Supervising a Non-Lawyer Who Uses AI for NJ Client Intake: What RPC 5.1 Actually Puts on You

Solo and small-firm attorneys in New Jersey have been quick to see the efficiency case for AI-assisted client intake. A chatbot collects the prospective client's basic facts, flags obvious conflicts, and routes the inquiry before you've even poured your morning coffee. The problem isn't the tool. The problem is the assumption that because a non-lawyer ran the process, the supervising attorney's ethical exposure is somehow reduced.

It isn't. And RPC 5.1 is the reason.

What RPC 5.1 Actually Covers (and Why It Applies Here)

Most NJ practitioners associate RPC 5.1 with supervising associates and partner-level oversight. The rule is broader than that framing suggests. It requires that a lawyer with direct supervisory authority over another lawyer ensure that the other lawyer complies with the Rules of Professional Conduct. But RPC 5.3, its companion rule for non-lawyers, applies when a receptionist or intake coordinator is doing the work. Read together, these two rules create a firm-wide obligation: the supervising attorney is responsible for the systems their office uses, including AI-assisted workflows that touch prospective clients.

If an AI intake tool gives a prospective client information that constitutes legal advice, or if it fails to flag a conflict that a reasonable conflicts check would have caught, the attorney of record doesn't get to point at the software. The question the NJ Office of Attorney Ethics would ask is whether the lawyer had a system in place that was reasonably designed to catch that outcome. An undocumented AI chatbot with default vendor settings is not a system. It's a liability.

The Intake Stage Is Where the Risk Is Underestimated

Here's what makes AI-assisted intake particularly tricky: the intake stage sits right at the edge of the attorney-client relationship. NJ courts have recognized that a prospective client can acquire confidentiality rights before a representation is formally established. That means the data your intake chatbot collects, the conflicts questions it asks, and the information it volunteers about how your firm handles cases all carry legal consequences before you've signed a retainer.

Some AI intake tools are designed to be helpful. Overly helpful. They'll answer follow-up questions from a prospective client ("Can I sue for that?"), suggest case viability ("This sounds like a strong personal injury claim"), or volunteer information about statutes of limitations. If a non-lawyer at your front desk said those things on a call, you'd retrain them immediately. When software says it, attorneys often assume it's the vendor's problem. RPC 5.1 and 5.3 say otherwise.

What Supervision of an AI Intake Workflow Actually Looks Like

Supervision here isn't reviewing every chat transcript after the fact. That's quality control, and it's better than nothing, but it's not a compliance framework. Genuine supervision under the NJ RPCs means the attorney established the parameters of the AI tool before it went live, and those parameters were designed with the ethics rules in mind.

Practically, that means a few specific things:

First, the intake chatbot's script or prompt configuration should be reviewed and approved by the supervising attorney, not just the vendor's customer success team. If the tool allows customization of what it says to prospective clients, that customization is a legal document and should be treated accordingly.

Second, the chatbot should be configured to avoid volunteering legal conclusions. Its job is data collection and routing, not case assessment. Any question the tool can't answer without crossing into advice should trigger a handoff to the attorney.

Third, your conflicts check can't be fully delegated to the AI. Tools that advertise AI-powered conflicts screening are a useful first pass, but the attorney needs to do a confirmatory review before the intake is treated as cleared. The software doesn't know about the matter you discussed informally with a potential adverse party six months ago.

Fourth, document what the AI tool is configured to do. If you're ever in front of the OAE explaining your intake process, "I relied on the vendor's default settings" is a worse answer than "Here is the intake policy we implemented, and here is how we trained our staff to use the tool within those parameters."

The Vendor Agreement Gap

One more thing most NJ solo practitioners overlook when they subscribe to an AI intake platform: the vendor's terms almost certainly disclaim liability for any legal advice the tool gives. That's standard. But it means the indemnification you might assume exists doesn't. If a prospective client later claims they relied on something your intake chatbot told them, the loss lands on the firm.

Review those terms before your next intake tool renewal. Specifically look at what the vendor claims the tool is designed to do (information collection vs. advice) and whether their terms conflict with how you're actually using it in practice. If they do, you have a gap that needs closing before someone else finds it for you.

The AI part of this is new. The supervision obligation isn't.