Beaumont Vance, vicepresidente senior de IA de Paychex: Cómo las pequeñas empresas pueden mantenerse competitivas con la IA

Gene Marks (00:00)

Hey, everybody, it's Gene Marks. And welcome to this week's episode of the Paychex THRIVE Podcast. Thanks so much for joining us. This week I spoke to Beaumont Vance. Beaumont is the Senior Vice President of Data Analytics and Artificial Intelligence at Paychex. He is in charge of all of AI. If you're running a small business or a medium sized business and you are looking to invest in AI, how will AI impact your business? The importance of data, making your data private as well. We cover all of these topics. This is a conversation that, for anybody that I know that is interested in AI and also want to know how AI will impact their business, who better to talk to than somebody that is well versed in it and also immersed in it at a large company like Paychex? So we're going to get into all of that with Beaumont. Let's go to the conversation. 

Announcer (00:50)

Welcome to Paychex THRIVE, a Business Podcast. Where you'll hear timely insights to help. 

You navigate marketplace dynamics and propel your business forward. Here's your host, Gene Marks.

Gene Marks (01:01)

Hello, everybody. So, I'm here with you, Beaumont. I'm very, very happy that you are here as well. Your title, SVP Data Analytics and Artificial Intelligence. Let's start with that. What exactly does that mean? 

Beaumont Vance (01:20)

Well, it's a mouthful and let me just put on the record first, I did not create the title. The title was given to me when I showed up by the board. So basically, what it means is if you think about the ecosystem of the information age we're in right now, especially, you know, everybody's focused on artificial intelligence mostly. Really, artificial intelligence doesn't work at all unless you have tremendously good data and a lot of it. So, data analytics and artificial intelligence are all a way of saying leveraging the data that we have and getting value from that to help our clients and to increase our revenue, basically. So, it's meant to cover the entire umbrella of not just the data, but making sense of the data, building automation systems of the data, building AI tools out of the data, and putting all that together into products for our customers and a more efficient company. 

Gene Marks (02:09)

You know, we're going to get into, we have a lot to talk about, you know, in this segment, but we're going to get into some of it. But you mentioned about having data and having the integrity of the data being there. I talk about this a lot when I talk to business owners about how AI like doesn't work if your data sucks. Right? 

Beaumont Vance (02:27)

No, you know, I came up with an analogy that seems to work, which is if you think of AI as a magic librarian. So, imagine some type of librarian who, if you put them in a library, they immediately know every single thing that's written in the entire library. Not only that, they can remember who wrote it and they can cite it. The page in paragraph. Right. So really, really valuable. But the problem, the downside is that they only know what's in the library. So, if you go to a medical library and you ask them a question about medicine, you'll get a very good, accurate answer that reflects the data in the books in the library. But if you ask them about music, no idea. You ask them about, hey, how do I terminate an employee? No idea. Can't tell you about that. And AI is exactly the same way. It's actually quite dumb until it has the appropriate data to train it and teach it what to give an answer to. And so a lot of the AI out there I see is scraping the Internet. And it was trained on the Internet, like the GPTs, ChatGPT4, and everything have been trained on all the information that there is out there in the entire web. The problem is, well, how reliable is the information on the internet? Does it ever contradict itself? Does it ever lie? Is it ever inaccurate or out of date? Well, yeah, of course it is. It's terribly out of date. And the AI doesn't know what to do with that because it's not truly intelligent. It's kind of a misnomer to call it artificial intelligence. So, what it does is it just looks at it and it says, hey, I've got, like, five answers to this question. What's the best vacation spot for a family of five in February? Well, it's got, like, 20 different places. What it does is it'll give you the average . And it might be like a spot somewhere in the middle of the Pacific Ocean where there's no island, but it will average it. And that's what we call hallucinations. Right. So, data really is the foundation. And one of the things that attracted me about coming to Paychex was I've been doing this. I've been working in artificial intelligence for, since about 2007 when I was at Fidelity Investments, was I understood the importance of the data. And Paychex is one of the few companies that not only had the data, we have the best data set in the HCM industry in the world, basically. And they understood that, and they understood that that was the foundational aspect that they needed to build upon to be able to provide for their clients. And so, they understood AI in the context of data, which I have to tell you, is very rare in corporate America right now. Most people are still sort of focusing on the AI piece and they're just starting to realize, especially as the lawsuits come rolling in from various data providers and publishing companies like Wiley and The New York Times who are starting to file these lawsuits because they're saying, hey, you trained all of your AI on our copywritten data. Right. People are starting to realize, hey, it's the data, right? 

Gene Marks (05:04)

What's your plans at Paychex? I mean, you know, you have, you have vast amounts of data to work with. It's impressive to me that a company like Paychex are making the investment and bringing somebody in at your level to be responsible for AI and the data that they have. What exactly does that mean? Like, well, you know, will your focus be on developing more products for your customers that are leveraging AI, or will it be more of an internal focus to help, you know, to help employees at Paychex use AI better? 

Beaumont Vance (05:36)

It's both. I have a very, very broad remit. So, we are actually looking at what are what our clients are telling us. And I can tell you this is a great AI story. This is a great success story. So, we, we talk to our clients roughly 36 million times a year. That's just emails and phone calls and chats. Right. That doesn't even include all the in-person conversations and everything else. But at least 36 million times a year we're talking to our clients and in those conversations, they are telling us what their challenges are, they're telling us what their needs are, they're telling us what they like and what they don't like. Now here's the problem with that data. Up until a few years ago, that data was stored on recordings and it comes out to about 450 years’ worth of recordings every year. So, if you want somebody to go through and listen to what the clients are saying, the voice of the customer. Everybody talks about voice of the customer. Right. If you want to actually go through all the voice of the customer, you would need 450 years of man hours working 24/7, 365 days a year. Right. It's impossible. Well, the big unlock with the AI revolution was when the transformer models came out in late 2017. Really, it allowed us to process linguistic data. 

Gene Marks (06:51)

Explain to us what a transformer model is. 

Beaumont Vance (06:53)

So, the Allen Institute first came out with one called Ernie. No, it was Ernie. Yes, Ernie. It was Ernie first, which essentially it's an advanced AI model that was a big breakthrough that allowed it to do a lot more than it could before. That's all you really need to know. And then Google quickly released one called Bert. And for the listeners out there, it's a funny story. So, you know, engineers have kind of a sense of humor. And so everyone who was working in this field, every time they came out with a new breakthrough AI model. 

Gene Marks (07:23)

Let me guess. Sesame Street. 

Beaumont Vance (07:24)

Right, Sesame Street. So, you had Big Bird and Big Bert, and you'll Never guess what ChatGPT was supposed to be named in the pantheon here. 

Gene Marks (07:33)

The Cookie Monster. 

Beaumont Vance (07:34)

Close. That would have been a good one. Snuffleupagus. 

Gene Marks (07:38)

That's even better. 

Beaumont Vance (07:39)

But apparently the marketing department got ahold of it and said, no, you can't call this serious thing Snuffleupagus. But anyway, when those came out, what happened is, up to that point, we were able to process, you know, going back to the 60s, we could process numbers really well. But you couldn't do anything with a book or a contract or a freeform conversation like we're having right now. Right. All of a sudden, it gave us the power to go through that. And what we've done at Paychex is built an AI system that is dedicated to understanding what our clients are telling us.And so, in addition to the NIFB surveys and the surveys we put out in focus groups that we do and feedback we get from our sales reps, I mean, we're really dedicated to what our clients want. We also have the direct voice of the customer that we can get daily that tells us what our customers want. So our plan with AI as far as customers go, is to understand what their wants and needs are, what their biggest challenges are, and then do what we've always done at Paychex for the past 50 plus years, which is, you know, help the small guy compete against the big guys, you know, and that David and Goliath struggle that we're all going on, going through every day. And we were talking earlier, I grew up in a small business. I don't think my dad ever had more than 10 employees. So I know. And he was fighting against these big chains that were starting to come through. I know exactly what that struggle looks like. And in that David and Goliath battle, Paychex has always been the slingshot. It's been the great equalizer. Right. And so, what we plan on doing is using our expertise and knowledge and vast repositories of data to build tools that solve our clients most pressing issues and needs. 

Gene Marks (09:12)

It is, you know, AI in the past two years, three years has really been like a big corporate play. So, it is a, you know, it's companies that have the resources like a Paychex to hire somebody like yourself. And then to invest obviously in the team that you need to actually take. This is taking, you're taking just reams and reams of voice data and like turning into something that could be used and managed. Right. You, for smaller companies, how, how can they take advantage of something? Say you were running a smaller business. Is it still something within their... Their perfect... I mean can a small business do what Paychex is doing right now, or do we have to wait until more tools arrive? 

Beaumont Vance (09:53)

It would be very, very difficult in the AI world. I mean, you know, ChatGPT3, when it came out, I think initial training, just the training, yeah. Cost $110 million and was something like 350,000-man hours to do that. So, it's already building an AI platform is out of reach of even Paychex. Right. That's a huge difficult investment. And it gets to, you know, business specialization. There's no reason that we should be developing our own AI system when there's plenty of startups out there. There are, there are over 70,000 AI startups globally right now. Right. And there are all the brightest minds in the world are working on, you know, the AI platform. And then you've got the issue of you got the AI platform now you need the data to train it and give you answers. So, what I would say is small businesses don't have to. I mean they're already struggling with more challenges than any one human being or team of 20 human beings could handle. And you think even just the regulatory climate, there's got to be hundreds of thousands of pages of regulations and laws that apply to a small business, especially if they're multi stated. It's really complicated. And what those businesses have always done, the smart ones, is they've always figured out that's not my core competency and they have found a good partner like Paychex that they can partner with, and they can leverage that expertise. Now with AI, it's going to be the same kind of play. I would say everybody's doing this already. Anyone who's using Google Maps, I used it to drive here today. I'm not worried about getting a map repository for all of Rochester. Any place I travel, Google has that, it's free. But low-cost providers have these things, and I outsource that type of work to Google in order to navigate a city that I've never been in before. 

Gene Marks (11:37)

So, you hear...when I talk to clients all the time. And I was speaking to this venture capitalist just a couple weeks ago for a piece I wrote in Forbes and his company says that big companies like yours, there is huge opportunity for startups that are in the area of AI, because a lot of big companies don't have the resource or the knowledge to do this. So, they wind up like, you know, partnering with a lot of startup companies to develop the solutions for them. And that's all great. But he also said, saying, like, if you're running a business of any size and you don't leverage AI, you're going to be out of business in a few years. You can't, you know, I mean, you're not going to be able to compete with others. And when I look at my clients, Beaumont, I mean, like, they're like little manufacturers or mom and pop stores or family-owned businesses. You know, I say to you like, you know, I'm like, say you're a client of mine and I'm like, if you don't leverage AI, you're going to be out of business in private. And he's going to, what am I going to do? 

Beaumont Vance (12:25)

Yeah, what AI, what do I do? My dad back in, I think this was in the 80s, my father, as I mentioned, had a small business and his brother was a, was an engineer, a computer engineer, I believe at Kodak at the time. And he said, you got to get in this computer thing. And my dad, computer what? So, he bought a Commodore 64, and he had it in his office for 25 or 30 years. Stared at it, I think I didn't turn it on once. His brother came by and showed me, he got like the little prompt and he's like, what am I supposed to do with this? And that was the end of that. Right. And so, I think that's asking the wrong question. The question is, what tools can I have that match the biggest challenges and problems that I'm facing? You know, we can talk about it in a bit, but you know, one of the ones we've heard from our clients is finding qualified candidates. Right? 

Gene Marks (13:07)

Yeah, let's talk about that. 

Beaumont Vance (13:08)

Sure. And so, finding qualified candidates is the perfect example. Right. Everyone is trying to solve this thing, and you'd say, well, where are you going to leverage AI? Well, right now, or up until we launched HR Copilot in September, for the past 4, 5, 600 years or so, the way you got candidates was you posted a job. It was a posting. Either stuck it to a tree or you stuck it on a website, but it's still posting a job. And I grew up on Lake Ontario, and it always reminded me of is as a kid, I know there's trophy fish in Lake Ontario, right? And I'd go out onto Lake Ontario, but then you stand there with your fishing rod in hand, and you look at the lake and you can't even see the other shore, right? It's a vast lake. So, you know they're in there. So, what do you do? You bait your hook, you throw it out, and you sit and wait. 

Gene Marks (13:53)

And maybe you get lucky. 

Beaumont Vance (13:54)

Maybe you get lucky, but most times you catch stuff you don't want to catch. You catch small fish, you catch carp, you catch things you don't want. And that's the way we have approached the job market for hundreds of years. So back when I was in private equity, I came across this company called Findem. And Findem basically was a bunch of very, very skilled database engineers, probably the best database engineers in the world and very, very good AI engineers as well. And they said, this is backwards. What we should do is we should be able to go in and make a query and just get a list of every single qualified candidate in the world, wherever they are, whether they're employed or unemployed, whether they're looking or not looking. And instead of sitting there and waiting, we should find them, get a fish finder, right? Go find out where the fish are and then go try to lure the candidate in. Right? And so they've got a database with over 750 million candidates globally. And just to give you a scale for the U.S. there's about 167 million working adults in the United States right now as of last, last April, right? Give or take. Their database has over 230 million adult candidates in that database. So, they have more candidates than there are working people, which means they have all the people who aren't looking and people who retired recently and people who, you know, who are still in the systems but are maybe looking or maybe doing a side gig or something like that. So you can go into this database. And I use it myself. In fact, over the holidays, I was hiring someone who I used finding this system. This is what we use to find our candidates now. Within a few seconds, I can get a list of every single qualified candidate to my specifications. Geographical, industry, background, education, whatever, whatever parameters I put in, and I can get a definitive list of all those candidates with their contact information. I can press a button and AI will write a customized email outreach to them and reach out to them. 

Gene Marks (15:50)

And, and these are people that are looking for jobs or not looking for jobs. 

Beaumont Vance (15:53)

Right. No.

Gene Marks (15:54)

They're just qualified. 

Beaumont Vance (15:55)

And the thing that attracted us to this in private equity, number one, is this is so disruptive. 

Gene Marks (15:59)

Yes, it is. 

Beaumont Vance (15:59)

It completely turns the entire concept of recruiting on its head, right?

Gene Marks (16:01)

Because normally you're looking for people that they're out there looking for a job. I mean those are the ones that are going to the recruiting websites right now. But why not be talking to people that might not be looking but could be a great candidate for you? 

Beaumont Vance (16:12)

Right. You know, prior to us having this up and running, I posted several jobs. I got in one job, I got 284 applications over the course of six weeks. So, it took six weeks to get 284 applications. Now I have to plow through 284 resumes to figure out, out of that group I had zero qualified candidates. Zero. I used to Findem and I'm looking in Rochester, which is a smaller hiring market, and I got 10 qualified candidates. None of them were looking for a job. And in fact, after I reached out to the person I'm going to hire, he said, you know, I never would have looked at your company because you know, I'm doing my own thing and I'm, you know, I've got this great thing going. But man, now that you tell me what you've got going on there and what the company's like and now that I've seen it a little bit, he's like, I'm in. I'm really excited, I'm excited to join you. So not only are you getting people who are not looking, you're getting people who might not ever consider you. And for a small business, I'm telling my three sons are young and just starting their work careers and I keep telling them, you know, guys like, like all people their age, they've heard of Meta, and they've heard of Google and they've heard of Apple and they all want to go there and it's really impressive. But the companies that are doing really well and the companies that are flying under the radar that have, you know, great opportunities, especially for young people, are small, medium sized businesses and they're all over the place, but they're not on the radar. So how do they compete when they have no name recognition, they now have no brand recognition, and they have no recruiting staff to go talk to these people and sell them. They're not doing job fairs; they don't have time for job fairs. So, what this does, it turns it all, the whole industry on its head and you can actually just target the people you want. 

Gene Marks (17:48)

So, Beaumont, this gets back to like the early question I asked. When people say to small or mid-sized business owners like, oh, you need to be competitive with AI, you need to have whatever. I think what you're saying is you actually need to be aware of AI applications and tools that are out there that you can use and leverage, like, for example, recruiting. I mean, we have a very tight job market. It's going to continue to be tight. And every single client that I talk to, their biggest issue is finding good talent. Right. So, no one says that you have to build your own AI solution to do that. When there's tools out there that will do that for you, does that make sense? 

Beaumont Vance (18:25)

Thanks for bringing me back. I will digress. Please feel free to give me some discipline. No, that's exactly it is that there's, there's going to be a million AI solutions thrown out there as well too. Right. And it gets very, very confusing. What I would say is that there will be, think of more like apps. Right? There are going to be AI applications that solve business problems on the market. And staying abreast of those and taking a look at those and, you know, doing a few demos or reading about them a little bit is probably all that a business owner has to do. They really just have to understand their business, what problem they have, and then go look for the solutions just like they've been doing, you know, running a business all the time. AI is no different. 

Gene Marks (19:04)

More questions for you. You're talking about leveraging these apps and I completely agree with you. And that's the advice that I've been giving as well. And first of all, you know, anybody that's running a business, they've got their core applications that they're using. If you're in manufacturing, you're using Dynamics or Epicor or Sage or something like that. You know, those are the companies that are spending hundreds of millions of dollars on developing AI features and functionality because they want to make sure that their customers are taking advantage of them and liking their products. That's how they're staying competitive. Yeah, but when we first started this conversation, we were talking about data. So, even if you buy into like AI functionality of an app that you're using or a platform that you're using, if your data isn't great, then it's still going to be a problem. So, I want to, I want to get back to what you said earlier about here at Paychex, like you're transcribing voice, you know, whatever. If you have some advice that you could give to business owners about cleaning up their data, like, what are you doing here? I mean, I am sure that you took this job and like any company, you had your challenges in getting your, the data here accurate and clean and whatever. What are you doing about that? What advice do you have? 

Beaumont Vance (20:17)

It's a tough problem, especially if we're talking about like CRMs. 

Gene Marks (20:20)

Sí. 

Beaumont Vance (20:20)

And if you're talking about manually entered data. 

Gene Marks (20:22)

And by the way, CRMs are the worst because there's a sales databases and salespeople are awful when it comes to doing good data. Look, I'm a salesperson. 

Beaumont Vance (20:30)

I have talked to many, many people in many audiences. In every talk I give, I always ask a show of hands of anybody who likes to do data entry. And I have yet to see a hand fly up in the air and somebody enthusiastically be like, me, me. Let me enter data all day long. Right. So actually this is one of my life's missions is to eliminate manual data entry. Definitely. And so, we're building tools. The flip side of the coin from the clients is what am I doing for Paychex employees? And one of my missions is to eliminate this drudgery of the manual data entry for all of the employees. Because usually what happens is people want the data and they get bad data in the systems because it's manually entered. And then so they flog the employees and say, no, you do a better job and the employees hate it and don't want to. And it's a big job dissatisfier and you never ever get to nirvana where you get good data. 

Gene Marks (21:14)

By the way, I have a client with a CRM system that they actually, they're a pretty well-known company that sells deli meats to delis, you know, like all around the country. And you imagine, you know, person selling deli meats. Literally there was like people like, they would have like men in old Chevrolets smoking cigars, playing with deli meats in their trunk and delivering it to like their, you know, customers. And these are not the guys to do data entry. You know what they did, they actually had, we set up a system where the salespeople, while they were on the road would call and leave a message on a voicemail box. And then they had like now they would just transcribe that message. I guess that's where, similar type of direction. 

Beaumont Vance (21:52)

Yeah. You know, so my wife has a small medical practice, and she worked for this one startup company for a little while, and they had just put AI in. It was all telemedicine. So, they had all Zoom calls, and they put AI in that would listen to the calls and it would enter all the information into the electronic medical records. And, you know, a large percentage of medical professionals quit the profession because of charting and entering data entry. And the data entry, it actually drives people out of the profession after they've gone to medical school and everything. Right. So it's no joke. So she started to work for this company and she didn't have to do that anymore. And I'm not kidding, she cried. She cried tears of joy. She was so relieved. And she stayed with that company as long as she possibly could because she didn't want to have to go back to the old world of having to enter data again. And so, I'd say, you know, for, again, it gets back to, for these businesses, what solution out there... Is there a solution out there that helps me get my data more organized, that captures it better? And now that we've got AI, that's happening more and more frequently. So what you're going to look for is you want systems that are integrated. So, if you're using, say you're using, using a platform like HR Recruiting Copilot, right, where you get candidates, you want that to integrate with your API, right. So that the data just flows straight through. Right. People are already doing this all over the place, even without AI. The system integration piece of it is very, very important. And the other important part of it is to look at the software that you're using. Does it capture the data the right ways? Now that we have AI, there's really no real excuse for having a lot of manual data entry. There's no excuse for anyone typing something from a form into a system. Right. There's scanning technology. You can scan resumes, you can scan order forms, whatever it may be. So, I would say the first place to look is look for systems that solve that problem and capture the data accurately for you. And then what's left over, I guess, flog your employees and tell them to do a better job. 

Gene Marks (23:46)

Do you think that voice is a big part of this? I'm going to get back to that same venture capital guy that I was talking to recently. He says they're investing in voice startup AI companies because he feels that's where the money is. 

Beaumont Vance (23:57)

You know, voice coming in or voice going out. 

Gene Marks (23:59)

Like voice coming in. Like being able to just immediately transcribe, recognizing it and then moving that data into where it needs to be, you know, into some large language model. I mean, that's what you're doing here. Right. I mean, you're transcribing a bunch of voice calls and putting that in. 

Beaumont Vance (24:14)

Yeah, absolutely. 

Gene Marks (24:15)

And do you think that, that, I mean, again, if businesses are looking to leverage AI, do you think that is, that's the place to leverage? 

Beaumont Vance (24:23)

I think it's a good place to look. I would say if you have to, you have to understand the AI revolution, sort of, you have to step back a little bit to frame it correctly. As I mentioned before, for about 60 years or so we've been focusing on information, IT as compute. Right. And it's been all numeric. Right. It's like we wouldn't be having this discussion about what do you do with your, with your accounting. Right, right. Like that's a solved problem for the most part. Right. But 95% or so of all data that's stored in business is actually linguistic data. And so, the place to look is where is their linguistic data. So, I think voice is a big chunk, especially for us. You know, we have, we have a call center. So, we take lots and lots of phone calls. Not every business is going to be voice focused. Right. What we're seeing too is more and more people are shifting to text and emails. So, a majority of our incoming contacts are through email. And we've seen that shift. It's not changing. Right. It's going in one direction. It's just more and more and more people prefer to write rather than talk to somebody for the most part. But I would say, you know, voices, if it's part of your business, is huge. But don't stop there. Look at everything that's written that you haven't in the past been able to like, compute. Because things you can't get in a spreadsheet, you can't get a contract in a spreadsheet, you can't get a sales order in a spreadsheet, you can't get customer information in the spreadsheet. All those things were real pain points for business. And once the data was in there, you could capture it, but you couldn't get it back out. Like I used to work a couple companies ago, I worked, we had 372 years of conversations stored all on, you know, all on servers that we were paying to store. And when I asked the question, well, what's the main dissatisfier amongst our customers? I don't know. Let's send out a survey. We got the data right there. How come we can't get it? You can't extract it. Why? Well, because it's just spoken word. So that's where voice is like super, super valuable. But then we transcribed it. Well then I got a different problem. Now I've got written communication that I can't put into a spreadsheet. That's where AI comes in. And AI is great for regulatory stuff, compliance laws, contracts, forms, speech, any, anything like that. And I would say right now, voice to voice to text transcription, huge. It's baked into almost everything. 

Gene Marks (26:33)

It is. 

Beaumont Vance (26:33)

It's a free feature, you know, of almost every platform. 

Gene Marks (26:36)

And it works quite well and quite accurately, as well. Couple more. I do want to ask you about privacy of data as well in a second, but before we even get to that, you had mentioned about call centers that you know that Paychex has. Right. I've seen that the biggest use of AI over the past couple of years by corporate America has been customer service and studies. McKinsey did one, Forrester did one recently saying about like it's like that is like the majority where the money is going to for customer service. And when I mean customer service, the example I can give is a Klarna, which is a buy now, pay later company and they have automated their customer service for like 85% of their calls now are being taken. 

Beaumont Vance (27:13)

Actually, the firm I used to work for was one of the early investors in Klarna. 

Gene Marks (27:16)

Yeah, right. I mean they're amazing. Swedish company, you know, so and they are, you know, they're, they leaned into it heavily and now they're having automated. You asked about voice in and voice out as well. And they're resolving calls way before is that, do you agree with that? You know, again, I'm talking about the, for the small businesses that are listening now, and the midsize companies is the customer side of things where they should probably be focusing on first. It seems like that's what you're doing here at Paychex. 

Beaumont Vance (27:42)

Again, I think it depends. It depends on what kind of business. Right. I mean it's very, very context dependent. Okay. If you're a landscaping company and you're working with homeowners, right. And you're not don't have a call center, you probably want to maintain the personal touch. Right. You probably have somebody who shows up at their house every weekend and that's where your touchpoint is, that's great advice, you know. Yeah. Because AI is really good for information retrieval. And if, let's say I built a really, really good, like the perfect AI call center solution, and you called and you didn't have to talk to a person, you asked your question and they just gave you an answer like you'd be thrilled. Right. But you might want to keep using that company because you can get your answer so quickly. Right. And you might develop... AMEX was this way. AMEX, actually, their philosophy was to develop a relationship with the technology. And as an AMEX customer, I can tell you their tech, Their tech is very good. And I'm reluctant to leave because they have very good user experience. I can get my answers very quickly without talking to somebody. But they also have a concierge service, and they also have really skilled customer service reps. And there is a point where data retrieval is not what you're looking for. That might be 90% of what you're doing in customer interactions, but there's a point where it goes beyond that. And if you don't have that, that customer interaction, especially for complex things, if your business has any complexity whatsoever, or any emotional component, like if you're a tax attorney or a CPA, right. People have emotion, get emotional, when they get, you know, letters from the IRS or whatever. Getting receiving information isn't what you're trying to do there. What you're trying to do is maintain relationships. So I'd say that that's one area. The other area, I would think, you know, where small business can really compete well now with AI is in analytics. Because before you needed to have data scientists, like, you can have great data, but you need some guys who did statistics and math with it. Right. We have analytics platforms right now. In fact, HR Analytics is a platform we just recently launched where you can go in and you can query your own data and just ask questions, and you can do it in a very natural way, because AI is doing it for you, and it can return the answers for you and basically give you. We started rolling this out in about 2011, 2012 at this company, at Fidelity Investments. And one of the things it was so good at, doing what a data scientist or an analyst did, that I had a friend who was teaching classes at a local college, and I said, you better tell your master's students to have a backup plan, because AI has gotten so good that it can give you insights to data so fast that it's better than having five dedicated data scientists working on it right now. And so, I think those types of platforms are where business can really start to compete. Because when you're running a business day in and day out, your head's down in the trenches. Right. You're not seeing the patterns. Right. You don't see if you got 10 employees, you're not seeing the patterns that that employee is starting to show an absenteeism pattern that's indicative of them being a flight risk or something or something else happening where there might be a flight risk. And my team actually spent a lot of time doing this where we've got tons of data, we can see the patterns across, you know, the entire SMB universe. And we can. And we built a tool, it's called Retention Insights. It does this. Right. But you can leverage a tool like that and query and watch it, and it can tell you about insights into your own business that help you run your business better. Right. And so, I think those tools are very... AI is great in those places, but I would be very cautious. I would look at it for cost savings, I would look at it for insights, I'd look at it for getting answers from data. That's what AI is great at. That whole world of humanity out there that makes business the glue that makes business stick together. Like, don't ever lose that. 

Gene Marks (31:20)

It's funny too, because I heard an interview recently with Sam Altman, you know, the CEO of OpenAI. And, you know, people talk about, like, oh, well, you know, is just everything going to be replaced by bots? And he said exactly what you said, that like, you know, humans want to talk to humans. You know, there's a balance between operations and, you know, HR and sales that, yeah, you can automate some things. You can get trends, you can get analytics out of some. You can have certain level AI bots answer questions. But ultimately people do want to deal with human beings. 

Beaumont Vance (31:50)

Yeah. And the CEO of Nvidia, I think, said, AI is not going to steal your job. The person who uses AI is going to steal your job. AI is like a force multiplier for human beings. And if for customer service, definitely give. Give the tools to your customer service people so they can answer questions quickly, efficiently, and be as helpful as possible. Right. But that doesn't mean get rid of the customer service person who's actually talking to your contact. 

Gene Marks (32:16)

Yeah. And your example also of, you know, depending on the kind of business that you're in. Perfect example. So, you know, my mom passed away recently. She has like, you know, like an annuity with like an insurance company. So, I had to call and find out, okay, what's the story with, you know, with this annuity and how does that get cashed out or whatever? And when you called that company, they were smart enough to know that it was an automated system. But the minute that you said choose three if you're reporting a death, when you choose three, immediately a human being came on. Because the last thing you want to do is talk to a bot about that. 

Beaumont Vance (32:54)

That's the best example I've ever heard. 

Gene Marks (32:56)

Yeah, I mean, they totally got it. And so, but it applies, like you said, even, you know, not in that extreme situation, but if you're in a landscaping business, I mean, people do want to talk to a human being about their, about their needs in their home and whatever. So that makes complete sense. Talk to me, Beaumont, about privacy of data. Talk to me about policies that you think a business should have, because this is, it's crazy stuff. And again, here at Paychex, I mean, you're doing analytics, you're leveraging a lot of data from hundreds and thousands of hours of phone conversations. Who knows that private data could be exposed? 

Beaumont Vance (33:35)

Right, sure, absolutely. It's a big question, one that goes back for a while. I actually used to work for Sun Microsystems, and about 20 years ago, I don't know if most people know this, they invented cloud computing, and our clients were mostly medical, medical insurance companies and Wall Street financial companies. So, lots of regulatory constraints, very, very sensitive data. The most sensitive data there is. Right. And one of the reasons that the cloud computing at Sun failed was because at the time, people were not ready to allow their data to be moved off of their on-prem servers into somebody else's servers. And they were, they were completely secure. And as a matter of fact, I was the head of enterprise risk management at Sun, which is why I was on the, on the team, was to look at the risks involved with data getting out. Because back then it was really new days of big data and data breaches. We hadn't had a lot of the big data breaches yet. And so now, fast forward, everybody's on cloud and now we've got AI and there are real concerns people should have. And I put them into two categories. One is your data going out. So, Samsung famously found this out the hard way. Their coders were using, I forget which platform. They were using an AI platform to do the coding for them. And they would load their proprietary code onto the platform. It's a cloud-based platform and say, hey, can you fix this code for me. And the AI does it really well actually. But the problem is that data then became a training set for and became publicly available to everyone else using the platform. So if I went in and said, hey, can you just code me up something that does a Samsung thing, it would give me back the Samsung proprietary code. Big problem, big problem. So you have to be very careful about who you're dealing with. And really the important thing here is counterparty risk. You have to be aware of where your data is going and who you're giving your data to. And in a world of 70,000 AI startups globally, that means you have to work with trusted providers. 

Gene Marks (35:33)

How do you figure that out? I mean, even if you're say I'm giving my data, I'm in Google Workspace or I'm in, you know, I'm in Microsoft Copilot. And by the way, all these companies they outsource, I mean they've got, like you said, some of those 70,000 technology, whatever. What do you do if you're a business and you're just concerned about your product, how do you figure out who's trusted or who's not? 

Beaumont Vance (35:54)

You ask and you know, get us back to AI. Ask the web. What I would do personally, I would Google this and I would say, what are the privacy policies and data use policies for Microsoft Copilot. 

Gene Marks (36:06)

I see what you're saying. 

Beaumont Vance (36:07)

And it will return to you with the answers on. They're very public about this and as it evolves, people have been stung by this, lawsuits have been filed. So, and they've got teams of lawyers working on, working on this to protect themselves and to protect their clients. So they're very, very public about how they're protecting you or not protecting you. If there's a question, what I would say is don't share your data. Right. If they're not going to give you reassurances, don't share your data with them. That doesn't mean you can't use AI to answer questions. Just don't load your proprietary information into it. Don't say, you know, here's all my clients or my employees' Social Security numbers and wages and addresses. Can you tell me like, should I send them a Christmas card? 

Gene Marks (36:45)

Right, right. 

Beaumont Vance (36:46)

Something like that. Right, right. So, so that's the one side and that's going out. The other side you have to be careful of is, is coming in. 

Gene Marks (36:52)

Yeah. Your own customer data, you're saying? 

Beaumont Vance (36:54)

Well not just your own customer data, but I mean, when you're asking AI questions and it's giving you answers, or you're running your business on an AI platform, where is the AI platform getting those answers? What was it trained on? 

Gene Marks (37:05)

Good point. 

Beaumont Vance (37:05)

Right. So, so I could today, if, if I did not have the high standard of ethics that we have at Paychex, I could just have a... I could just take chat GPT, relabel it and say, here, here's a new HR tool. Go Ask, Go Ask HR. And you could ask it and I would just pull whatever, whatever came off the internet. Right, right off the public information, and it would be as good as what you would get off the internet, which is to say, not that good, not reliable. If you're relying on that, then you're going to have a problem for a business. On the other hand, and this is where, you know, one of the reasons I joined Paychex, I told you I came because it really attracted me that they had such an incredible database. If we're going to be building AI tools, we're going to build them on the best database in the HCM industry. That data we can vouch for, that data is clean, that data is protected. We have some of the highest standards of ethics in the industry. And when we build a tool, we know that we're going to try that out and we're going to test it and we're going to make sure it's 100% before we roll that out to anybody else. And you know, for example, the HR Copilot I've mentioned a few times, that's what we use for our recruiting internally. We're not going to be giving our clients something that's a slapdash kind of effort that we can't vouch for. And because we have the data, then we can vouch for it because we know what we've got, it's our data, and we're in the spot we're sitting in. We're the only people who can do that because most people don't have the data. So I'd say for anyone, if you're a small business person, all you need to know about AI, two things. Ask what they're doing with your data and then ask where they got their data. And if you know those two things, you know whether you've got a solid AI company. When I worked for private equity, we used to review over 200 companies a month, and a lot of them were AI companies. In fact, everybody started putting AI into everything a few years ago, right? And the first question, I was on the due diligence team, I would ask whenever we had a new founder in, I would say, they would tell me about their AI and how smart it was and how it was better than Google's and it was better than Open AI and everything else. And I would say, where's your data? Where'd you get your data? Where's your licenses for the data? Where'd you get it from? You know, how do you clean it? How do you... How do you... How can you assure me that your data is clean and it's correct? And if there was any stammering or hesitation, then no, not an investment, because anything, anything downstream from bad data is bad. You can't fix bad data with good AI. 

Gene Marks (39:28)

Great. 

Beaumont Vance (39:28)

You can, you can fix bad AI with good data, but you can't do it the other way around. 

Gene Marks (39:33)

Well, my fans, I want to thank you so much. It was a great conversation. I learned a lot about AI we're which I will summarize at a later point. But I really appreciate your time and thanks for joining. 

Beaumont Vance (39:42)

It's been a pleasure, Gene. Thank you. 

Gene Marks (39:45)

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 (40:22)

This podcast is property of Paychex Inc. 2025. All rights reserved.