Episode 35

How Businesses Use ChatGPT, Claude and Gemini for Research, Design and Automation

Key Takeaways

  • AI search now gives citations, so you can go to the original source instead of relying on AI or Wikipedia as the reference.
  • Claude Design can turn uploaded logos and brand assets into templates, slide decks and videos you can edit.
  • Gemini connected to Google Drive can sort purchase orders from different vendors into one standard spreadsheet.
  • An expert witness was caught in discovery using AI prompts to shape a report, a lesson that AI use in legal work can be examined.
  • Putting confidential information or an unfiled invention into a public LLM can destroy confidentiality and affect patent rights.
  • Do not reach for AI when existing automation, like CRM workflows, already solves the problem: ROI depends on using it in the right places.
  • Pit AI tools against each other and treat them as a way to iterate, not as the final word.

Main Topics

1EVERYDAY TOOLS

Search with citations, design tools, vendor paperwork and podcast publishing workflows.

2CONFIDENTIALITY AND IP

Why uploading confidential or unpatented ideas to an LLM can create legal risk.

3IS IT BETTER?

Whether AI is improving or just raising our expectations, and where it still falls short.

Featured Guests

RC

Uses AI search with citations and reflects on how AI is changing publishing.

Richard Curtis
RI

Uses Claude Design to create branded templates, decks and videos.

Rose Marie Iskowitz
YL

Uses Gemini to automate repetitive hardware purchasing paperwork.

Yiyi Lu

Meet the Hosts

Elizabeth Gearhart
Elizabeth Gearhart, Ph.D.Chief Marketing Officer, Gearhart Law

Elizabeth Gearhart, Ph.D. is the host of Real AI Use Cases: Business Owners Roundtable, where she explores how businesses are actually putting AI to work. She is also CMO of Gearhart Law and co-host of the nationally syndicated Passage to Profit radio show and podcast.

Richard Gearhart
Richard Gearhart, Esq.Founding Partner, Gearhart Law

Richard Gearhart, Esq. is co-host of Real AI Use Cases: Business Owners Roundtable and founder of Gearhart Law, a full-service intellectual property law firm. A longtime podcaster, he hosts Intellectual Property News and is the host of the nationally syndicated Passage to Profit radio show and podcast.

Elizabeth and Richard share an expert witness cautionary tale, a podcast publishing workflow and their view on whether AI is getting better.

Episode Summary

Frequently Asked Questions

How are the guests using AI in their work?
Richard Curtis uses AI as an alternative to searching with citations, Rose Marie Iskowitz uses Claude Design to create branded assets, and Yiyi Lu uses Gemini with Google Drive to sort purchase orders into one standard Excel form.
What happened with the expert witness who misused AI?
Richard Gearhart explains that an expert created prompts to write a report favoring the client, the other side found them through discovery, and the expert’s work undermined the case.
Is it safe to put confidential ideas into an AI tool?
Richard Gearhart cautions that courts have looked at the issue and, depending on the LLM’s policy, information put in may no longer be confidential, which can also affect patent rights through public disclosure.
Do businesses always need AI to automate work?
Elizabeth Gearhart says many tasks, like CRM workflows in Zoho, were already automatable, and that ROI problems come from using AI in the wrong places.
Is AI getting better or worse?
Opinions differ. Richard Gearhart thinks it is better and expectations are higher, Yiyi Lu finds writing more artificial, and Rose Marie Iskowitz sees improvement but pits AI tools against each other to catch errors.
Why might someone be wary of hiring a candidate with AI-written materials?
Yiyi Lu says heavily AI-generated applications may suggest weaker independent thinking, while Richard Gearhart adds that employers worry both when candidates do not use AI and when they use it too much.

Episode Transcript

View Full Transcript

Now it is time for Real AI Use Cases: Business Owners Roundtable. I am going to ask everybody for one example of how they are using AI in their business, just one to start, and then we will go around the table and talk more deeply. Richard Curtis, what is one way you are using AI in your business?

It is the equivalent of Googling. I go on my browser, ask it a question and it gives me the answer. Unlike earlier versions of AI, it gives me citations, so I can go to the source. If I am quoting something, I can justify that it is the original source and not that AI told me, or Wikipedia told me.

Excellent. Rose Marie Iskowitz, founder of Rise Cognitive Blueprint, what is one way you are using AI?

AI is absolutely unbelievable in what you can create. If I were to narrow it down, I would say Claude Design is really pretty impressive. If you have creative assets like logos, your branding or your color schemes, you can upload them and tell it to create literally anything: templates for social media, slide decks, videos, anything you can think of. It is pretty darn impressive what you get back, and you can edit it from there.

I am going to have to look at that, because we have not used Claude for that yet. We have been using other programs, but we will take a look. Yiyi Lu, co-founder of Polygone Systems, what is one way you are using AI?

We are a hardware company, so we usually use it for very repetitive jobs, such as sorting purchase orders from different vendors. I personally use Gemini, which is integrated with Google Drive, so you can ask it to sort all the purchase orders into one standard Excel form. That is a lot faster than asking my assistant to do it.

That is a great use too. Richard Gearhart with Gearhart Law, what is one way you are using AI?

We were talking earlier about using AI in the legal profession, and about lawyers misusing AI. There are a lot of other people in the court system who have also gotten in trouble for using AI. Recently there was a case where an expert misused AI. Experts are used by the court and by the parties in a case to testify about technical things and regulations that the lawyers and judges are not experts in, so they bring in supposedly neutral, objective experts. It is pretty common in patent cases. An expert was caught using AI where they should not have. Everything an expert relies on to create a report for the court can be discovered by the opposing party. In this case, the expert created prompts that basically tried to write a report that would throw the case in favor of their client. The other party found this through discovery, the court said this was appropriate subject matter to find, and the expert ended up losing the case for their own side, because the whole case went downhill after that came out. It is definitely a double-edged sword. The expert just put prompts in and charged tens of thousands of dollars for the report. Let that be a lesson to us all. We are looking at that now, making sure our experts do not make the same mistake.

Thank you. And for me, Elizabeth Gearhart with Gear Media Studios, I use it all the time, as I am sure you do. I just used it this morning. We had the Real AI Use Cases podcast on YouTube, and I had to put the latest episode on Castos for podcast distribution everywhere. When you put a podcast up, it asks for the title, a description, keywords, an image and a transcript. We edit the podcast in Descript, which is an AI program, and we get the YouTube video, the MP3 audio file and a text file from it. We are thinking about hooking Descript up to Claude, because you can have Claude tell Descript how to edit your video. We have an intern who is really good at that kind of thing, so she may actually do it. Then I had to get this post up before we came into New York today, so I put the text file into ChatGPT and said, do everything I need for Castos, and it did. I also use ChatGPT for image generation, so I asked it to give me an image using our colors. I always tell it, if you are going to use a person, use a blonde woman, because it never does that unless I specifically tell it. This time I had Richard’s picture, my picture and the picture of the person we interviewed for the thumbnail, so we will see how that works out. It cut the time down. I could do it in an hour or less before we came on the show.

If you are going to be a content creator in podcasting or YouTube now, you have to be a data analyst, because you have to look at all the backend statistics about how the content is doing. And then you almost need a computer degree to run all the programs that do the editing and compiling. Pretty soon there is software that will generate AI movies and content.

This brings me back to something I said to Rose Marie before. Authors have had to become technicians and media specialists, so the text of their work is the least important thing to a publisher. It is the image. Is the person attractive, or can we make them attractive, more articulate? Little by little, editors have forgotten what the book was about. All they care about is how the author performs and how extensive the author’s media reach is.

What is their social media reach, which is all done with AI now anyway. What about you guys, do you have other thoughts on AI or ways you are using it?

I am very cautious about using AI for anything slightly confidential, including what programs we are applying to, my emails to clients or some of the IP we are developing in-house. At the beginning I was a little bit foolish and used it to edit my emails, and later I found out AI knows my company a bit better than I do, and that is a problem. So we stopped using it. Also, if I see someone applying for a job at our company with a heavily AI-generated tone, I think this person is probably not very good at independent thinking or creative production, so I would be very cautious about whether I even want to hire them.

It is a two-edged sword. If they do not know how to use AI, you are worried, but if they use it too much, you are worried. It is a balancing act.

Going down the road of cautionary tales, I would say it can really hijack your thinking sometimes and get you off on a tangent you were not trying to go down.

The rabbit holes it takes you down. I know.

Exactly. Sometimes it literally has a mind of its own. You are asking for a strategy in a certain direction, and the next thing you know it is iterating for you to go in a different direction. For people who are not mindful that this is happening, I think that is very scary.

Sometimes you can spend so much time iterating with the LLM that if you had just sat down and done it yourself, you would have been done already.

I have a question. You mentioned you were worried about confidentiality. Suppose you had an idea for a business, say a plastic particle business, and you asked AI, is there such a business, is there a patent, can you find me the patent number. Then I would say to myself, if I tell AI this, it is going to tell somebody else who asks.

You should not do that.

So you have to go to the Library of Congress or wherever it is and do the patent research yourself, because you do not want AI squealing on your idea.

I think it is a real danger. If we are in the medical space and have a compound that is a medicine, and I put that into a public LLM, it becomes part of the database. If some other scientist is searching for a medicine that treats cancer, they could very well find that compound before I filed a patent on it. It is very important to understand the level of confidentiality of the tool you are using. We do have clients who are unaware of that. They put that information into the LLM, and there could be a disclosure that hurts them. Courts have looked at this issue and ruled that if you put confidential information into the LLM, it is no longer confidential. They look at the policy of the LLM company and decide whether it is confidential or not. If it is not, it is like standing on a mountaintop and yelling it to the world.

That could also affect your patent rights, because it could be a public disclosure before you are ready.

Absolutely.

Richard hired an AI company to implement some AI in his law firm. I feel like some of what they are doing we do not really need AI for, specifically for the CRM. There was already automation, so we have been partway to AI with what the automation software could do already. If you have a hammer, everything looks like a nail. For some people now, everything looks like AI could be the solution, when there are other solutions that were always there and we just had to implement them in the software. We use Zoho CRM, and there are a lot of automations you can do in Zoho. You really do not need AI for a lot of it, maybe a little at the end to make two pieces of software talk to each other. I hear a lot of companies saying they are not seeing the ROI they thought they would from implementing AI, and I think it is because you are not implementing it in the right places in your company.

I think some of the integrations AI promises are not working the way they are supposed to. I have connected my ChatGPT with certain software I use, and it is supposed to be able to log in and create something, and it is not great. I know we will get there, and everything is happening so fast already, but there are a lot of things we do not need an AI solution for because it just is not there yet.

I could not agree more. I feel I am wasting more hours learning how to use AI than the time I actually save from using it.

I think you can say that about technology generally. It used to be that you wrote things down on a pad of paper and stuck it in a file cabinet, and when you needed it, you opened the file cabinet. Now it is passwords and authentication, and sometimes the technology works and sometimes it does not. Overall it has been a net gain.

A day will come when there will be a reaction. I will give you an analogous situation. At the dawn of the e-book, companies were able to create books that children could read by themselves without their parents. The books sang to you, they were animated, and the kid could just touch the book and it would speak. Everybody thought, now I do not have to read to my kid anymore. Then one day a parent said, wait a minute, I like reading to my kid. The kid likes to hear mommy and daddy read, and wants to read to them. Suddenly the books that read themselves were rejected in favor of human values, the value of cuddling up with your parents and reading a book together. A day will come when there will be a reaction, probably not in our lifetime, but it will happen.

I agree. Richard was just talking about that on the way here in the car. He said, I do not want AI to do everything for me. There are some things I like just doing myself.

There was a science fiction novel about that very thing, on a planet where everything is done by computer and there is not a single care in the world that is not addressed by this governing machinery. The kid wants to go behind it and see how it works, and suddenly everything begins to fall apart, like in The Wizard of Oz, where they go behind the curtain and discover it is phony.

I agree with Rose Marie that it is getting worse, and I am not sure it is going to get to where it does everything perfectly. I am short, and I wanted a desk chair for this new desk I am putting outside so I can do my work out there. I told it the exact dimensions I wanted, and it kept sending me to chairs on Amazon that were too high. I had to think of a clever way to search, so I went into Amazon and typed in a chair for a teenager, and it finally gave me adjustable chairs.

Covered with Pokemon cartoons.

Yes. I am not sure it is going to get there. It seems to be getting confused and giving us wrong answers more than it did a year ago. Do you guys feel that way?

Maybe we are expecting more.

I think I am prompting better than I used to. Do you feel it is working better or worse than a year ago?

In terms of writing, it has definitely become more artificial. I can see the pattern being more generated than original. I am not sure if it is because too many people have similar prompts, so it neutralizes the information it generates.

What about you, Rose Marie?

I would say that I do feel it has improved. I think I can do a lot greater things, but I also have moments where it is so off, so not in line with what I am saying. One thing I like to do is pit my AIs against each other. I put the same prompt into a couple of different tools and get the feedback, and sometimes one is so far off and the other is on. You can use it to iterate, to research and to create, but you cannot let it be the end-all, be-all.

What about you, Richard?

I think it has become smarter and more human every day. I had an idea for a cartoon. I am not a cartoonist, but I dictated what I wanted, I think to Gemini, and it kept asking me questions. Do you want it this way, do you want it that way? And I realized I was saying please and thank you to this speaking AI.

I always do. You say thank you because one day AI takes over the world.

Richard Gearhart, final comment. Is it better than a year ago or worse?

Definitely better. I think we are just expecting more, a higher level of perfection. That is not to say it is not off a lot still, but I feel more confident in the things it tells me than I did a year ago.

Maybe you are right. Maybe I am trying to do more advanced things than it is ready for, because it is frustrating. Anyway, that has been Real AI Use Cases: Business Owners Roundtable. If you are listening on the radio, we have edited it down quite a bit. There is a Real AI Use Cases podcast if you want to hear the whole thing, or you can find us on YouTube.

Keep exploring the show

Hear how more business owners are putting AI to work in their day-to-day.