The Truth About AI-Generated Client Communications and NJ RPC 1.4

Most conversations about AI and legal ethics orbit a familiar cluster of concerns: confidentiality, hallucinated citations, billing transparency. Those are real. But there's a quieter compliance gap spreading through solo and small-firm practices in New Jersey, and it lives inside something attorneys do dozens of times a week: communicating with clients.

RPC 1.4 requires New Jersey attorneys to keep clients reasonably informed about the status of their matter, promptly comply with reasonable requests for information, and explain things well enough that the client can make informed decisions. It sounds basic. In practice, when AI tools enter the communication workflow, "basic" gets complicated fast.

How AI Slips Into Client Communication (and Why It Feels Fine at First)

The starting point is usually efficiency. A solo handling a residential real estate closing, a simple estate, or a personal injury file gets a client email asking for a status update. The attorney has 40 other things open. They open ChatGPT or their AI-assisted practice management tool, type a quick prompt, and send a lightly edited response in 90 seconds instead of 10 minutes.

That works fine, until it doesn't.

The failure mode isn't usually a hallucinated fact. It's subtler: the AI drafts a response that is technically accurate but pitched at the wrong level of specificity for that client's actual concern. The client asked whether their offer was accepted. The AI-drafted response explains the general contract execution process. It reads well. The attorney, skimming before sending, misses that the actual question went unanswered. The client is left without the information they needed to decide whether to call their lender.

That's an RPC 1.4 problem. Not a dramatic one, not a grievance-level event on its own, but a pattern of this kind of drift is exactly what the NJ Office of Attorney Ethics has historically flagged in communication-related complaints.

Where the Specific Risks Stack Up

Intake and initial consultations. AI-generated intake acknowledgments that over-promise on response timelines, or that describe the scope of representation in vague terms, can create misaligned expectations before a formal engagement letter even goes out. NJ courts have found implied attorney-client relationships in far less.

Settlement communications. Drafting a settlement summary for a client is one of the highest-stakes communication tasks an attorney handles. An AI tool that softens or generalizes the actual terms, even slightly, can leave a client unable to make a genuinely informed decision about whether to accept. That's the exact function RPC 1.4(b) is designed to protect.

Status update emails on contingency matters. In PI and workers' comp practices, where clients often have little legal literacy and high anxiety, a generic AI-drafted "your matter is progressing" update fails the client even if it's technically inoffensive. "Reasonably informed" under RPC 1.4 isn't a floor, it's calibrated to what this client, with this matter, actually needs to know right now.

Automated follow-ups via CRM integrations. Several practice management platforms now let attorneys automate client-facing touchpoints. When those touchpoints are AI-generated and fire without attorney review, the attorney has effectively delegated a core RPC 1.4 duty to software. NJ RPC 5.1 supervisory responsibility may catch some of this, but the predicate failure is still the communication gap.

What a Disciplined Workflow Actually Looks Like

The fix here isn't "don't use AI for client communications." It's building a review checkpoint that's substantive, not reflexive.

Before sending any AI-drafted client communication, an attorney should be able to answer two questions from the client's perspective: what did they specifically ask or need to know, and does this response address that specifically? If the answer to the second question requires the attorney to re-read the draft carefully, that's the checkpoint working. If the attorney's review is just a grammar scan, the checkpoint is decorative.

For automated outreach through CRM or practice management tools, the attorney should audit the message templates themselves, not just the individual sends. Templates that were AI-generated and approved once, months ago, have a way of drifting out of sync with how the firm's matters actually develop.

One practical step: create a short internal category for client messages your AI tool should never draft without a full attorney compose, not just a review. Settlement communications, adverse updates, and anything involving a deadline the client must act on should be on that list.

NJ's RPC 1.4 comments don't mention AI, and the ACPE hasn't issued specific guidance on AI-drafted communications yet. That silence isn't permission. It's a gap that individual attorneys fill with either disciplined practice or eventual exposure.

Start with the gap.