Should I Hire an Estate Planning Attorney or Just Use AI?

For more than ten years, Ruth came to my office with essentially the same instructions.

She had four children. She loved them equally. When she died, she wanted them to divide her property equally.

There were other familiar things about her visits. Ruth used a powered wheelchair and sometimes bumped into the side of my doorway on her way in. I would point to the marks farther down the frame and joke that those were hers, too. She would apologize and offer to pay for the damage. I would tell her she had already paid for that doorway many times over.

When you work with someone for a decade, you accumulate these small exchanges. You learn how she jokes, how she explains herself, and which decisions she returns to without hesitation.

In 2019, Ruth arrived with her son. She was wearing an oxygen mask and struggling physically. Her son explained that she wanted to change her plan. The hundred acres would go to him. He was the one taking care of her.

Ruth nodded.

I asked him to leave so I could speak with her privately. Alone with me, she said she wanted him to have the property.

A client can change her mind. My job was to understand her decision, including a decision that differed from everything she had told me before. But this conversation left me with serious concerns about whether the change reflected Ruth's independent wishes. Those concerns were about the circumstances of the decision, not her use of a wheelchair or oxygen.

When her son returned, I told him I would not prepare the changes. He offered me $10,000. I refused again.

After they left, I reported the incident to Adult Protective Services.

Years later, when I began worrying that artificial intelligence might replace estate planning attorneys, I thought mostly about the documents we produce. Eventually, I realized I should have been thinking about that afternoon with Ruth.

The experiment I kept repeating

For years, I conducted a small, informal experiment. Every few months, I opened an AI program and gave it the same assignment: create a will for me. I supplied my information and examined the result.

I had a professional interest in the answer. I am an estate planning attorney. If a computer could do in seconds what clients paid me to do, I wanted to know.

For a long time, the results were reassuringly bad. Some drafts failed to address basic signing requirements. Under Texas Estates Code Section 251.051, an ordinary typed will generally requires at least two credible witnesses, age fourteen or older, who sign in the will-maker's presence. Those requirements matter regardless of how polished the document looks.

I would find problems, close the program, and return to work feeling that my profession had a little more time.

Then I tried Claude with a legal skill. The result surprised me. I thought it was a good document. The obvious defects I had come to expect were missing.

That required some honesty on my part. I could keep evaluating AI against the mistakes it used to make, or I could acknowledge what I was seeing. Its drafting had improved substantially.

I use AI in my practice now. It helps me edit language, review work, and consider whether I have overlooked something. Our firm still uses professional drafting software, and I remain careful about what goes into a finished estate plan. But I no longer find it useful to dismiss AI as incapable of producing good legal language.

With careful instructions and informed review, it can produce a strong draft.

For someone considering an estate plan, that creates an understandable question: if software can write the document, what am I paying an attorney to do? For a while, I had been asking myself the same thing.

Who will be able to explain your decisions when you cannot?

What a finished document cannot tell you

The difficulty is that an estate plan is unusually hard to evaluate when you buy it. You can read the pages. You can check the spelling of your children's names. You can see that the paragraph dividing your property says what you asked it to say.

What you cannot easily see is how the plan will function years later, under circumstances you have not anticipated, when the people involved may disagree about what you intended.

The finished document gives you something tangible to judge. Much of the work that makes it useful is less visible: understanding the family, identifying problems, coordinating the plan with the property, and establishing how decisions were made.

Ruth's family eventually needed that less visible work.

After her visit with her son, I received a call from her daughter. She thanked me for refusing to make the changes and explained what had been happening.

According to her daughter, the family was trying to get Ruth's son out of the house. He kept two large dogs there, and Adult Protective Services had been unable to meet with Ruth. He had also taken her to another attorney. That attorney prepared a deed transferring the property to him.

Her daughter told me he had taken more than $100,000 of Ruth's money, traveled, and left Ruth at home with the dogs. The other children eventually moved Ruth into one of their homes and helped get her affairs out of their brother's control.

Then Ruth died.

When the plan was tested

Her family now faced a problem that could not be solved simply by reading the trust I had prepared. There was another document to contend with: the deed transferring the property to her son.

I told the children to hire another attorney. I expected to be a witness.

At the hearing, I testified about the Ruth I had known, the instructions she had repeatedly given me, and what happened during that final visit.

The judge set aside the deed. The property was divided under Ruth's trust. Five million dollars went to her four children, divided equally, as she had wanted.

The trust mattered enormously. So did the family's intervention, the attorney handling the dispute, and the court's decision.

My contribution included something that had never appeared as a paragraph in Ruth's estate plan. I could describe a relationship that extended across more than ten years. I had heard her explain her wishes repeatedly. I had observed the circumstances surrounding the proposed change. I could answer questions about what I had seen and done.

That history became evidence.

When people compare an attorney with AI, they often compare the cost of obtaining a document. Ruth's experience shows why the comparison needs to include what happens after the document is signed.

The people behind the paperwork

There is another complication I had not thought much about when I began my experiments. The same technology that helps someone draft an estate plan can help someone else examine it for weaknesses.

A disappointed family member can upload a document, ask questions about it, and request possible arguments against it. Whether those arguments have merit is a separate issue. But the ability to generate them is becoming easier to obtain. As these tools improve, I expect people to use them on both sides of estate disputes.

That changes how I think about the work. Better drafting tools are valuable. They also make it more important to understand the circumstances behind a document and to preserve useful evidence of the client's decisions.

An attorney can meet privately with a client. Ask why a longstanding plan has changed. Notice who answers the questions. Explore whether the client understands the proposed decision. Supervise the signing and document the process.

Sometimes that work leads to a better paragraph. Sometimes it leads to another conversation. In Ruth's case, it led me to refuse the assignment.

None of this makes an estate plan immune from challenge. Hiring an attorney cannot guarantee family harmony or a particular court result. It can, however, help create a plan supported by careful advice, an appropriate signing process, and people who can explain what happened.

AI can be useful as you prepare for that process. It can help organize your questions, clarify unfamiliar terms, and put your initial wishes into words. I welcome clients who have thought carefully about what they want and arrive ready to discuss it.

But producing a convincing draft does not establish that the plan fits your circumstances or will accomplish your goals.

The more difficult questions often concern the people around the document. Who depends on you? Who controls access to you? Who might object? What happens if your health changes? Who will be able to explain your decisions when you cannot? Those questions were present in Ruth's life long before anyone stood in court.

I once worried that my value as an estate planning attorney would disappear when a computer learned to write a good trust. Ruth's case helped me understand that value more fully.

Her family needed the trust. They also needed someone who remembered the woman who had made it: the woman who came back every few years, offered to repair my doorway, and told me she loved all four children equally.

When the time came to explain her wishes, I had more than a document to offer. I had been there.