Scratching the Surface: Real‑World AI Use Cases in Business

In this episode of On the Surface, Seth is joined once again by Brad for a wide‑ranging, practical conversation about how AI is actually being used in everyday business—and why most people are still only scratching the surface. From Copilot and ChatGPT to Claude, Gemini, and agentic AI tools, Seth and Brad break down what works, what doesn’t, and where confusion still exists as these technologies evolve at lightning speed.

They dig into real, hands‑on use cases rather than hype: creating presentations and emails faster, analyzing data, building Excel models, summarizing long documents, turning contracts into podcast‑style audio, and setting up automated systems that pull information every week without repeated prompts. Brad shares firsthand experiences using AI agents to monitor industry projects and generate reports, while Seth explains the differences between creative AI, analytical AI, and autonomous agentic systems—and why prompt quality still matters, even as the tools improve.

Together, they tackle the bigger question many organizations are quietly wrestling with: what does AI mean for jobs? Seth and Brad offer a grounded perspective on efficiency, automation, and workforce change—arguing that while AI won’t replace entire roles overnight, it will reshape how work gets done. They explore which tasks are most at risk, why repetitive and administrative work is the first to be automated, and how early adopters who learn to use AI effectively will be best positioned as companies inevitably move from experimentation to expectation.

The conversation also highlights the human side of adoption—resistance to change, security concerns, and the reality that many people don’t even know the tools they already have can do these things. From personal examples like planning vacations and visualizing home renovations to professional workflows that save hours or days of work, this episode emphasizes one core takeaway: AI is just another tool—but curiosity, experimentation, and willingness to adapt will determine who benefits the most.

It’s an honest, approachable discussion about AI in the present—not sci‑fi futures or fear‑based narratives, but the real efficiencies, limitations, and opportunities unfolding right now for people willing to engage with the technology rather than ignore it.

Thanks for listening!

Check us out:

deltacos.com

Instragram

LinkedIn

Transcript

SPEAKER_02 0:04

Welcome back to On the Surface. I'm your host, Seth Stevens, and this week Brad is back with me to discuss real life use cases for AI in everyday business. But first, I just want to take a moment to thank everybody for listening to the podcast so far. Please make sure that you're following, rating, and leaving comments on our show on whatever platform you're listening to. This helps drastically circulate the podcast. It shows some love. And uh please make sure that you are sharing uh the podcast episodes on social media, LinkedIn, Instagram, Facebook, whatever you use. Make sure to give us a shout, please. It helps a lot. All right, let's get into AI. So, AI, it is confusing. What are you using it for? Are you confused about it the same way I am? Yeah, it is super confusing because there's so many different platforms and softwares out there. And uh they're all good for different things. So, like the co-pilot is good because we use Office 365. So it ties into PowerPoint, Outlook, Word, Excel kind of. It's getting better, like it's drastically better than it was six months ago at all those things. And it'll help you write emails, help you edit Word documents, create PowerPoints, whatever. Like it's a super efficient solution for getting that stuff started.

SPEAKER_01 1:53

But if you don't know, like creating PowerPoints and things like that, if you don't know how to put the right prompt in, it doesn't give you something that's very usable. Yeah. I mean it's gotten better since I've started using it, but you've really got to find it.

SPEAKER_02 2:10

Well, the better it gets, the less polished your prompts can be. Yeah. Like over time. It's getting better. Um outside of Copilot, I mean, I've used Claude quite a bit for like problem solving stuff and analysis. It's pretty good at making Excel files and like creating a multiple tabs and formatting it and stuff so that it's actually a usable Excel file. It understands like all the math and the problem solving. Um, and then I've used like ChatGPT for a lot of like creative type stuff. Like Claude is not great about asking it to create podcast episode titles or social media posts or whatever. And chat is really good for that. Like you just put in a prompt about, you know, maybe a tone or a feel that you want and like the information, the baseline information, and it'll come up with something really good usually.

SPEAKER_01 3:21

Still just scratching the surface, those types of things. Yeah. Right? It's still editing your your thoughts. You're still having to put in quite a bit of work. It's not actually creating after the fact or doing a task for you. It's editing. It's like uh it's like the difference between uh whatever that is and an AI agent. Oh, well, yeah.

SPEAKER_02 3:46

Yeah, for sure. An AI agent is actually training them to do the work. It's kind of right, which is what kind of like robots or like an RPA was.

SPEAKER_01 3:55

Which is what I I guess that's the newest or the best version of it right now.

SPEAKER_02 4:01

Yeah, like replicating repetitive mundane tasks because you're training it to do something. Well, does Manus use Agentech?

SPEAKER_01 4:12

You've used Manus? I've used Manus. I mean, it's it's really cool. I actually had it what I was wanting was in a weekly email of an intelligence brief. Say on Monday, I want to know all of the uh projects um that could involve construction or aggregates in the few states that we operate in. Um if any new information on existing projects in the area, uh who who those projects are being done by, you know, contact information, stuff like that. That's that's what uh my prompt, I had all of those things in Manus. And then it asked me if I wanted to create a website. It created the website in about five minutes, and it will send a weekly update directing me to that webpage with all that information.

SPEAKER_02 5:07

Yeah.

SPEAKER_01 5:08

Yep. I think Mana is known for their web website or web page creation.

SPEAKER_02 5:16

Okay, yeah, yeah.

unknown 5:17

Yeah.

SPEAKER_02 5:18

So a lot of its solutions is hey, I'll just make you a website. Yep. But I th I mean, I'm no expert. I'd have to even look this up. Maybe I will real quick about what I agentic AI really is because I feel like that's getting into some of it where it's like now it's not waiting on you to give it a prompt. It's automatically working each week to pull new information. Learning from information. Yeah. Learning from the things, the data that it's collecting from the session.

SPEAKER_01 6:03

But I don't know.

SPEAKER_02 6:07

I'll have to, my internet's being slow.

SPEAKER_01 6:10

I still don't I feel like we're not being like our company in particular is not pushing the use of it. They want us to utilize the tools that we have in our program, copilot tools. Because it it can help you create emails and presentations and things like that.

SPEAKER_02 6:26

Well, I think it's a great introductory efficiency tool for AI, plus it's secure. So copilot is secure with our data because it runs through our Office 365 and it doesn't go out into the world wide web for people to see and for AI to reference and stuff like that. Whereas open AI, like anything that we've been talking about, Claude, chat, manus, is all using your information and like storing it. So you have to be really careful about putting sensitive data in there. Like you don't want to put our company name and like financial information or you know, personal information data, like anything payroll or social security numbers or anything like that. And honestly, the majority of the world just is not uh self-secure enough to handle that. So, like as a company, we have to say you can't use that officially. Like you should use Copilot because we still have people click on test spam emails in their entire and then you have 50% of the people in the company that don't want to learn something new.

SPEAKER_01 7:50

Yeah, they don't they don't want to go through the pain of having to learn how to use it. There's there's that, and then there's also belief that it can't do my job better. It can't make me, it can't help me do my job.

unknown 8:04

Yeah.

SPEAKER_02 8:06

Yeah. There's some people that are just like resistant to change or scared of change because of what it might bring. It's the same it's like kind of the same mindset of people saying, well, I'm not gonna train that person because they're just gonna know enough. They'll one day know what I know and then they'll replace me. Well, that's not realistically how it's like uh so now instead of it saying, um instead of people saying, oh, well, you'll be able to that if I train that person and give them all my knowledge, they'll replace me. Now they're thinking like, okay, well, I'm scared of AI because it could potentially replace me. So I'm just gonna act like it doesn't exist and then we'll be fine. But realistically, you're kind of just hurting yourself because the more that you would know about AI and your position, the better positioned you would be because it's going to happen to keep your role and job.

SPEAKER_01 9:06

Because the technology is coming. Yeah. And there will be a point in the future where we are not not just asked to experiment with it, but mandated to figure out how to use it to do certain tasks.

SPEAKER_02 9:19

Yeah.

SPEAKER_01 9:20

Yeah. Yep. 100%.

SPEAKER_02 9:23

I did finally get it pulled up. Manus is like exactly agentic AI. It's like it says refers to autonomous goal-oriented AI systems that move beyond passive content generation to actively reason, plan, and take actions using external tools to achieve complex complex objectives with minimal supervision. So it's kind of like the entry level. I wouldn't say like pulling market data is complex, but it's working by itself with no supervision in that case on a weekly basis. So you created the first Terminator.

SPEAKER_01 10:14

Do you think that I mean a controversial question? Will it replace, can it, could it replace jobs in our industry? Yeah, for sure.

SPEAKER_02 10:24

I mean I think it's like a question that people continuously ask and try to address. And a lot of depending on who you talk to, a lot of people say like it's not going to eliminate jobs, it's going to change jobs. And other people will say, like, the scare tactic of it's going to eliminate a bunch of jobs. I mean, I think realistically it's somewhere in the middle, like most things are. You can't become more efficient and not eliminate jobs because jobs, people and jobs are the highest cost typically of any organization. So you can't become more efficient and then expect to just carry all that overhead cost. But I think it won't be like, okay, well, we used AI and automation to change or replace 50 jobs, so we're eliminating 50 people. I think it'd be like, okay, well, we used AI or automation to change or adjust 50 jobs, and now we're more efficient at that, but now we need to monitor these areas of work, or we need people to prompt AI to do these things, or we need people to focus on people or actually physically being somewhere so we can reduce our headcount by five of the 50, right? Like the other 45 people, we need to know AI and we their job is just different. So I think it's somewhere in the middle.

SPEAKER_01 12:03

And those that are early adopters are learning about it, yeah, searching it, figuring out how to use it to help our company will be the ones that have those jobs. Right. What do you think? Or any company. I agree with you. I definitely think that you know this it's gonna increase efficiency for sure for a lot of different positions. And I mean you increased efficiency has to result in less people um for a business for it to be a successful efficiency increase.

SPEAKER_00 12:38

Mm-hmm.

SPEAKER_01 12:40

Or less cost or more business. You know, maybe you do the same, maybe we're able to utilize this this technology to increase our the scope of the work that we do or the footprint of the work that we do. Um and and grow our business with the same amount of people. Yeah, that's a good point. Yeah. So, you know, that's that's the way that's what I would want to see. Yeah, for sure. But it's I mean, increasing efficiency and reducing costs are are things that this is gonna bring, for sure.

SPEAKER_02 13:14

What do you think becomes like some of the first stuff that's that it's it's realistically good enough or comprehensive enough to replace?

SPEAKER_01 13:28

The first positions?

SPEAKER_02 13:30

Yeah, I guess. Like, I mean, realistically, it's not all gonna happen in one position. Does that make sense? It's gonna be a bunch of different processes, but maybe that allows you to come to take all those out and then combine remaining processes into one role.

SPEAKER_01 13:48

I think the easy answer and the answer that everybody probably already knows is these uh monotonous um like positions that or functions, tasks that are repetitive. Um they're that's almost busy work for a lot of people. Uh it's a it's the same process, it's just repeated over and over again. Transactions, a lot of what comes to mind honestly is administrative type work, accounting type work. Um data entry. Data entry. Yeah.

SPEAKER_02 14:28

Um data management. I mean, really, it should be able to like look at old orders or data and prompt questions to operations or handle it itself, where it's like confirming whether an order needs to stay open or be closed, basically. And today a person is having to do that and say, oh, well, this stuff is old, or should we close it and then actually go physically close it and yada yada. Like all that data entry stuff is very basic and should be able to be automated, right?

SPEAKER_01 15:13

Right. I think what I'm most excited about right now is just the amount of time that I think that it could free up for a lot of folks that sit in management or uh office roles that are getting a hundred emails a day and have a lot of meetings and are having to, you know, look at different documents and things like that. I mean, just uh like for myself, I've created a couple of prompts, scheduled prompts throughout the week. Like Copilot will send me on Monday morning, it will send me uh a list of the meetings that I have for the next two days. I've got it set to where it sends me this, like a couple days in advance. That's a good idea throughout the week. Like here's a list of your schedule for the next two days. These are the meetings that you need to prep for. Um this is all the information that I can find in your email that has to do with this certain so if you wanted to review old emails that had to do with that topic or documents that had to do with that topic, you know, you can do that through in that in that's really good. And then like the documents. So if I wanted to review a a lease agreement that we're working on, in Copilot right now, you can go to it and and have it create an audio transcript or audio version. It's like a podcast. It's got basically two people talking back and forth and telling you what is in this document. Really? In a summary. Yeah. So it like I took a lease agreement that's 20 pages yesterday, and it created an audio file for me to listen to. I'm a I'm an audio. I learned. How long was it? It was 12 minutes. I didn't even know you could do it. It was like two people talking back and forth saying, This is what's important in this section, this is what's important in this section. Yeah, that's legit.

SPEAKER_02 17:11

It's pretty cool. I didn't know that it would turn it into audio. I mean, I've used it kind of similarly for like a summary, like this is what's important.

SPEAKER_00 17:18

I think it's a new function.

SPEAKER_02 17:21

That's super cool. Um, I think to your point, we're still like a ways. I shouldn't say a ways. This happens like exponentially faster every time we go, but I think like we're seeing the possibility of, you know, like reducing positions or fully changing jobs, but it's not necessarily there currently. Like I think in present day we're still like pretty squarely in efficiency boosting phase. Like you're taking management roles, like you said, or executive roles, and historically they'd have to delegate a lot of things, or it would take them more time in the day to do all of the tasks that they were trying to handle. But now you can like significantly increase that because you can get a lot of the groundwork taken care of, you know, for you. Like you don't have to go read a 20-page document at pretty quickly. Gives you like a 12-minute audio podcast file that you could be like working on something else if you're good enough to do that, or like up walking around or something uh without having to, you know, grind through 30 minutes of reading and then rereading since it's a legal contract.

SPEAKER_01 18:51

But this stuff is happening so fast that most people don't even know. I say most people, most people in our organization, our company, don't know those functions even exist.

SPEAKER_02 19:00

Yeah, in the programs we have. No. You have to kind of just like keep trying stuff like what you're doing. Yeah.

unknown 19:08

Yeah.

SPEAKER_02 19:09

Like you're putting it in and saying, hey, what can you do to make this easier on me? And I've tried to do the same thing is just like have tasks on my list to create like a dashboard or create an analysis, and like a decent amount of time would be creating like a standardized file or just laying the groundwork. And I just rather than doing that for a day, I've just spent five to ten minutes like creating a detailed prompt into a software and then it giving me all of the groundwork. Like some of the formulas and the things inside of it may not be correct, but like the thought process behind it was correct. So it has a lot of the correct um like outline, basically. And I saved a day's worth of time from creating an Excel file and putting formatting in it so that it's readable and all this stuff, and now I can just focus on like the correcting formulas and the actual data analysis and like entering data, changing things for different scenarios.

SPEAKER_00 20:28

A lot of time.

SPEAKER_01 20:31

And again, we don't know. We're just playing with it at this point.

SPEAKER_02 20:35

Yeah, no, I'm not that good at writing prompts, honestly. Um, I kind of take the approach of like treating it like somebody on our team or an employee, and I'm just trying to give it like a paragraph's worth of stuff saying like, here's what I'm trying to get to. Here's some of the information that I know as background that I would think would be helpful for you. Because you have to give it like scope and some context, right? If you want something good to come out of it. But I still don't think I'm good at writing prompts.

SPEAKER_01 21:11

Oh, I'm not either at all. I don't spend enough time. They they tell, I mean, each I guess each AI platform has different settings, but in Copilot, I noticed there's a setting that you have to change for it to remember what you the information that you put in. And I've learned that it says that you should go in there, change that setting, and then tell it who you are and all about you and what you do.

SPEAKER_02 21:37

And oh yeah.

SPEAKER_01 21:39

So it knows to talk to you. You don't have to reinform it every time.

SPEAKER_02 21:42

Yes. Yeah. Yeah. The benefit of Copilot being connected to 365 is that and the data being secure is like you said earlier, it has all of that background that it can like search through and find for you. Um and it's all safe to use because it's secure on our work stuff, so it's uh you know through teams, referencing peace agreements and emails and things like that for you. I have noticed that like even if you use an open AI, and I'm not putting, you know, like private secure data in it, but it will it can reference back to chats that you've had before. So like it is still keeping some of that background. Like it's not like you're talking to a to totally new person every time. Every time. Okay. Uh what else is out there? Gemini. I haven't really played with Gemini.

SPEAKER_01 22:46

I use Gemini more than Chat GPT. Just like Oh, you swapped over? I don't I don't Google things anymore. I go to Gemini. Yep. What do you what uh here's a good question, could be interesting. What do you use it for personally? What's what's something cool you've used it for in your personal life?

SPEAKER_02 23:07

Uh so we've used like Chat GPT and Gemini because we're trying to do some updates to the exterior of our house and stuff, like needing a new roof and gutters and things like that. So it's sometimes it's hard to envision colors, but Chat GPT and Gemini are pretty good if like you take a picture and then you tell it what you want changed or recommendations that you want for color schemes. It's good at like image creation. Yeah. So it'll make a new photo with the changes in it. And it most of the time it looks pretty good. Yeah. Uh you have to be careful about telling it to adjust things from that point forward, though, because it um gets a little confused. Like you're almost better off to start a whole new chat with it if you want it to test something else on the photo.

SPEAKER_01 24:07

I did the same thing. We have a our kitchen table has white chairs, and Natal was looking at black chairs. And I took a picture of our table and told it to change our chair color to black. And it did, and we decided not to buy a black chair. This is pretty cool.

SPEAKER_02 24:25

It's a great because think about what you had to do before. Like you could still test it in you know, prior to AI, but you have to go to the store and like buy three chairs, bring them home, see what you think, and then if you don't like them, you gotta pack them back up, you gotta return them. It's just a whole hassle compared to taking a picture and saying, Hey, uh Jim and I make these chairs black instead of white. I'm gonna see what it looks like, and bam. Um, an example for like Claude is so I've kind of gotten into watches, like analog watches. And so I have some and had been wanting to make like a list of them and keep track of like the rough dates that I purchased them so I know when to service them in the future, which is like typically every five years or whatever. But I similarly, that probably take you know an hour of work to put together and like have it nicely formatted in like a Google Sheets file that I can just put in my Google Drive. Uh, so I spent five minutes writing a prompt just saying, hey, I want like a tracker style list. Uh here's stuff that I own, here's some stuff that I would like to own in the future. And that's really, I just said I want to know, like, I want to be able to see purchase date, maybe reason why I purchased it, next service date, that kind of thing. I didn't tell it like when I wanted it serviced, whatever, and it just went out, was able to find all those watches based off of me putting in the make and model, knew what type of movement was in it, knew that you should get automatic watches serviced every five years, used the purchase date that I gave it, made this whole like Excel file table that said, here's your watch, here's when you bought it, here's why you bought it, here's your next service date five years from the date you bought it. It's teaching you things.

SPEAKER_01 26:40

Yeah.

SPEAKER_02 26:41

Yeah. For sure.

SPEAKER_01 26:42

Yeah.

SPEAKER_02 26:43

And then like left spots for notes. Uh, it may have said, well, this watch is quartz, so you don't need a service or a movement service. You just need a new battery in like two years instead of we're gonna have to have a whole nother podcast on watches. Uh yeah, I think there's sounds like plenty of those already. I think there's plenty of those. What do you use at beside your table chairs? It took you five minutes to do that. Yeah.

SPEAKER_01 27:13

I mean, it saved me a ton of time. I've used it for like vacations, trips, planning trips. Oh, yeah. Um, I I want to do, I don't have it scheduled or booked or anything, but I was playing with the idea of renting an RV and and going driving west and hitting some of the parks and then driving home. But I had it map out, and so if I want to do a seven-day, ten-day trip to Yosemite from Paraguld, you know, where should I go? Where should I stay? Where should I eat all of those things particularly? I think it'll tell you ten days isn't long enough. Yeah, it ten days isn't long enough, but it it did make a plan around what I told it. Yeah. That's pretty legit. Like how long it would take to get from here to here and where to stay and the reviews based on reviews.

SPEAKER_00 28:07

Mm-hmm.

SPEAKER_01 28:08

Really cool. I've heard of people using it for like logistics type stuff like that, and I've never done it. The agents will then like it's to where it's to the point now to where it will go and book everything for you if you I didn't even think about that. You can have it reserve and book things.

SPEAKER_02 28:27

Like the recommendations and stuff. Yeah, you can probably find decently or the deals. That's sweet. I I didn't even think about that. It's like a super advanced version of MapQuest. Super advanced. Yeah. Yeah. Pretty cool. What else?

SPEAKER_01 28:52

Any final thoughts around AI? I mean, uh like I've been telling our team, we need to be playing with it. Everyone needs to be. Figuring out how to how to use it if not it work personally. Just play with it.

SPEAKER_00 29:10

Mm-hmm.

SPEAKER_01 29:11

Because it it's definitely going to affect our lives. It's going to affect the way that we work. And it's happening fast. It's just another point.

SPEAKER_02 29:19

It's just another tool. Like you're use people are using it in personal lives, probably more leading up to this point than in business. But I mean, I think pretty much all of your skills that you've learned in the past are because you've been curious about it and like played with it even personally. So I don't think it's any different than that. That's right. You just gotta, it's just new and different, and you kind of gotta get into it. One way or another. If y'all enjoyed the episode, please rate our show and leave a review on Apple Podcasts, Spotify, or wherever you listen, and check out Delta on all social media platforms at Delta Companies and our website at Delta C O S looks like Deltacos.com. Thanks for listening, and we'll see you next week.