AI Expert Dan Chuparkoff: Try It Before You Fear It

Dan Chuparkoff (00:00)
One of the great things about AI is that it gives you an army of productivity with a very small team. So you don't need, you know, 150 software engineers anymore to build something that is useful. So, it's kind of a democratizer in that way, and that's amazing. And it's giving everyone some superpowers. We can do things for customers that we never could have imagined before. So, you probably should be, if you have technologists or even light technologists, you should be working on something.

Announcer (00:36)
Welcome to THRIVE, a Paychex Business Podcast. Your blueprint for navigating everything from people to policies to profits. And now your host, Gene Marks.

Gene Marks (00:48)
Hey everybody, it's Gene Marks. And welcome to another edition of the Paychex THRIVE Podcast. Thank you so much for joining us. I am joined today by Dan Chuparkoff. Dan is an AI and innovation keynote speaker. And Dan, first thing I have to say is you didn't need to get dressed up for this conversation, okay. No tie is required. I feel bad. I should have told you this in a bit, you know, and are you out of, like, an office, office right now, or are you talking to me from a, from like, your home office?

Dan Chuparkoff (01:16)
I am from my home office, but immediately after we get off the call today, I am running off to an event where I'm then going to speak about AI to a stage of people, so.

Gene Marks (01:28)
Got it.

Dan Chuparkoff (01:29)
The tie's not just for you. It's actually for the sound guy. It gives him a great place to clip a mic.

Gene Marks (01:32)
Gotcha. I felt like, you know, all you AI and tech speaking guys would, like with the turtleneck, you know what I mean? Or the, you know, the cool Steve Jobs kind of thing. But I guess we're back to something a little bit more formal, so. Well, thank you again. It's fun. And so you're off to go and speak about AI. I'm curious, so, where are you speaking and what are you speaking about? That'll get us started.

Dan Chuparkoff (01:55)
Yeah, so I speak to teams all over the country, North America, sometimes internationally, but mostly it's people from every industry because AI is, you know, is tackling all of us in various ways. And so this week, I'm going to talk to some technologists first, and then immediately after that, I'm running off to another city where I'm going to talk to transportation administrators.

Gene Marks (02:20)
Wow.

Dan Chuparkoff (02:21)
People that run, like, city buses and trains and all those things down in Texas. So, the first one's Raleigh, North Carolina. Second one is Texas. That's what this week looks like.

Gene Marks (02:33)
And is most of your presentations to for the most part, to employees of organizations? To teams?

Dan Chuparkoff (02:40)
No, it changes. It's, different all the time. So there are essentially three kinds of big-event gatherings. First is an employee corporate meeting where we're having the, you know, 2026 strategy rollout or something like that. Something like that. Second kind of group is an association meeting. A bunch of us, we all join an association to get better in our field. Those groups like the transportation association that I'm about to go to on Thursday, those people all get together and share learnings. And then the third kind is a client meeting where the kind of company that has a bunch of clients, they host a meeting to bring some of their customers and prospective customers to learn more about how AI is changing their field.

Gene Marks (03:28)
Got it. All right, that's great. So tell me a little bit about your background. So how did you get to the point where you're talking about AI? You certainly weren't doing this five years ago. I mean, this has got to be like fairly, at least the topic has got to be fairly recent. So, how'd you get here?

Dan Chuparkoff (03:43)
So first of all, I was a software product manager leader for most of my career. I started as a software developer, started building software for the company I worked at, then I started to become the leader. Then I started to lead both product and engineering people and spent three decades, essentially, started in '92 and, you know, ended up at Google at my last hop. Always just building products, software products for users in all kinds of industries. Those particular 30 years, we saw a lot of changes. We saw PCs, then we saw the internet, then we saw mobile, then we saw cloud, then we saw data science. Now we see AI. And so I was just building products through a time of change, and AI just happens to be the next one of those changes. It's really the same talk I'm giving now is the talk that I started writing when we were adopting PCs back in, you know, 1994. So, it's really been a talk a long time in the making.

Gene Marks (04:54)
And it's all about change, isn't it? I mean, I'm assuming that you're, you know, the people in your audience are somewhat, some of them are terrified about AI, some of them are more embracing of AI. What are you seeing out there? Do you, do you see a lot of people being dubious? Do you see them, you know, do you see businesses embracing AI more and more? Do you see people holding back? What's your, like, your general take?

Dan Chuparkoff (05:17)
Yeah, I see the world is sort of bifurcating to the two extremes because our ways of getting information these days, whether that's through social, through news, through friends talking who got their information through social or through news. Our media channels these days are click optimizing and the best way to get clicks is to say something extreme. Right. Because that gets a bunch of people disagreeing with you.

Gene Marks (05:52)
Right.

Dan Chuparkoff (05:53)
And that drives clicks up. So in most of the media channels, we get people saying AI is going to be the greatest thing humanity has ever seen. It's going to do everything, it's going to cure cancer, it's going to put people on Mars, or they're saying AI is consuming all the water, it's consuming all the electricity, it's going to kill us all. There's no pragmatic middle. I can write a post that says, hey everyone, it's going to be okay, but no one's going to click on that. So people don't get that information. So, people are at one extreme or the other generally. And I'm trying to bring them a little bit more closer to the middle.

Gene Marks (06:29)
And what is your message being in the middle? Okay. I mean, you know, we're, we're a branded podcast, so we're not looking for anything in the extreme either. But we, you know, but, so are you generally a glasses-half-full person when it comes to AI?

Dan Chuparkoff (06:43)
I think that it's going to be okay because for a long time I was working on software to make sure that everything's going to be okay. There's a lot of people working on cyber security protection. There's a lot of people working on Google Cloud to make it more efficient and more sustainable. Like, you know, all those things are being worked on with, you know, thousands and thousands of people's energy and so, it's going to be okay. I also don't think AI is capable of taking your work because you have an infinite amount of work. There's no Thursday afternoon when you run out of work. Right. If you get done with all the work you thought you have, I promise you're going to find a little bit more work that you weren't getting to before. And so no matter how efficient we get, I think there's always more work that your customers want from you.

Gene Marks (07:34)
Yeah, I do agree with you on that. I also agree with you just work for employees. I've seen like listings of 50, 60, 100 job titles that didn't even exist 20 years ago. Right? So, it just seems like humans managed to find stuff to do to keep ourselves busy, you know, and so I guess that sort of terrifying aspect of it is there. Do you, what is your message to employees when it comes to AI? I mean, should they be afraid of it? Do they think it's going to put them out of a job? Like what do you tell, like, you're speaking to these transportation workers, people, you know what I mean? Like, what's your message to them?

Dan Chuparkoff (08:11)
I think my biggest message is one, if you just start playing around with it a little bit, you will see the kinds of things that can be helpful. And so first and foremost, before you decide if you're afraid or not afraid, try using it. That's my first message. And one of the prompts that I usually tell people to get started with is go into whatever AI you like. Whether that's, you know, Perplexity or ChatGPT or Claude or something else. Go in there and say, hey, AI, give me one new AI thing that I should try every single Monday. And then every Monday you get a new idea. And maybe you like that idea and maybe you don't, but you'll get it. And some of them you'll try and you'll start to see the kinds of things that AI can do for you. And then, you know, then you'll start finding your way. You'll start to see that AI can be super helpful. You'll see that AI can make you a little bit better at some of the things that you try to do.

Gene Marks (09:15)
Right.

Dan Chuparkoff (09:16)
You will also probably see that AI doesn't usually make you faster.

Gene Marks (09:20)
Right, right. In some cases, it creates more work for people.

Dan Chuparkoff (09:23)
Yeah, yeah, yeah, exactly. And so, as you start using AI more, you'll have more and more and more and more stuff you could use AI for and so, you'll start to less afraid.

Gene Marks (09:33)
When people talk about AI, it's a very general term. So can you be, like, more specific? Like if you were to say, like, hey, you need to be using AI more so that for all the great reasons that you said, what do you mean by using AI more? Be specific.

Dan Chuparkoff (09:52)
Yeah, so people do treat AI like it's a piñata. Like, there's all this cool stuff inside the AI piñata, but no one really knows what's in there, so we're all just sort of banging at it with a stick and that. And that creates confusion. So when I when I say AI, I really mean using an AI assistant. So asking ChatGPT a question and getting an answer back, or asking AI for, I'm sorry, or asking ChatGPT for a piece of content. I want to write a LinkedIn post. Can you give me a suggestion for a good LinkedIn post? Right. Asking AI for some help, whether that's content creation or feedback for yourself or a strategy idea or a brainstorming topic. That's what I mean when I say AI. Beyond that, once you're already using AI in that way every day, you think of AI as a assistant sitting next to you, that you ask for questions all the time. Then there will be some other things too. You might have that AI go off and do some things on its own, like book a flight to, you know, Orlando for you for your next conference.

Gene Marks (11:05)
Right.

Dan Chuparkoff (11:06)
But first, start with the assistant thing.

Gene Marks (11:08)
Yeah, that makes sense. How about connecting your AI assistants to other data? ChatGPT has connectors to QuickBooks and Slack and Gmail and you know, SharePoint and, you know, a bunch of other, you know, platforms out there and Dropbox is another one. Talk to me a little bit about that. Like, do you ever, do you have clients that do that? Do you have any advice or thoughts on connecting to other data sources?

Dan Chuparkoff (11:34)
So I think most people, most people don't understand the permissions and the privacy implications of doing that. So, for most people, I recommend don't connect it. Use it as an AI tool where you, if you need data in that thing, copy and paste it into that thing so you know exactly what's going in there and exactly what's not going in there. I do see some people giving access to their email. You know, Claude Cowork makes it super easy to connect your inbox and give it permission to look at all your email. Inside of your email you have some stuff that you're forgetting is there. Like you know that time you sent your bank account info to that company that was going to give you money? For most people, I wouldn't do that. I have very few connections to data sources between my AI tools and those data sources because there are some privacy challenges with doing that and there aren't, the management of those privacy challenges is not very mature yet. So I manage that privacy by copy and pasting myself when I need it.

Gene Marks (12:52)
Isn't that just like a risk-reward conversation though, Dan? I mean, you know, these connectors do exist for a reason and there really is a lot of great ways that you could be using these AI assistants if you're using it on your data because it can be analyzing, helping with your company's data. And there's no question that there are privacy concerns. So, you know, so for starters, like, you don't feel that, that if you connect to this to you connect say ChatGPT to your company's Dropbox or OneDrive or Google Drive, again, do you feel that a company is compromising their privacy? Is that, is that true?

Dan Chuparkoff (13:30)
I don't think that like by default, just connecting doesn't mean you're compromising your privacy. What I do think, though, is you could be in some cases. Right? If you're not controlling, if you're connecting to the whole data store, first and foremost, you might not need to. If you need to, then sure, do it. But you could also just create one shared folder where you connect that, and then you put stuff that your AI needs to know in that folder. Right. And so that, that's, you don't need to give it access to 45,000 documents from the last 27 years. Um, you can. And if you are actually looking at the privacy policy and deciding which AI tool you're doing that with, then you can get some immense benefits. But your, the amount of data you connect should be proportionate to your understanding of the privacy policy.

Gene Marks (14:36)
And, and I guess the next question, you know, related to privacy is, you know, do we, you know, is that such a big risk? I mean, don't take this the wrong way, Dan, but like, let's say that, you know, ChatGPT has got all of your information about you or your business or whatever, you know what I mean? Like, you know, like, who cares? I mean, like, is that...

Dan Chuparkoff (14:57)
Yeah, that is a, that's an amazing, excellent point. It's probably fine. ChatGPT isn't really going to use your bank account number to train their algorithm. There's nothing useful from that number. Right. So that's all probably fine. I think that we're in the middle, though, in this slippery slope where people will then grab a new product that just came out because they're like, oh, people are using this one tool that does something cool, and then that tool then by proxy has access to your data and...

Gene Marks (15:33)
Oh, I see. Right. Indirect impact. Right. Yeah. Now that, that, that makes total sense. It does. You know, I also hear a lot from my clients that their nerves when it comes to privacy and dealing with like, say we're picking on ChatGPT, but it's all about AI systems. It's just that the they're like, well you know, if I connect it to my Dropbox folder of quotes and then I, you know, I'm using it to you know, ask questions about quotes I've done in the past or draft a new quote or whatever say, can't my competitor then see my pricing that I'm using? you know? Now I always, you know, I thought about that. I'm like listen, anything's possible, nothing is 100% secure. You are, you know, But like, you know, the way I kind of argue is I'm like, you know, most of my clients' data is already in the cloud. I mean it's on Azure, it's on Google, it's on, so it's on AWS. So like you know, we can't do that now. You know what I mean? So I don't, you know, I don't know if that's really like a feasible thing. Like I can search on ChatGPT for my competitors' customers, you know what I mean? That kind of thing.

Dan Chuparkoff (16:39)
Yeah, it would certainly, it would certainly be really hard for that to happen. Right. It's probably not impossible, but super, super hard. Like first of all in the train the algorithm training process, when you are trying to build an LLM, you are looking for bits of information to remember and if you only see something on the Internet one time, it's probably not worth remembering. So, like the fact that this particular customer is a customer of that particular service provider is an anomaly in the database that is not worth saving. And so, it's only helpful if you know, something gets repeated over and over and over again so much in different contexts. And so, it would be really, really hard for a privacy thing like that to come out or financials. Like my customers will know my revenue. That's, first of all if they did know it, there would be a time lag.

Gene Marks (17:45)
Right.

Dan Chuparkoff (17:47)
It's not going to happen instantaneously. It still takes like four or five months to train these algorithms.

Gene Marks (17:53)
It does.

Dan Chuparkoff (17:53)
And not everybody even uses the new version so.

Gene Marks (17:57)
No, no, you're absolutely right. It's also interesting, there's so many, your point about there can be indirect access to data is definitely well taken. You had mentioned about like if you're gonna, you know, build or do something similar on your own with AI and I, you know, it's funny you know Dan, I've a growing number, it's very small still but starting to grow of my clients, small companies, you know, 100-person company, 50-person, whatever, are starting to dabble into, you know, building their own thing, you know. Just a couple weeks ago, Mark Cuban said, you know, you know, people asked him about the opportunity of AI for small businesses. And you know, he was like, listen, there's, there's 34 million businesses in this country and every single one of them has an AI need and they don't know what to do. Which is a massive opportunity for people that build AI solutions.

Dan Chuparkoff (18:50)
Yeah, for sure.

Gene Marks (18:51)
So, and I guess the question is like, is that, you know, do you think that's a viable a, you know, a reasonable road for a smaller company to take even in 2026 is to build something internally? And if so, how would they, what should they know about building their own AI? It sounds precarious.

Dan Chuparkoff (19:13)
I think the, yes, one of the great things about AI is that it gives you an army of productivity with a very small team. So, you don't need 150 software engineers anymore to build something that is useful. So it's kind of a democratizer in that way. And that's amazing. And it's giving everyone some superpowers. We can do things for customers that we never could have imagined before. So you probably should be, if you have technologists or even light technologists, you should be working on something. I think that the caution that I give people is as you're working on stuff, people tend to overvalue how much they will learn from learning on their own data. When Google Transformer, which is generally speaking the first transformer model, the first thing that kind of did what GPT does now, that was built By Google in 2017, we trained Google Transformer on 50 million examples of data and it wasn't very good. Nobody heard about it, nobody knew about it. There were hardly any examples, right. It wasn't until we got to about 13 billion examples of content that GPT3 finally started to surprise people with its results. And so if you're going to pull all of your data and try to build your own custom AI on your data, realize that you don't have 13 billion examples of anything and so don't over index on that. You should, however, you should use chat as an interface to look stuff up in your knowledge base, right? You should integrate, you should have some sort of interface that looks stuff up in your own data sometimes and looks stuff up in the world's brain. Now ChatGPT or Claude or Gemini, right? Toggling back and forth when sometimes you need like industry data or world knowledge and sometimes you need client doc data.

Gene Marks (21:33)
Right.

Dan Chuparkoff (21:33)
You should build that switching mechanism on your own. I think that's where the high ROI is right now.

Gene Marks (21:39)
Can you think just on that point, like if you're a typical small business and again, every business is different but like, can you think of an example or two where you think like a quick, you know, AI application might help, you know, a business owner grab some low hanging fruit and provide some ROI? Anything come to mind or anything you've seen?

Dan Chuparkoff (21:59)
Yeah, I think, I think first and foremost, most organizational IP is not in documents. It's spoken in meetings and conversations. It evaporates in chat. And as soon as those threads are gone, people forget about that stuff. And it's important. And so as quickly as people can. I think we should institutionalize the recording of conversations with meeting notes. Even when those conversations are one-on-one, you know, even, even maybe when those conversations are in person between two people. Like I bring up, I bring a recorder around with me everywhere and I'm recording notes all the time. Those notes should go in knowledge bases that are searchable later. And right now, retrieving stuff from knowledge bases is searchable. It's a pull thing. But quickly, what happens is when people start using a, an AI assistant to help them do work. I'm working on this presentation for my vp. Hey, AI assistant, can you help me make this presentation? That AI assistant can go back through all your old notes and say, remember last October you had that conversation with this team. That's important right now. And it'll pull those old institutional memories into your current work stream where you would have forgotten. And that's where our work products get dramatically, dramatically more effective.

Gene Marks (23:31)
Yeah, I think you couldn't be more correct. You know, there are so many conversations that go on. It can be phone online or just face-to-face meetings that they disappear into the ether after they've happened. And you imagine just think about grabbing them all. I talked to some guy who developed a little AI application for himself. He does like Windows and doors and he has a showroom and he has his. He had somebody build an AI application so that when a sales say you were going to go and buy like some Windows from this guy and you're walking around the showroom, the salesperson is just recording the conversation as they're walking around and you're like, you know what? I think I'd like to get this, you know, in this color and this, you know, dimension or whatever. And the sales guy's like, that's fine. You know, I'll send you a quote, you know, on that. Give me a day. Their AI application could take that whole conversation it just had and create a draft quote for the guy, you know what I mean? So that, which because otherwise he would have gone back to his desk and he would have to look at his notes and he would probably would have forgotten some things and whatever and then take some time. I think that's like a very like straightforward low hanging fruit application for AI that just like boom, really addresses a need and it's not that difficult to do.

Dan Chuparkoff (24:43)
Totally, totally agree.

Gene Marks (24:44)
Yeah, be fun. Okay, so we're almost out of time here. Let me close up this other question that I had for you because I've been looking forward to asking you this as well as about data. You know, I mean, Dan, I, you know, you know our data is terrible. My company's data is terrible. Most of my clients data is that, you know, it's like not terrible terrible but you know, we're missing data fields are open, things haven't been updated in a while, that kind of thing, you know, and yet you're like, well, you know, if you're going to run AI, you know, you really, you know, data is really cool, you know, it's really critical. And yet I, I spoke to another guy who develops AI solutions and he's like, you know, you're not wrong. I mean data obviously needs to be in good shape for AI to be trained on it well, but he feels that that's becoming less and less of a problem. He feels that today's AI, you know, you know, platforms and future ones, very near-term future, it can understand and fix data or understand even bad data better than your misspellings or wrong locate. And then also AI can, applications can be built to check on data with external sources too. So my CRM system might have this address for the customer, but before anything gets done, it's going to check a couple of external sources to validate that address and then say yes, that is the address. Anyway, I did a lot of talking, I apologize, but what are your thoughts on data as we head into.

Dan Chuparkoff (26:14)
Yeah, I agree with him. Anomalies in data don't matter as much as they used to. We used to spend a lot of time cleaning up our data, transforming it as it came in, you know, the whole ETL, you know, extract, transform, load. Like there's whole teams, there's whole companies that exist to help, you know, cleanse data on the import so that when people query it in databases they never get wrong answers later. AI doesn't really care about that stuff. The, the metaphor that I use sometimes is, you know, these, these new AI things, like some people say, AI has actually been around for a long time. The example to illustrate that it actually has, that I use sometimes is AI just looks at the trends of past data and uses that to predict answers right now. And there's a version of that that we all are moderately familiar with, regardless of our industry. And that's life insurance. Right. If you are a male, if you're a male in the US the life insurance actuary people predict that you will live to age 78.2. 78.2, Great. All right. That's how I buy life insurance. That's how health decides all that stuff. They're doing the exact same thing our AI assistants are doing now. They're looking at those trends. So imagine life insurance actuaries. Imagine that. I don't know. There's 7,000 people who we didn't write down the day they died or their birth certificate was fuzzy. We couldn't really read it. So we don't exactly know. Does that matter? Does that really materialistically Change the number 78.2? It would have to be millions of mistakes. And so you're probably fine. Just get, get some data, collect more of it over time. And the more new stuff that you collect will overshadow your, your, your data anomalies and your missing, missing fields and stuff like that.

Gene Marks (28:18)
Final question. I'll let you go. It's been great. Just listen. I again, don't think this is the wrong way. You can't, no one can really predict the future. Like, we don't know, you know, what's the, you know, you have some ideas, but, you know, so I'm going to just ask you at least for the next maybe 12 months, you know what I mean? Like, at least looking forward to say, the middle of 2027. And that's hard enough given how fast everything is moving. But where businesses are concerned, what do you expect that they will be? What do you expect I will be used within businesses?

Dan Chuparkoff (28:49)
I think the biggest thing that is untapped right now is AI as a coach, people, right now, when I use AI. So for instance, at the end of this podcast, I'm not recording this one, but eventually it will come out and then I'll have the recording and I will, I'll be able to take, I'll be able to ask AI how. How did I do in this podcast? What could I have done to make this better. What? When did I ramble on? When did I. Was I not concise? When did I use the wrong metaphor? AI will give me feedback so that the next podcast I'm on will be better. I ask AI for feedback on everything I do. Every client call, every speech, every email that I write, everything. There are no things that I do where I don't say, AI, how could I have made this better? I don't ask AI to make it for me at the beginning. I do it myself and then I ask AI for feedback. That's, I think, a thing that people will do. It will be doing with way more frequency than they're doing right now in 12 hours.

Gene Marks (29:57)
I love that. I love that, you know, you have to have a bit of a thick skin so you could take your criticisms. And some of these AI platforms are being like, ChatGPT is getting a little bit less chirpy and cheerful and a little bit more, you know, critical, which is good. And I do the same thing with you. When I write columns, I right now, I always upload them first to ChatGPT and say to one thread that of all my columns, I'm like, given what you've advised, you've given me in the past, give me, you know, how can I make this combat? And it will suggest, you know, you might want to rephrase this, you might want to change. So just like you, I will write the whole thing first and then say now, you know, give it back to me. And. And it works really well. Well, Dan, this is great. I really appreciate all of your time. You've really helped us prepare, and obviously, in the nature of what you do, things are going to be changing quite a lot. So it would be great to speak with you again. But thank you very much for your time.

Dan Chuparkoff (30:48)
Yeah, it was a great conversation for me, too. I appreciate the opportunity to be here and let me know if I can ever be helpful in the future.

Gene Marks (30:55)
You will. And you know, your AI assistant will give you top marks on this conversation, so you have nothing to worry about when you ask it. Everyone, Dan Chuparkoff is an AI and Innovation keynote speaker. Dan, first of all, where can we find you online?

Dan Chuparkoff (31:09)
I'm pretty easy to find if you can spell my last name. C-H-U-P-A-R-K-O-F-F for the audio listeners. And LinkedIn is my most active platform, so follow me there. I'll accept your connection request. I don't care who you are. And we can talk more in the future about AI changing work.

Gene Marks (31:27)
That sounds good, and we'll put your put that information into the show notes as well. So thanks again for your time and everybody. Thanks again so much for listening or watching. My name is Gene Marks and you've been listening or watching to the Paychex THRIVE Podcast. Do you have a topic or a guest that you would like to hear on THRIVE? Please let us know. Visit payx.me/ThriveTopics and send us your ideas or matters of interest. Also, if your business is looking to simplify your HR, payroll, benefits, or insurance services, see how Paychex can help. Visit the resource hub at paychex.com/WORX. That's W-O-R-X. Paychex can help manage those complexities while you focus on all the ways you want your business to thrive. I'm your host Gene Marks and thanks for joining us. Till next time, take care.

Announcer (32:12)

This podcast is property of Paychex, Inc. 2026. All rights reserved.