AI Is as Smart as Its User

How I used AI to prepare for professional advice—and learned that better questions still need verified answers.

Many of us remember looking things up in an encyclopedia before we could type a question into a search bar. We learned new tools along the way. AI is another opportunity to learn, and we do not have to approach it empty-handed.

Our experience counts. So does our willingness to question an answer that sounds a little too neat.

The title is my reminder to stay involved. Good instructions help, and even a well-directed tool can make mistakes. That is where our judgment comes in.

Researchers studying several language models found that small changes in wording and formatting could change their responses. That is a reason to be clear about the assignment, not a promise that the perfect prompt makes an answer true. Read the 2024 POSIX prompt-sensitivity study.

Learning to work with AI

My understanding developed through AI for Business and AI Prompting courses included in my MBA certificate program at Augment Entrepreneur School. The courses gave me a foundation. Practice taught me how much my own participation mattered.

I learned to explain the job: what I wanted to accomplish, what information mattered, how I wanted the findings organized, and what needed checking. A broad question became a focused assignment.

The most useful lesson was simple: getting an answer is not the same as establishing that it is correct. NIST identifies confident but false output as a generative-AI risk, often called a hallucination or confabulation. Fluency is not proof. Read NIST's Generative AI Profile, section 2.2.

When the work became personal

AI became especially useful to me while I was researching intellectual property for my startup.

My patent attorney charged approximately $450 an hour. As a bootstrapped founder, I wanted to arrive prepared enough to use that time well. I spent nearly a year learning patent terminology and organizing preliminary research with AI's help.

I directed it toward official USPTO materials and used it to help explain unfamiliar terms, classifications, prior art, and questions I needed to take to counsel. I wanted to understand the conversation I was paying to have.

A preliminary search has limits. The USPTO explains that an examiner may find relevant material that the applicant's search missed. It also recommends a registered patent practitioner because preparing and prosecuting an application requires legal, procedural, and technical knowledge. Read the USPTO's guidance on preliminary searches and professional assistance.

AI helped me prepare. It did not establish patentability or replace qualified legal counsel.

I asked it to challenge me

I did not need a machine to applaud my idea. I needed help finding the questions I had overlooked.

I asked it to consider the analytical perspective of a patent examiner. That instruction did not make it an examiner. It gave the research an organizing approach.

My assignments included:

• Find information that might contradict my assumptions.

• Identify potentially relevant prior art and explain why it might matter.

• Separate a source's actual statement from an interpretation.

• Flag missing information and questions that require legal review.

• Provide the underlying sources so I can read them myself.

This helped me organize difficult material and prepare more focused questions. The responsibility to check the results stayed with me.

AI researched. I evaluated. My attorney advised.

Those roles mattered. The USPTO’s official AI guidance warns about inaccurate submissions and confidentiality risks. Using AI does not transfer responsibility away from the people submitting the work.

In my experience, doing this preparation reduced the paid research time I needed and helped me focus the conversation.

It also clarified what I needed from counsel. I felt that the research I had assembled was not receiving the substantive review I wanted. I was not asking my attorney to trust AI. I wanted the underlying sources examined, the mistakes identified, and professional judgment applied.

Eventually, I ended that attorney-client relationship. AI had not removed my need for a lawyer. The experience helped me recognize what I needed from the relationship: careful evaluation, clear explanations, and mutual respect.

The lesson beyond patents

For me, the value was becoming a better-informed participant. That principle reaches beyond this particular project.

Try a modest first assignment using information that is already public. Ask for a plain-English explanation, a comparison, or questions to discuss with a professional. Then read the sources and decide what holds up.

Before entering confidential material, examine the service's privacy terms, retention practices, and relevant settings. The USPTO specifically warns that entering invention details into an AI service may disclose sensitive information to third parties. Telling a tool to keep a secret is not a substitute for understanding the service. See the confidentiality section of the USPTO guidance.

Bring your experience with you

I think many Gen X readers have more of a starting point than they give themselves credit for. We have already learned unfamiliar tools, sorted good advice from bad, and asked people to explain what they mean.

Whatever your starting point, being a beginner is allowed. Your questions and experience belong in the process.

You can be curious without handing over your judgment. Start with something useful, check what comes back, and build from there.

That is the spirit behind the title: making informed decisions with another tool at your disposal.

Author's note

This is a personal account of AI-assisted preparation, with linked background sources. It is not legal advice. Patent matters should be evaluated with an appropriately qualified professional.

Related reading: The Career Pivot After 50.