketing workspaces.
- Use synthetic examples only when developing or demonstrating prompts.
- Review each vendor’s training, retention, access, and deletion settings, and turn off training on inputs where you can.
- Restrict intake-file access to the people who need it.
- Require a human to approve the brief before it goes into a tool, and the draft before it goes public.
- Store the approved source brief with the resulting draft.
- Brief the whole path, including intake coordinators and lawyers who might paste a note into a chatbot.
Pick one recurring campaign type this week, build the template, and audit the last few briefs against it. The goal is to let automation draft from inputs that were safe to share, so review is about quality, not about hunting for a name that should never have been in the brief.
I used AI assistance in researching and drafting this article and reviewed, edited, and fact-checked every line myself.
Author:
David Oscar is Editor at Vivid Algo, where he writes about AI tools, automation and marketing for small businesses. He has worked in content, marketing and software since 2014.His contact details are at – Vivid Algo | LinkedIn





