Drafting Client Retainer Agreements With AI: Is What NJ Attorneys Are Producing Actually Competent Under RPC 1.1?

There's a workflow spreading quietly through NJ solo practices: attorney opens ChatGPT or a legal-specific tool, types something like "draft a retainer agreement for a New Jersey personal injury matter," and has a usable-looking document in sixty seconds. Then the attorney skims it, maybe edits a paragraph, and sends it to the client.

That workflow has a problem. And it's not the one most attorneys expect.

The hallucination risk is real, but it's actually the secondary issue here. The primary issue is that a retainer agreement produced this way may be technically coherent and still fail the competence standard under RPC 1.1, which requires not just knowledge of the law but thoroughness and preparation reasonably necessary for the representation. For a document that sets the terms of the entire attorney-client relationship, "reasonably necessary" covers a lot of ground.

What AI Gets Right, and Where It Stops

General-purpose AI tools are decent at producing boilerplate retainer language. Fee structure, scope of representation, termination rights, file retention, basic confidentiality language, most tools handle these adequately when prompted well. If you're starting from a blank page, AI gets you to a first draft faster than any template library you've maintained since 2017.

But NJ-specific retainer requirements are where things get uneven, and attorneys often don't realize how much the tool missed until something goes wrong.

A few specific gaps worth knowing:

Fee agreements in contingency matters. Under RPC 1.5(c) and New Jersey Court Rule 1:21-7, contingency fee agreements in NJ have specific formal requirements, the written agreement must be signed before or at the time the attorney commences representation, it must disclose the contingency percentage and how expenses are handled, and certain case types (like medical malpractice) have statutory fee caps that must be reflected. AI tools regularly produce contingency language that is plausible but not compliant with the NJ court rule structure. The output looks fine. It's not.

Arbitration clauses. Some AI-drafted retainers include broad arbitration provisions for fee disputes. In New Jersey, fee arbitration under R. 1:20A is a client right, not just a contractual option, and retainer language that obscures or contradicts that right creates exposure. AI doesn't flag this tension because it doesn't know you're in New Jersey unless you tell it explicitly, and even then it may not apply the right framework.

Scope limitations in multi-matter representations. When a small firm represents a client on both a business dispute and a related employment claim, the retainer needs to address scope carefully, including conflict waivers, billing allocation, and whether a conflict in one matter affects the other. AI produces a single-matter template by default. Attorneys often just send it without adapting it.

The Competence Question Isn't Just About Errors

RPC 1.1 doesn't only kick in when a document contains a wrong statement of law. It also applies to the process. The NJ Supreme Court's general interpretation of competence, consistent with the ABA Comment to Model Rule 1.1, includes the requirement that the attorney apply the legal knowledge, skill, and preparation that the matter requires.

Sending an AI-generated retainer you haven't genuinely reviewed for NJ-specific compliance is arguably a competence issue even if no client ever complains. The standard is what a reasonably competent NJ attorney should have done, not whether the client noticed a gap.

This matters practically because retainer agreements are the document most likely to be scrutinized if a fee dispute goes to the District Ethics Committee. A committee reviewing a complaint about an attorney's fees will look at the retainer. If that retainer has generic language that doesn't match NJ-specific requirements, the attorney is now defending both the fee and the document.

A Practical Review Checklist for AI-Drafted Retainers

Before any AI-drafted retainer agreement goes to a client, work through these specific questions:

1. Does the fee structure comply with RPC 1.5 and, if contingency-based, with NJ Court Rule 1:21-7 including applicable fee caps?
2. Does the agreement accurately describe the scope of representation with enough specificity that a client could reasonably understand what's included and what isn't?
3. Does the arbitration or dispute resolution clause, if any, accurately reflect the client's rights under NJ's fee arbitration program rather than contracting around it?
4. If the matter involves a vulnerable client (elder law, guardianship, certain family matters), does the agreement account for NJ ethics opinions addressing those representations?
5. Is the termination and withdrawal clause consistent with RPC 1.16's requirements, including the attorney's obligation to take steps to protect the client's interests?
6. Does the document reflect your actual billing practices, not generic billing language the AI assumed?

That last one catches attorneys more than any other. AI will often draft "billed monthly in arrears" or "retainer replenished when depleted to 25%" as defaults. If that's not actually what your firm does, the agreement now conflicts with your real practice, and that conflict becomes ammunition in a fee dispute.

The Better Use of AI Here

AI should be the drafting engine, not the compliance engine. Use it to generate structure and first-pass language, then run your own review against your NJ-specific checklist. If you handle retainer agreements regularly across a few practice areas, build practice-area-specific prompts that include the NJ rule citations you actually need reflected. A well-constructed prompt takes thirty minutes to build and saves you the compliance review gap on every subsequent draft.

The competence obligation under RPC 1.1 doesn't disappear because a tool did the first draft. It just shifts entirely to the review step. Make that step real, not a skim.