AI-Generated Wills and Estate Documents in a NJ Solo Practice: A Case Study in What Goes Wrong

A solo estate planning attorney in New Jersey gets a new client: retired, two adult children, a modest brokerage account, and a house held jointly with a spouse. Straightforward stuff. The attorney uses an AI drafting tool to generate a will, a durable power of attorney, and an advance directive. The documents look clean. The attorney reviews them, makes a few edits, and sends them to the client for execution.

Six months later, the client's spouse dies unexpectedly. The house passes by operation of law, but the brokerage account, not titled in trust, no beneficiary designation updated after the intake, is subject to probate. The will the AI generated used boilerplate residuary language that didn't account for the actual asset structure the attorney had noted in intake. Nobody caught it.

This kind of failure isn't dramatic. There's no smoking gun, no obviously bad output. It's the product of a workflow where AI did most of the drafting, the attorney reviewed for correctness but not for fit, and the gap between the two went unnoticed until it mattered.

Where AI Document Automation Actually Earns Its Keep

AI drafting tools have a real, defensible use in estate planning for NJ solo attorneys. They're good at generating first drafts of standard instruments, populating client-specific variables into tested language, and flagging provisions that may need attention. If you're doing volume work, simple wills, healthcare proxies, basic POAs, the time savings are real.

The problem isn't the tool. It's the assumption that a document that looks correct is a document that works for this client.

What RPC 1.1 Competence Actually Requires Here

New Jersey RPC 1.1 requires that a lawyer provide competent representation, which means "the legal knowledge, skill, thoroughness and preparation reasonably necessary for the representation." The ABA's Formal Opinion 512 (2023) extended this to AI use explicitly: attorneys must understand enough about how their AI tools work to identify when the output is unreliable.

But competence in document automation isn't just about catching hallucinations or bad case citations. In transactional and estate work, the more common failure mode is a document that is legally accurate in isolation but wrong for the client's specific facts. An AI tool trained on standard estate planning forms doesn't know that your client's brokerage account has no TOD designation. It doesn't know the house was already jointly titled. It generates what it was designed to generate: a technically sound document that fits the average fact pattern.

The attorney's competence obligation, under RPC 1.1, is to close that gap. That means the review step can't just be "does this say what I intended to draft." It has to be "does this actually accomplish what this client needs given what I know about their assets, family situation, and goals."

A Practical Intake-to-Execution Workflow for NJ Solo Estate Practitioners

Here's what a tighter workflow looks like in practice.

At intake: Build a structured asset inventory into your intake process. Don't rely on a client questionnaire alone. Specifically ask about beneficiary designations on retirement accounts and brokerage accounts, how real property is titled, whether there are prior estate planning documents, and whether any assets are already in trust. AI tools can help you generate this intake form, but the data collection itself needs to be thorough before any drafting begins.

Before you run the AI draft: Feed the tool the actual client facts, not just the standard variables (name, date, executor). If your tool allows custom prompts or intake fields, use them. Specificity at the input stage reduces the gap between what the AI generates and what the client actually needs.

During review: Compare the dispositive provisions in the will and beneficiary structure of non-probate assets against each other. This is the step most attorneys skip when they're moving fast. A will that leaves everything to the surviving spouse is fine, unless most of the estate passes outside the will anyway, and the residuary clause is doing almost nothing.

Before execution: Run a plain-language summary of the estate plan for the client. AI tools are actually useful here. Have the tool generate a summary of what each document does, then review that summary for accuracy. If the summary reveals a gap, you'll catch it before the documents are signed.

After execution: Build a follow-up trigger into your calendar, not just a "call us if anything changes" disclaimer. Life events, deaths in the family, new accounts, property purchases, create exactly the kind of asset-structure changes that make yesterday's clean estate plan a problem tomorrow.

The Supervision Layer That's Easy to Skip

If a paralegal or legal assistant is doing any part of the intake data-entry or the AI drafting setup in your solo practice, NJ RPC 5.3 requires that you have adequate supervision in place. That means the attorney, not the staff member, is the one doing the dispositive-provisions review described above. This isn't just an ethical formality. In estate work specifically, the risk that a non-lawyer will miss a structural mismatch between the will and the client's asset profile is high, because that mismatch requires legal judgment to spot.

The practical fix is a two-step sign-off: staff handles intake data and document generation, attorney handles the structural review against the client's actual asset picture before anything goes to the client.

If your AI-generated estate plan is going to hold up, that review step is where it happens. Build it into your process before a client's family finds out it was missing.