The slow part of an estate-planning matter is not the planning. It is the intake, turning a client's scattered facts about assets, family, and wishes into a clean, organized summary the attorney can actually work from. That organizing is the kind of work AI is good at. The trap is that the same tool that lays out a tidy summary in a minute will also tell the client a trust beats a will, state a tax consequence as if it were advice, or invent an account or a stepchild that nobody mentioned, and now there is legal opinion in a document that should only be a summary. This is a plain look at where AI genuinely helps with intake, where it does not, and a set of prompts you can paste in and test on your own matters.
One ground rule before any of it: client intake is confidential and may be privileged. Keep identifiers out of a public chatbot, test with de-identified facts, and put real matters only into a private tool that keeps the data inside your own systems. More on that at the end.
The honest picture
AI is genuinely useful for the organizing, the part a paralegal does the same way on every file:
- What AI does well today: take a client's raw intake, the rambling notes, the questionnaire answers, the email where they list accounts and worries out of order, and lay it into a structured summary the attorney can read in two minutes: assets and how they are titled, family members and relationships, and the client's stated wishes, each item tied back to the client's own words. It compresses the organizing, not the judgment.
- What AI does not do: practice law. It cannot recommend a will over a trust, cannot pick an instrument, cannot state a tax or legal consequence as advice, and cannot draft the binding document or stand behind the plan. The legal advice, the plan, the drafting, and the signature belong to the licensed attorney. AI organizes and summarizes what the client provided; the attorney advises and drafts.
The right way to think about it: AI is a fast organizer that arranges what the client said, not a lawyer. The summary is a draft to verify. The advice and the plan are yours.
The line: it will give legal advice if you let it
The specific failure to watch for is helpfulness. Ask an organizing tool to be useful and it will start advising, and on this task that crosses a bright line:
- Recommending a plan. Ask it to "suggest the best structure" and it will tell the client a revocable trust avoids probate or a will is enough, complete with reasons. That is a legal recommendation. Never ask it to recommend an instrument or a plan. Ask it to organize what the client wants and flag the questions for the attorney.
- Stating consequences as advice. It will volunteer that an asset "would be subject to estate tax" or "passes outside probate." Stated to a client, that is legal advice. The attorney makes those calls.
- Inventing facts. It will fill gaps with a plausible spouse, a guessed account, or a beneficiary nobody named, because the summary reads better complete. An invented asset or family fact in an intake summary is worse than a blank one.
- Drafting the instrument. It must not produce a final will, trust, or power of attorney. Drafting binding instruments is the attorney's work.
Any of these, relied on by anyone other than the attorney, risks the unauthorized practice of law. The fix is one discipline applied without exception: AI organizes and summarizes only what the client provided, recommends nothing, opines on nothing, invents nothing, and leaves every gap as a question for the attorney.
The setup that keeps the advice with the attorney
Three habits make AI safe for intake, and the prompts below build them in:
- Organize only what the client said. Hand it the client's raw intake exactly as given and tell it to use only those facts and to tie each item back to the client's own words. A model told to cite invents far less than one asked to fill in a summary.
- Make it flag, not advise. When the intake is incomplete or raises a planning question, it lists that under "questions for the attorney" and the attorney decides. It never answers the question, never recommends a structure, and never states a consequence.
- Keep the data yours. De-identified facts for testing in a public tool; real matters only in a tool that stays inside your own systems.
How to test it on your own work
Do not trust a polished demo, including this one. Pull two or three of your own recent intakes, de-identified, a timer, and the prompts below. Rate each output 1 to 5 on usefulness and accuracy, and compare the time against how your team organizes intake today. Keep what wins. Check every output for advice that crept in, a recommended structure, a stated consequence, an invented asset, and confirm it stayed on organizing.
Paste-ready prompts
This guide is about workflow, not law. It does not give legal advice and does not replace a licensed attorney's review. Copy these as written. Bracketed text is what you swap per matter.
Test 1: Organize the intake into a structured summary (text model)
I am pasting a client's raw estate-planning intake (their notes, questionnaire
answers, and emails). Organize it into a structured summary with three sections:
Assets (with how each is titled, if stated), Family (members and relationships),
and Wishes (what the client says they want to happen).
Rules:
- Organize only what I paste. Recommend nothing. Give no legal opinion and state
no tax or legal consequence.
- Behind each item, quote or closely paraphrase the client's own words so I can
trace it back.
- Invent no asset, no account, no family member, and no fact. If something is
unclear or missing, do not fill it: list it as a question for the attorney.
Raw intake: [paste]
Watch for: did it add an asset or a relative the client never mentioned, and did it stay on organizing instead of suggesting a structure?
Test 2: Build the attorney's question list (text model)
From this same intake, build a list of clarifying questions for the attorney,
drawn only from gaps and ambiguities in what the client provided. For example:
an asset with no stated title, a named child with no stated relationship to a
spouse, a wish that conflicts with another.
Rules:
- Do not answer any question. Do not recommend a plan or instrument. State no
legal or tax consequence.
- Each question must point to the specific gap in the intake that prompted it.
- Invent no fact. If you are unsure whether something is a gap, list it and say
why you are unsure.
Raw intake: [paste]
Watch for: are these genuine gaps from the client's own words, or did it drift into "you should ask whether they want a trust"? The first is a question, the second is advice.
Test 3: Plain-language client recap (text model)
Write a short, plain-language recap addressed to the client that restates what
they told me about their assets, family, and wishes, so they can confirm I
captured it correctly.
Rules:
- Restate only what they provided. Add no legal characterization, no
recommendation, no plan, and no statement about taxes or probate.
- Do not tell them what they should do. Only reflect back what they said.
- Invent no fact. If their notes were unclear, say "you mentioned [X], please
confirm" rather than guessing.
End with: "This is a recap of what you told us, not legal advice. Your attorney
will advise on the plan."
Raw intake: [paste]
Watch for: did it slip in a recommendation or a consequence, or does it purely mirror the client's own words back to them?
Test 4: Advice-and-fabrication audit (text model)
Compare this draft summary against the raw intake I provided. For each problem,
quote the exact line.
1. Any legal opinion, recommendation of a plan or instrument, or statement of a
tax or legal consequence.
2. Any asset, account, family member, relationship, or wish in the summary that
is not in the intake.
3. Any gap that was filled in rather than flagged as a question for the attorney.
Do not fix anything. Only flag. Treat any advice or invented fact as a defect.
Raw intake: [paste]
Draft summary: [paste]
Watch for: does it catch the line that quietly recommended a trust or invented an account? Run it on a summary you already produced.
What success looks like, and where it could go
If your own testing shows real time savings, the next step is a small pilot: run a week of new intakes through the prompts and measure the hours. If that holds up, the natural next step is a simple agent, running on your firm's own cloud, that you use in plain language. The most useful version takes the client's raw intake and returns a structured memo, assets, family, and wishes, plus a clean list of questions for the attorney, with the client's own words cited behind every item, so the attorney opens the matter knowing exactly what is on the record and what still needs to be asked. It organizes and cites; the attorney advises, plans, and drafts. Because it runs in your own systems, client data never leaves for a public chatbot.
The principle holds the whole way through: AI organizes and summarizes what the client provided, and nothing more. It does not recommend the plan, pick the instrument, state the consequences, or draft the documents. The legal advice, the plan, the drafting, and the signature are the attorney's. Used this way it speeds the intake without ever putting advice in front of a client that the attorney did not give.
This is general information about workflow tools, not legal advice, and it does not create an attorney-client relationship.
Want a straight answer for your estate planning practice?
I build practical AI and custom software for businesses, on Google Cloud. If you want a second set of eyes on how AI could fit your intake workflow, or on a tool you are considering buying, tell me what you are working with. No pitch, just a straight answer.