AI Is as Smart as Its User

How I used AI to research intellectual property, prepare for legal counsel, and become a better-informed founder—without confusing the tool with the expert.

Generation X has already lived through more than one technological revolution.

We researched with encyclopedias before we searched with Google. We remember when computers were separate from everyday life, when the internet was optional, and when nobody carried the collective knowledge of the world in a pocket.

Now we are being told that artificial intelligence will change everything.

Maybe it will. But before AI can change anything for us, we need to understand what it is—and what it is not.

I believe AI is as smart as its user.

Not literally. AI has capabilities and limitations that do not change simply because we become better users. But the usefulness of what it gives us depends heavily on the direction, context, judgment, and scrutiny we bring to the conversation.

Research has found that large language models can produce significantly different results in response to relatively small changes in prompt wording, format, and examples. That makes the quality of the user’s instructions consequential, although even a well-designed prompt cannot guarantee accuracy. A 2024 study published by the Association for Computational Linguistics examined this sensitivity to variations in prompts.

If I use AI like Google, it responds like a more conversational search engine. If I give it a clear assignment, direct it toward reliable sources, ask it to challenge my thinking, and show it how I learn, it becomes something much more useful.

It becomes a working partner.

Learning How to Work With AI

My understanding of AI 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, but learning to use AI effectively required practice.

I had to spend time explaining what I was trying to accomplish, what information mattered, how I processed new material, and how I wanted the findings presented. I learned to narrow broad questions into focused assignments.

Most importantly, I learned that receiving an answer is not the same as establishing that the answer is correct.

Generative AI can produce false or internally inconsistent information, misrepresent sources, miss important context, combine unrelated facts, and present uncertain conclusions confidently. The National Institute of Standards and Technology identifies confidently stated false or erroneous content—sometimes described as “confabulations” or hallucinations—as a significant generative-AI risk. NIST recommends appropriate human review, oversight, testing, and documentation. Read NIST’s Generative Artificial Intelligence Risk Management Profile.

AI-generated work has to be questioned and verified.

That is not a weakness we can afford to ignore. It is part of the user’s responsibility.

When AI Became More Than a Search Engine

The moment AI became truly valuable to me was during my intellectual-property research.

As the founder of a bootstrapped startup, I had to do much of the preliminary work myself. My patent attorney charged approximately $450 an hour. I did not want to spend those hours delivering incomplete information, correcting avoidable misunderstandings, or asking questions I could have researched beforehand.

I wanted to use my attorney’s time for the work that genuinely required professional legal expertise.

So I spent nearly a year learning about patents and developing a structured research process with AI.

I directed my AI to study official United States Patent and Trademark Office materials and other reliable patent-law resources. I used it to help me understand terminology, processes, patent classifications, prior-art research, novelty questions, claim concepts, and the legal framework surrounding patent protection.

The USPTO explains that an examiner’s prior-art search ordinarily includes textual and classification searching across patents and other published documents. The agency also encourages inventors to conduct preliminary searches using its patent databases. Review the USPTO’s explanation of patent searching and prior art.

I preliminary search, whether conducted by a person or assisted by AI, cannot establish patentability or guarantee that every relevant piece of prior art has been found.

Patent examination involves much more than locating similar ideas. A USPTO examiner reviews the application and the claimed invention for compliance with applicable legal requirements. Preparing and prosecuting a patent application requires knowledge of patent law, USPTO rules and procedures, and the scientific or technical field involved.

The USPTO allows inventors to represent themselves, but it recommends using a registered patent attorney or patent agent because an applicant may not obtain the full scope of available protection without experienced professional assistance. Learn more from the USPTO about applying for patents and working with registered patent practitioners.

That professional expertise remained important to me. My goal was to arrive prepared enough to use it intelligently.

Teaching AI to Challenge Me

I did not ask my AI to tell me that my assumptions were correct.

I prompted it to approach the research from the analytical perspective of a USPTO patent examiner. That did not transform the AI into an examiner, attorney, patent agent, or legal authority. It simply gave the research a more disciplined point of view.

I instructed it to:

• Search for information that might contradict my assumptions.

• Help identify potentially relevant prior art.

• Examine novelty questions and possible similarities.

• Explain patent classifications and unfamiliar terminology.

• Identify inconsistencies or missing information.

• Separate established facts from interpretations.

• Flag questions requiring professional legal review.

• Cite the underlying sources so I could verify the work.

I wanted the AI to find weaknesses—not validate what I hoped was true.

That approach helped me organize my research, understand difficult material, recognize better questions, and reduce possible errors in the information I presented to my attorney.

AI Researched. I Evaluated. My Attorney Advised.

Those roles mattered.

AI was a research tool. I remained responsible for evaluating and verifying its findings. My attorney remained responsible for providing professional legal advice.

The USPTO has issued formal guidance addressing the use of AI-based tools in matters before the agency. The guidance recognizes that AI may be used while preparing patent and trademark applications and other submissions. However, using AI does not displace the responsibilities imposed on applicants, attorneys, agents, and other parties.

The guidance warns about risks that include inaccurate or misleading AI-generated material, disclosure of confidential information, improper signatures, and reliance on unverified facts or legal authorities. Read the USPTO’s official guidance on using AI-based tools in practice before the agency.

That aligned with the process I was developing. I did not automatically accept what AI produced. I examined the sources, questioned its conclusions, and continued refining the assignment when an answer was incomplete.

In my experience, this preparation reduced the number of paid research hours I needed from my patent attorney. It allowed me to reserve more of that expensive professional time for legal analysis and decisions.

But it also revealed a problem in the relationship.

When Preparation Is Not Taken Seriously

My attorney did not identify a specific error that invalidated the entire body of research. Instead, I felt that the work I had spent nearly a year developing was not being taken seriously.

I was not asking him to trust AI.

I was asking him to evaluate the research, examine the underlying sources, identify what was accurate or inaccurate, and apply the professional judgment I was paying for.

I welcomed scrutiny. In fact, I had deliberately instructed my AI to challenge my assumptions. What I could not accept was having disciplined, source-based preparation dismissed without meaningful engagement.

Eventually, I ended the attorney-client relationship.

AI had not replaced my need for qualified legal counsel. It helped me understand what I needed from that professional relationship: critical thinking, mutual respect, careful evaluation, and a willingness to examine the evidence.

The Lesson Was Bigger Than Patent Research

This experience changed how I think about AI.

AI can help a bootstrapped founder reduce the cost of becoming informed. It can translate unfamiliar language, organize complex research, identify gaps, explore opposing arguments, and prepare us for more productive conversations with experts.

But it cannot assume our responsibility.

We still have to:

• Choose trustworthy sources.

• Protect confidential information.

• Verify important claims.

• Recognize when we have reached the limits of our knowledge.

• Consult qualified professionals when professional judgment is required.

• Make the final decision ourselves.

Protecting confidential information requires more than telling an AI system to keep something private. Before submitting sensitive business, invention, legal, medical, or financial information, users should understand how the particular service stores, retains, uses, and potentially discloses submitted material.

Lawyers have separate professional obligations when using generative AI, including duties involving competence, confidentiality, client communication, supervision, verification, candor, and reasonable fees. The American Bar Association addresses these obligations in Formal Opinion 512.

For inventors, confidentiality is especially important. Public disclosure of invention details can affect patent rights, and the privacy protections, data-use terms, and retention practices of AI services are not all the same.

The goal is not to remove human expertise. The goal is to become a more knowledgeable participant in decisions affecting our businesses, finances, health, and future.

Why Generation X Is Ready for This

Generation X may be better prepared for AI than we realize.

That is my assessment, not a claim that every member of our generation has the same experience or ability.

We remember doing research without instant answers. Many of us learned to question sources before social media turned every opinion into apparent expertise. We entered adulthood before technology began anticipating our every move—and then adapted to each new wave anyway.

We know how to learn independently.

We know how to be skeptical without becoming paralyzed.

And many of us have reached a stage of life where we are building businesses, changing careers, protecting ideas, and creating second acts without unlimited time or money.

AI can help—but only if we stop treating it like magic.

It is not an oracle. It is not automatically an expert. It is not a replacement for our judgment.

It is a powerful tool that becomes more useful when we give it a defined role, reliable information, meaningful context, and clear standards for its work.

That is what I mean when I say AI is as smart as its user.

The real advantage does not belong to the person who asks AI the most questions. It belongs to the person who learns how to ask better questions, challenge the answers, verify the evidence, and decide what to do with the knowledge that comes back.

Author’s Note

This article describes my personal experience using AI-assisted research. It is not legal advice. AI-generated legal or patent information should be independently verified, and inventors should consider consulting a registered patent attorney or patent agent.

The USPTO also provides an Inventors Assistance Center, a Pro Se Assistance Program, and a qualifying Patent Pro Bono Program.