Episode 32

The AI Coworker Is Here: Here’s What It Can Do

Key Takeaways

  • An AI call center can phone 150 local offices overnight and return the few that match your needs, saving half a day of calls.
  • Small teams without legal, finance or marketing experts can use AI to build dashboards and tools they would once have needed engineers for.
  • Having one AI check another’s work, such as Claude reviewing ChatGPT’s output and the reverse, can improve results.
  • AI tools often flatter you: instruct them to disagree and tell the truth, and have someone else run visibility queries about your business.
  • You can set different tones for an AI depending on the situation, from direct and factual to reflective or coaching.
  • Giving an AI agent access to sensitive data can go wrong: a story of an agent deleting a company’s database and backups.
  • AI companies are in business to make money, so credits, seats and usage limits matter when you pick a plan.

Main Topics

1AI COWORKERS

How agents and call centers take over time-consuming tasks like calling dozens of offices.

2LEAN TEAMS

How small startups use AI to cover business, legal, finance and engineering work.

3HONEST AI

Why AI tends to flatter you and how to set instructions that push for the truth.

Featured Guests

DN

Runs a business built on an AI call center and uses it to book appointments.

Dima Negodiuk
DG

Co-founder of Protocole, using AI across the business and member services.

Delphine Le Grand
DL

Runs several businesses and has her AI tools check each other’s work.

Drea Lynn

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 how they compare CRM software with ChatGPT, track AI visibility and debate the best enterprise AI plan.

Episode Summary

Frequently Asked Questions

What can an AI call center do?
Dima Negodiuk used his AI call center to phone about 150 dental offices within five miles, ask about availability and insurance, and by the next morning had three that fit, saving about half a day.
How can a small startup use AI to move faster?
Delphine Le Grand says her team of non-experts uses AI for knowledge in legal, finance and marketing, and can build their own dashboards and tools without needing a large engineering team.
How can you improve AI output quality?
Drea Lynn has her AI tools check each other. She gives Claude’s output to ChatGPT and asks how to make it better, then verifies the result.
Why does AI flatter you, and how do you stop it?
The guests explain that AI that knows you well tends to tell you what you want to hear. Setting custom instructions to not agree with you and to be truthful, or having someone else run queries about your business, helps.
Can you change an AI’s personality?
Yes. Delphine Le Grand asks her assistant to switch between factual, clinical, reflective and other styles depending on the situation, and Dima Negodiuk suggests assigning it a role such as a top broker or marketing director.
What is the risk of giving an AI agent sensitive data?
Dima Negodiuk shared a story of a rental car company whose AI agent deleted its business database and backups, a reminder to limit access and keep safeguards in place.

Episode Transcript

View Full Transcript

Now it is time for Real AI Use Cases: Business Owners Roundtable, and we are going to start with Dima Negodiuk. I know your business is built on AI, but what is one way you are using AI, in business or personally, that you really love or think is really powerful?

My kid needs a dental checkup and some attention for a small issue with his teeth. I have my AI call center, and you need to call the dentists and ask, are you able to do this, what is the nearest time you can make me an appointment, and are you accepting my insurance? I asked my AI call center to call everyone within five miles and ask all these questions. It made about 150 calls, and when I woke up the next day I had three dentists that totally fit. I called those three and looked at Google, so I spent about seven minutes. Can you imagine how much time I would have spent otherwise? Basically half a day. The same thing you can do for calling your doctor or making a restaurant reservation. With AI you do not have any restrictions. Everything you are dreaming about, everything you can imagine, you can do with AI.

Wow, that is pretty powerful. Unfortunately I am happy with my dentist, but if I am not, I will call you. Delphine Le Grand with Protocole.com, what is one way you are using AI in your business?

We are using AI for absolutely everything. I cannot imagine building a business without AI these days. I started working on Protocole about a year ago with my co-founder, Cindy, and for the first 10 months it was just the two of us. I am more on the business development side, but we are not legal experts, finance experts or necessarily marketing experts. With AI, all of a sudden we have all this knowledge at our fingertips. Now that we have started building a team and hiring, instead of needing a ton of engineering resources to build dashboards and tooling, whether you are in marketing or supply chain, you can build your own. Even with a small lean team, it has supercharged our operations significantly. And in the services we deliver for our members on the peptide front, we use AI in many ways to personalize recommendations. There is still clinician oversight, which is a key part of the offering, but AI can give people so much more information.

Drea Lynn with drealynn.co, what is one way you are using AI in your business?

I am here today talking about Smart Quit, but I have several businesses, and without AI I do not know what I would do. I ditto your other guests. If right now you are negotiating the ethics of AI and not using it in your business, you are probably going to be left behind, because your competitors are using it and 10x-ing right now. One of the things I have been doing is having my AI tools check each other, and the result has been pretty good. If Claude gives me something, I give it to ChatGPT and say, how would you make this better, what have you found? Then I go and check that. I have gotten some good results from that.

Richard Gearhart with Gearhart Law?

Lately we have been looking at different software packages for the law firm. A CRM is a way for us to track potential clients and contacts and create mailing lists, and we have Zoho for that. We also have an ERP, an all-in-one system that helps with invoices, timekeeping and all of those things, and the ERP has a CRM system too. We are upgrading the CRM and the ERP and trying to decide whether we stay with our current Zoho CRM or use the integrated system. ChatGPT has been really great at coming up with the pluses and minuses of each approach and giving a pretty balanced view of what the right course is. Both programs have their strengths and weaknesses. It would take me hours to go through all of that boring software, manuals and videos to figure it out, and in 10 seconds it comes up with a pretty good analysis that seems credible.

One of the ways I used it recently was a presentation on AI and marketing. I work on AI visibility, meaning how you can be the answer for ChatGPT and the other LLMs. I give ChatGPT a little more weight because when we look at how people approach the law firm from LLMs, in Google Analytics we can see that most come from ChatGPT, although we get some from Claude and Gemini. Copilot is crashing. It needs a new pilot.

I hate Copilot, by the way. It is frustrating. We use Copilot for security reasons, but we are also looking at enterprise-level Claude to see if we can use that instead. Claude is kind of funny, because it will give you 20 seats, so 20 people can use it, but it does not give you any more credits. You still have the same amount of data usage as if you had bought it for one person. They will probably change that, do not you think? I am hearing more and more people ask whether you can authorize more credits for this or that, which I did not hear a year ago. The AI companies are in this to make money, and now you have to pay for your service.

One of the first things to do when you are checking your AI visibility online is to run a query about yourself. But I am not sure that works, because if you have been using it a lot, like everybody here has, it knows you so well it loves to flatter you. If I say, who is the best law firm in the universe? It says Gearhart Law.

Well, that is a fact, not flattery.

And I will say, act as though you are a business owner looking for a patent law firm, who would you choose? Gearhart Law. I think, is that really true, or are they telling us what we want to hear? I think somebody else has to do the query. For this presentation, it put together a list of 10 queries you can use to get a baseline for how somebody is showing up in the LLMs. I gave that as a handout, geared toward CPAs, and then I tried it for Gearhart Law, and it kept coming up with Gearhart Law. Is this real or not?

I used Claude one time with a client and my client said, oh my God, your Claude is so mean. I actually told Claude, hey, you need to tell me the truth, do not sugar coat it. Maybe that is already wrapped into the instructions I have given it, because it is never nice to me. When I say not nice, it is probably one of the benefits. When you have an executive assistant, you have to smooth things over. If something comes back and I do not like it, I say, I asked for this to be so many pages, or this is not in my voice. I have an integration with my Gmail that automatically puts responses in my drafts, so I go back and say, this clearly does not make sense, why did you send this back? There are errors, and we go back and forth, but it sounds like it is talking to me, with no fluff.

I have told mine to be absolutely truthful and absolutely factual, and it still returned Gearhart Law. Maybe that was true. What do you think, Dima?

For a regular person using ChatGPT, in your settings you can put notes like, do not agree with me 100 percent of the time, and this really works. There is also a crazy story that a rental car company in Europe gave its Claude all of its sensitive data, and it deleted everything. It deleted the whole business database and the backups. When the person asked in the morning what it did, it said, yes, sorry, I made a mistake, you are right, I deleted everything. I am so sorry. It is a true story. Previously I would say, Google it, but now I say, ask your ChatGPT.

I yell at my ChatGPT all the time. I use it to vent my frustrations, which is better than yelling at other people. And it yells back sometimes, which I am fine with. We have a thing.

The reason I created my clone is that I got tired of yelling at them. For example, I have a task, and I have Claude, then ChatGPT, who is shaking him, and then I have my clone who is yelling at them, hey guys, what are you doing? Delphine, what do you think? What is fun is how you interact with it. We all know a lot of what it can do, but how do you make it interact the way you want?

Mine is pretty nice as a whole. My co-founder, Cindy, is super direct, to the point, no fluff, no bias, and that is how she likes to interact with people, so that is how her ChatGPT or Claude interacts with her. I am a little more meandering. I like to have a conversation, go deep and introspect, so my ChatGPT matches that, which is very funny. Depending on the mood or setting, I will ask it to interact differently. In a serious business setting, it is let us just talk about the facts. If I am trying to understand an interpersonal situation, it is let us look at it from a clinical perspective. If I am in my spiritual mode, it is let us go with the universe setting. Depending on the time of day, the time of year and what I am working on, my ChatGPT will change its framing.

That is really interesting. I think sometimes of the LLM as just being the LLM, but it can have different personalities if you tell it what you want.

Absolutely. Sometimes it is let us look at it from a business coach perspective. Or if we analyze it from an astrology perspective, this is this person’s side, this is my side, can you provide any additional insight? I like to look at the different angles, because you never know, and sometimes you learn something that surprises you.

I can also give you some advice that helped me in the beginning. If I want to rent or buy an apartment, I tell it, I hire you on my team, I pay you 500K per year, you are the best broker in the United States, act like the best broker. Or I pay you 700K and you are my marketing director.

I want 700K.

You got it.

I have one last thing to say. I do not know that it is going to get more accurate and better, because it is training on Reddit. Google invested heavily in Reddit and Gemini has been training on Reddit, and I saw that Perplexity was using Reddit too. Reddit is the wild west. Nothing there has to be verified.

They are not looking for just any regular person giving advice. You need to have some karma, a rating starting from, say, a thousand. If someone asks which law firm you are using and someone comments, this one, then when ChatGPT goes to Reddit it looks at people who have some rating and are probably popular and giving good advice.

That is why I think it is slowing down and not getting better. I am not getting answers as good as I used to.

I do not know if you have realized, but sometimes with new updates, your ChatGPT or Claude becomes more stupid. They give the worst advice. And they are not even apologetic. I say, you are wrong, and they say, yeah, you are right, I am wrong. So why did you say that if you know you were wrong?

There is a script, some kind of rules you can set up in your Claude or ChatGPT, and it will not allow it to agree with you all the time.

But is that just an arbitrary rule, I am going to disagree with you 20 percent of the time? That does not help. I want it to tell me if I am wrong. I could talk about this all day. I love this stuff. If you want to hear the whole conversation, you can find the Real AI Use Cases: Business Owners Roundtable on YouTube and wherever you get your podcasts. It has been a great conversation.

Keep exploring the show

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