Episode 100 - Pat Barry

Erik Martinez:Hello and welcome to this episode of the Digital Velocity Podcast. As we start 2026. I wanted to do something different and talk about where AI is headed. 2025 was a huge year in AI as we saw a shift from lots of experiments and companies to the proactive use of this powerful technology into business operations.

The question everyone keeps asking, what should we expect as we move into 2026? Today, I'm excited to be joined by my friend and fellow AI nerd, Pat Barry, who'll be helping us unpack what to expect in 2026. Pat, welcome back to the show! Thanks for [00:01:00] having me, Erik. I really appreciate it. Pat, for those listeners who don't know you, would you give us a brief intro into who you are and what you're doing?

Pat Barry: Yeah. Uh, my name is Pat Barry, currently the president of AI Consulting Partners. We work to do a variety of things, train people on AI, build agents, workflow strategy around how to bring AI in, and most importantly, fractional work around data and AI. So, spend 20 years in the data science field working for big companies like Discovery Channel and Google, then moved into the agency space where I was able to work with clients like Unilever, McDonald's, UnitedHealthcare, whole slate of Fortune 100 companies.

So, excited to be here.

Erik Martinez: The only thing that matters is super AI nerd.

Pat Barry: Yes. That's the biggest thing.

Erik Martinez: All right. Let's talk about 2026. If you thought 2025 and maybe 2024 were the years of experimentation, what do you think 2026 will be?

Pat Barry: I think the more advanced companies that did their, hardcore experimentation or at least like, Hey, we bought chatGPT gave people [00:02:00] kind of a path to use it. I think you'll see some more advanced companies actually moving to managing a digital employee, like a real agent that has sub-agents underneath it. That's for the companies that were early adopters and we're probably doing stuff in late 2023, all of 2024 and 2025. That's where I can see things going. At the same time based on what I hear from people, a lot of folks will not be doing that, still deciding how to bring it in, what to do. Um, but I could see companies not hiring because they have an agent now to do that particular job or tasks.

What do you think it's gonna look like?

Erik Martinez: I think you're right. It's gonna be the year of automation. Automation is now a buzzword in our industry. When it comes to automation and agents, for those who don't know what an agent is, just think of, a little digital assistant who does a very specific thing for you.

A good example might be, every time an email comes in that has this, I would like a response that says [00:03:00] that. That's a simple example of a very unsophisticated agent. Right? And then we can build way more complicated ones. So, back to the answer, the answer is automation.

But what does that mean? I think the challenge for most people is, what does that mean in the context of AI within their businesses? And I think that's one of the things that we're gonna have to address and educate each other on what are the steps to actually implement AI? Because it's not software, right?

If I give a program that's designed to add and subtract, it's gonna give me the same answer every single time I give it the input.

Pat Barry: I sure hope so. Math is pretty straightforward, so yes.

I would agree. You need to be a great writer and be a great explainer of instructions, in English words, again, like I know we were chatting before this, I was showing you building a Gemini enterprise, and you can see it's getting there, but I have to give it very explicit instruction. So, I think too is people get more comfortable with it they'll be [00:04:00] able to do their own automations, build their own agents, and I think I could see confidence just picking up too. I know that's something you and I talk about quite a bit, is just like coming in and helping people get confident with the tools. Like open 'em up and use 'em. So again, I'd see confidence being another thing too.

Erik Martinez: I think the other thing, for 2026, is training. You'll hear the word upskilling a lot. Whatever term you want to use, teaching people in your organization, how to do basic prompting. Once they understand how to do basic prompting consistently and what the different types of outputs that they can get.

Because what they'll find is when they put in a prompt. They put in that prompt a week later, you're not gonna get the same answer. And that something that's a little bit disconcerting at first, but it's actually one of the reasons these tools are so powerful. So, teaching people how to prompt, teaching them advanced prompting frameworks that they can use before you even start to do automations.

You need to learn how to do [00:05:00] basic prompting, and I think training your staff and your team and finding people like Pat or myself who can help with that, is a really important thing for 2026. Is there anything else that you see on the horizon for 26?

Pat Barry: It's hard to tell, everything changes so rapidly. Google made big announcements just two days ago about new products that they're launching. We could go on and on about what might happen. I think it's hard to look at all of 2026.

I tend to think in terms of quarters. So I guess in my brain right now, I'm just thinking about Q1 2026. I don't know what is gonna happen after that. But, I think a lot more about workflows and business processes. And I know kind of getting to our next question, what do you think, like workflows or a business process, become more automated in 2026?

Kinda staying on the automations thing.

Erik Martinez: Yeah. I think, depending on what type of business you are, number one friction point that we see, is communication in an organization is fragmented and inconsistent. Even if you were the best organization in world, the idea of taking on a complex [00:06:00] idea, disseminating it to your team and distilling it in a way that the entire team understands and is moving in the right direction, is a little more art than science.

But, I believe that if you look at that workflow of any project, or any set of tasks from beginning to end and really scrutinize the steps that are involved, you can figure out opportunities to improve clarity through simple automations using AI tools to make sure that clear understanding gets distilled down to the entire team in a more effective manner.

The benefits, when we have more clarity we're gonna be a lot more efficient about how we do it. And I'm not one of those people that believes in efficiency for efficiency's sake. I believe in efficiency for the opportunity for my team to do more and better. That's kind of what I think. So, I'd start with communication 'cause it's the most fundamental thing that we do every [00:07:00] day in every business and any department.

What do you think Pat?

Pat Barry: I mean, comms are huge. Just being able to distill, large amounts of text data, whether that's, communicating deep research to a group and being able to sum it up quickly. It just depends on the business you're in, size of the company, how advanced you are, like what have you done Prior to this. I could see companies easily having full on agents to solve real problems, instead of somebody submitting a help desk ticket to, somebody in IT, depending on what that ticket is. I can see that help desk now being an agent. Routing it to someone that can solve it or solving it itself, depending on the issue. If it's a simple digital task, like somebody's locked outta their computer, that's something you can program an agent to do is go back in the backend system, reset their password and give them a new dummy password till they could go and reset their own.

It's tough though because that's for the advanced company thinking about this. I could also see it being dependent on what industry you're in. An advertising agency is gonna have much [00:08:00] different, more automated business processes than a logistics company. They do two totally different things. I could see the more digitally based companies, that do strictly digital business, whether that's SaaS, you know, whatever. I think they'll have more opportunities to automate just because it's digital.

Whereas your more brick and mortar type of companies will have less opportunity. It's just the field that they're in. But, I think it varies by adoption. How confident are the people there? Are they familiar with the technology or do they still think it's, kind of outlandish?

Erik Martinez: Which it is to a certain extent, right? I mean, we've seen some weird things. I've seen some weird things, but it is improving. I think you're right. In a world of E- retail specifically, you know, I think there's some really, really cool things that are happening.

Streamlining customer service with chat agents. I think it was Open AI announced, a search tool. Not their agentic browser, but a search tool feature within chat GPT. I have not played with this yet. But, there's [00:09:00] an agent that they have created that will ask you questions about what you wanna buy, and then it goes and finds the data, the research, the comparisons, and brings it back to you. And I've been talking about this for about six months now. Like, this world of agentic shopping, automating the shopping process for the individual consumer.

So, this is for everybody who's listening. You, are probably not gonna be going to the websites as much as you used to. This is just the very beginning. So, what does that look like in terms of your business processes? That's one workflow that's gonna get automated, but it's gonna get automated on the consumer side, which is gonna force businesses to rethink how they do a lot of things. What information needs to be on the website so that the agents can talk to each other intelligently and bring back the right set of information. I think that's, kind of back into that, what will 2026 bring? It's here, and it's gonna get more prevalent over time. So let's pivot to agentic [00:10:00] topic for a second.

What do you think agents are gonna be realistically capable of doing in 2026? And what would be the measure of value?

Pat Barry: I would say into more complicated things, like a multi-agent system. I know you and I have talked about that. We've built a couple things together that do that. I mean, you'll have some sort of intake agent let's stay with our ad agency, flow. Somebody sitting there and writing 50 different social media ads, you can have your main agent write the ads and then send the ads to your legal agent that understands all your compliance, privacy policy stuff. If the legal agent disapproves the ad, it just sends it right back to your writing agent to rewrite it based on the notes, then back to legal.

That loop might happen two times, 20 times, 50 times. Who knows? But I think those type of efficiencies, that's just a small example, but think about content creation. I could even see something similar with image and video creation. Those types of things can definitely happen.

Erik Martinez: We're working on something like that, right?

Pat Barry: I was gonna [00:11:00] say, both you and I know exactly, what could be done with image creation. In terms of value, that's the hardest thing, when you have somebody that has been doing this one task process for so long and

all right. I get my time back. Okay, that's great. If you're in an ad agency, we can look at time sheets, hourly rates, and determine, like calculate ROI, just based on that. But, for businesses that's, the hardest part. The businesses that don't have time sheets or aren't contract heavy, they just have employees. They get salaries. Like how do you measure that? I think there's obviously I'm data science background, so all I think about is exact metrics and calculations but I think you're just gonna have to move to some sort of survey based, value system.

Like, are the workers happier because they're more efficient? Are we getting more strategic opportunities done? Can the same individual do more work like take on another client, another project, another account, and remain at their same level of happiness and pay and not notice anything? Those types of things, they're harder to put a number behind.

Like I can sit here and calculate ROI for anything. But, [00:12:00] the assumptions you have to make along with factors coming up that gauging the value won't be a number. It'll be some sort of other survey type system says like, Hey, we introduced AI, our employees are happier.

That's great value. Our clients notice we're getting things done faster from client surveys. I think you'll move away from, not necessarily move away from, but they'll be less people banging on my door going tell me what the ROI of this is. But again, I think you'll start to see other measures of value out there. A lot of surveys. 'Cause you need humans with empathy to fill out those surveys and let you know how they're feeling.

What do you think? What's going through your head?

Erik Martinez: You know, I just want to comment on what you were just saying 'cause I was listening to, the Everyday AI podcast, episode with the CEO of Aria. Aria projects itself as the middleman between agents and the LLMs. Allowing companies to have a safe place for beginners and experts to use these tools a compliance safe way.

But, what was interesting about one of the things he said. [00:13:00] He kind of contradicts one of the things you just said, which is he thinks that the notion of AI sprawl, that in an organization and even within a small agency or a small e-com retailer, there is a lot of AI sprawl. And I'll give you an example. My wife, just took a job in a manufacturing plant. As an admin. And I'm not gonna say who, what, because I can't do that. But, one of the things that's really interesting inside this manufacturing plant, they are Microsoft shop, they have access to office tools and a little bit of copilot, but not everybody has access to copilot because IT hasn't enabled all the things. Because they're a manufacturing plant, what they're manufacturing has some proprietary, they can't take photos, they can't record meetings, they can't do a lot of the things that we just take for granted, like we [00:14:00] do every single day.

And yet, when you ask them Hey, can you use your phone during business? Oh yeah, we're allowed to, you know, do texting and stuff. And so that's the opportunity that creates this little AI sprawl. Because, if she has chat GPT on her phone, she could ask it questions while she's working.

And that's kind of this shadow AI scenario that Aria was talking about. I think CFOs are going to be asking that ROI question. I think everybody else in the organization's gonna have a hard time answering it the way you said. And I think there's gonna be some tension there, right?

Whether it's a CFO or a business owner who's like, how do I get real value out of these agentic systems? I think it's gonna be hard to measure until you understand what your objectives are.

Pat Barry: Yeah. And to that end too, I'm glad you brought up sprawl, and shadow AI. I like to think of it shadow AI. That sounds like kind of cooler, like a spy. It makes me feel like a cool [00:15:00] person when I'm doing my job. It's funny you bring that up, because I'm working with a company right now, big healthcare provider, and that's one of the questions they keep asking me is like, Hey, we haven't approved anything. We're looking at this one system. We're probably gonna have to work in just Microsoft or Microsoft Shop. At the same time they're we're all using chat GPT on our own.

They have no like written AI policy internally yet. And I'm like, you need to nail that down and tell people what can I use, what I can't use and what can I do with it?

What can't I do with it? I start bringing up the concept of the human in the loop and they're like, oh yeah, we should have somebody checking this. Yes, you need somebody to check the work. You still constantly need to check the work. The work might get done faster, but you need that human there to check it.

Erik Martinez: And that's one of the places you can improve quality, right?

Pat Barry: Yeah, exactly.

Erik Martinez: You can spend more time validating that the work you're producing meets your standards. I think that's important. So, back to the question of what will agentic [00:16:00] AI realistically be able to do in 2026? There's an agency owner, in my agency owner's group, who has automated 256 tasks this year.

Pat Barry: Good for her.

Erik Martinez: She works 10 hours a week. She's a mom.

Expecting a second child, and her drive to keep doing what she's doing 'cause she loves it, but yet have the quality time to spend with her current daughter and baby to be is what's driving her to automate some stuff. So, again, if you have clear objectives or a destination that you're trying to achieve. She's absolutely crystal clear what she wants to do.

You can use some of these tools, they don't have to necessarily be a really complicated agent. They can just take one small thing off your plate that takes and bleeds time from your day. And that's a good thing. The capabilities are only limited by our imagination and the tools available to [00:17:00] leverage that imagination.

So, before everybody gets overwhelmed, just think about what's that one painful thing? If I could just solve that problem would make me 10 times happier. Focus on that.

And I think that's the real power of AI agents in 2026.

Pat Barry: Well, it's a small tasks. I have an agent, all it does is build one slide for me. It's a single slide in my sales deck that I like to personalize depending on who I'm talking to. It would normally take about an hour to put together. I'd have to go research the company, figure out these four or five different buckets I need to show them of what ideas I have around data, AI, whatever it is.

I'd set up an agent took about two hours, and now I can just tell the name of the company, give them the website. Teeny bit of context. Builds the slide content. It doesn't produce images or anything. And it really produces in about 30 seconds, usually an edit or two, and it's off and running.

I don't have to spend the hour anymore. It's about five minutes, 10 minutes tops.

Erik Martinez: For the podcast, I had the same thing researching guests. I used to [00:18:00] spend two, three hours researching a guest to come up with the questions and rewriting the questions. Now, I've got, an agent I put the person's name in, I put their LinkedIn, their website, a brief description of what the topic is.

And because I've done it so many times, the agent knows what type of questions I want to ask. Couple minutes later I have the questions. Then I can go and edit it and talk with it if I thought I got it wrong. But that has gone from three hours down to, you can do that in about 30 minutes now.

Pat Barry: Yeah,

Erik Martinez: And the questions are as good as I would've come up on my own. And in fact, sometimes they're better.

Pat Barry: Yeah. That's what gets me sometimes oh, I wouldn't have thought of that. Glad I asked AI. I think it does depend on the tasks though. There are some tasks that can get a little bit more complicated. But, you know, again, I think as the systems get easier for the non-technical person to use, things will just expand.

I remember back in the early two thousands when Google came out, it was just a box on a screen, nobody knew what to do. And then it took two, three years and all of a sudden [00:19:00] became a verb in the English language. Everybody understood it completely. So, we'll definitely get there, but, you just don't know what's gonna happen a lot of the time.

Speaking of which, what do you think some of the biggest risks or blind spots are, heading into 2026 with AI?

Erik Martinez: Honestly, I think there's a couple things, like from a blind spot perspective, it is not the end all solution to everything.

Pat Barry: Correct.

Erik Martinez: If you push your agent or your chatGPT or Gemini or whatever it is, you will find where it has holes.

Pat Barry: Mm-hmm.

Erik Martinez: You've gotta stop asking generic questions and ask more sophisticated, well thought out questions, and then have conversations with it. One of the blind spots is, Hey, I can just ask it a quick question and get a good answer. Without context, you're not gonna get the answer you're really looking for. You really have to not try to rush through and answer the quick question quickly the first time. Take some time, give [00:20:00] it some good context. Teach it what you're really trying to get at, and then have a conversation to refine it. You do that two or three times by the fourth time you do it, it's honed in pretty well on what you're looking for.

We talk about training people, but we also have to take time to train the AI to do things. I would say another blind spot is, 'cause I've heard one of my team members say this, oh, i've got my AI trained so well, it knows exactly what I want. And it sounds like me. And I'm like, okay, but is that always good?

Pat Barry: You just gave yourself a case for replacing you.

Erik Martinez: Well, I think that's one possible issue. But I think the other issue for me is, I know I'm not perfect.

So if I've trained it to be just like me, what inherent weaknesses I have, it's probably gonna have. So I think the other big risk is training it to be too much like you is actually [00:21:00] counterintuitive and that you really should train it to also think like your opposition or a mentor or somebody else, so that you can find the holes in the things that you are doing.

What do you think, Pat?

Pat Barry: That is a good answer, that's pretty comprehensive. I mean, I think in terms of risks and blind spots, it's not too big of a deal right now, but as the AI marketplace, you know, there's a lot out there in terms of small tools. A lot of the bigger guys are kind of giving stuff away for free to get clients. For me it's cost, and I think at some point next year one of the biggest jobs somebody's gonna have is not only like, yes, I got to, write my prompts for my agents, but understanding how many tokens your prompt that you send the agent is gonna cost or generate. And then, what's the output? How many tokens is that output coming back? Depends on the tasks. If you're creating video and images, it just costs more money. But [00:22:00] I can easily see a lot of people, they just don't think about it. You got your $20 a month, chat GPT. I know you and I, bought a Gemini Enterprise license last night. I've been messing around with that. It all seems really cheap and inexpensive to me right now, and I think that's going turn, I think you're gonna have to have people, figuring out the token costs, optimizing your prompts to get exactly the results you want, but at the right cost.

And I think the reason I bring this up is, I pay attention to who's funding who and where all this is going. It just looks like these big circular investments like NVIDIA keeps giving chat GPT Money and chat. GPT looks like it's giving the money right back to NVIDIA to build chips and stuff like that.

And that's just one example, but I do think it's either one gonna slow down because we can't keep up with data center demand. And the cost will grow that way. But again, I think the one thing, outside of your answer, is cost.

I can see people blindly using these things, every company, as they begin to adopt AI, either needs to have a system in place. They understand that so they can monitor their costs.

It's almost like keeping track of your [00:23:00] family budget. How much money are we spending every month and I think that'll be someone's job. Yes, you can set an agent up to calculate those things and build it for you, but they're gonna have to give the results to a human. The human's gonna have to go like, all right, we gotta tell Larry in operations, he's gotta shut off Gemini for two more days because it's right at the top of his limit. I think that'll become a big function within a company. Now, depending on the size of the company you'll have somebody on that backend, whether that's somebody in data analytics or operations or warehouse, you know, in finance that's keeping track of that.

I do think that's gonna be a big deal coming up. And I think it'll be a whole industry around, jobs of people. All they do is test prompts. It is hard.

Erik Martinez: It's hard. Because you don't get the same answer every time.

Pat Barry: No, you don't.

Math is one thing. Like yes, if you two plus two is four, you should always get that same answer. But when you're prompting an agent, you give the same prompt to one agent the next day, you might give it to the exact same agent with same model, everything, and it'll just come up with a different answer. That's the human aspect [00:24:00] of AI, where it's like a human. As people get sharper with it and start to understand more, I think prompt optimization to keep costs down, around using AI, that'll be a big thing.

Erik Martinez: I think you're absolutely right. I was, in nano banana a week or so ago. 'cause I've been doing some experimentation for some clients, some light image generation, just showing them some possibilities just to show them ideas and it's one of the most powerful things.

Well, Google rolled out. Gemini three, and then they rolled out an upgrade to nano banana. And I noticed all of a sudden that they're asking me for my credit card. And so what Pat's talking about is actually very, very real. Now, the token costs aren't that expensive and generating 10 images is not that expensive. But he's right. This is going to happen more. Especially for, an organization like OpenAI who is burning way more money than they're bringing in, they're looking for ways to monetize everything they [00:25:00] do. There's, a lot of speculation that we will see this month in December. That we will see open AI launch their ad platform inside of chat GPT. Guys, we are there now. That is going to be a thing, you're all trying to solve how I get into the AI search results. Well now you're gonna be able to buy your position.

How is that gonna work? And by the way, chat GPT hasn't been doing this for 30 years, like Google.

Pat Barry: Right.

Erik Martinez: They don't know that ecosystem. Is that a good thing I don't know yet. None of us know how this is gonna play out. Perplexity is already doing it.

I don't know what kind of adoption they've had on their platform, but chat GPT has a massive audience. So I think there's gonna be a rush to figure out how to use these tools and get your sponsored ads up in the results. Will that drive us all crazy or will it go to a streaming model, which is like, you can do this tier, but it's gonna be ad [00:26:00] supported, or you can do a premium tier that's not ad supported. Who knows how this is gonna work?

Pat Barry: I know, they'll figure out a way to generate revenue. I'm glad you brought that up on OpenAI a lot of people don't know that. It's not a profitable business, at all. It was a nonprofit up until, two, three months ago. And they're heavy hitter in the field. But I also know google's had several releases a lot more people are downloading Gemini now than chat GPT. So chat GPT has to do something to keep up with the industry. Serving ads, that's marketing, that's a whole shift. How does it work?

I remember when I got into digital in the early two thousands. So the search auction was like this big. I'm making a very small, space between my index finger and my thumb right now. But watching that thing grow and the capabilities you could do just paid search over time, I mean, oh my God. And then the amount of data they have to improve their system. So, chat GPT's not going anywhere anytime soon.

Erik Martinez: It's almost too big to fail.

Pat Barry: Yeah

Erik Martinez: I hate to bring up that phrase, but, if there's gonna be a sacrificial lamb in the, [00:27:00] aI space, I don't think it's gonna be open AI, it's probably gonna be anthropic or meta.

Pat Barry: Yeah.

Ad buyers, people in marketing in that world you're constantly coming up with new ways to deliver, to any potential customer. If those people haven't been looking at AI or understanding just the logic, how it serves things, they could be in trouble. I'd add that to our blind spot list for 2026.

Erik Martinez: I think that's definitely a good one. So Pat, let's pivot back to, training for a moment.

Pat Barry: Yeah.

Erik Martinez: I think there's a misconception in this space. There's a lot of data out there that says corporations are investing in AI and they're providing tools to their people, whether it's chat, GPT, or Copilot or the Google suite of tools, but there's still today a very limited amount of training. It's like, go figure this out. I think the challenge that I see with that is. That there's really kind of three different kinds of users.

There's a power [00:28:00] user like you, pat does this, by the way. I've seen him do where he just goes in, digs into something and figures it out. But he also has that mind from the technology and working through the nonsense connections that don't make sense behind the scenes that make it more challenging than just, saying, Hey, connect this to this. Right. There's a lot more steps involved.

So there's a power user like Pat, then there's somebody a little more like me. I put myself in the middle i'm not as technical as Pat. I understand the tech speak and concepts. I just have a hard time weeding my way through all those painful steps that Pat seems to knock down.

Then there's third group that's like, I don't understand it. I don't know where to start. I'm afraid this is gonna take my job. And by the way, I'm not discounting any of those things. So just putting the tools in front of people isn't good enough. When we did that with my team, we gave them assignments. We were like, just play with it a little bit [00:29:00] and tell me what you learned a month later. Crickets.

Pat Barry: Yeah, because you need to kind of show people how to fish.

Erik Martinez: Yeah. But even when you're giving them knowledge, it's gotta be, relevant to their job the whole adage in marketing, right? What's in it for me? What do I get out of this that I should invest my time? Because I'm also really busy and investing in AI and learning it is an investment of additional time, you don't get the efficiencies from it until you learn how to use it.

Pat Barry: Exactly.

Erik Martinez: What's your perspective on that?

Pat Barry: I think one, i'm just a little bit different in the sense, i've been in the tech space in the data world since early two thousands, so for me, like learning something new, that's just part and parcel with the job. I've managed whole teams where I'm like, you guys are gonna have to do stuff at night and on the weekends to up your career skills and whatnot. So I think for me, that's just normal. But in terms of training, it's a lot of what you said. It's just sometimes showing people just click this thing open.

Using our example, you and I did our [00:30:00] training, was that two, three weeks ago now, where we showed people a custom GPT and how to build one, and they were like, oh, did you even know this existed? And then they're like, oh. So it's just typing words into boxes. There's literally no coding here, guys. We did do a teeny bit towards the end, but even that was nothing. We asked the AI to do it for us. So, it's getting over that fear. I think it's also as people get more into custom models and building their own agents, it's explaining to them listen, this is great. You want to connect to your Google Drive. Look at how your files are named. Look at the structure of the internal files. Sometimes the first step is going in and cleaning up your data.

And there's agents that can help you with that. But on a case by case basis for every company. It's not something that you can just pull out of a can and say, oh, it's this format. So I think that's part of it too. Part of training is getting people to understand this thing's great, but if your stuff underneath it isn't formatted properly, it's not gonna be so great. And I think that's another piece of the training is getting people to understand you know, one, find some tools, find what you like. 'cause there's so [00:31:00] many out there right now that you can get caught up in just trying all the tools.

It's happened to me all the time. And I think too, just communicating to people that. You've gotta look at this as we're at, ground zero and we've made a ton of progress this year, just in 2025. But these tools will get easier and easier to use. They'll become more common and just normal within the workspace. I liken it back to when Google came to be, search engines kind of came commercially available around late nineties, early two thousands. Google's homepage was just a box.

But I remember talking to people like my dad, you know, and he's engineer, like pretty technical guy. He is like, I don't, what do I do with this? It's just a box. I'm like, write a question, see what happens. And so I think a lot of people just need to adopt it. It just needs to become more mainstream. I mean, honestly, I think by probably 2027, 2028 people will be more advanced. There'll be agents everywhere. You'll have people just building their own agents. If it's within company policy, in the very near future because you don't have to code. It definitely [00:32:00] helps if you have a technical background, but, the better command you have of the English language and the better you are in, being able to tell these agents context, about what you're looking for, that's really where the training should come in, you know, and telling them about context, data, cleanliness, like those types of things too. So a little bit different than, just explaining stuff. You gotta give 'em the fishing pole and then show 'em how to use it. I mean, that's the biggest thing.

Erik Martinez: I think the other thing I would add to that, 'cause I agree with all of those things like garbage in, garbage out, that's been a principle in IT forever. So data cleanliness, data organization, all those things. I think on the softer skill side though, if you're really starting to think how to use these things, there's a lot more data that's coming out saying, Hey, you know what? People need to learn a little more about the humanities. Learn linguistics, learn another language, learn about history.

I think you're gonna be [00:33:00] speaking 90%, 95% of the things you do. Your input is not gonna be the keyboard. It's going to be your voice and your ability to think and brainstorm. Ideate and problem solve, I think are fundamental skills. So when we talk training, it's not just, Hey, how to use the prompt framework 'cause that is important.

Pat Barry: Oh, it is. Yeah.

Erik Martinez: How do I take that prompt framework and give it the proper context in the right sequence?

And there's gonna be a little bit of experimentation there, but if you're already trained on how to think that way, this stuff becomes a lot easier. And I think the backend technical stuff will get easier over time as the systems get smarter and more interconnected. But, from a training standpoint, I wouldn't limit yourself to just the prompt training also really start thinking about what skills we need in a world where. We're not gonna be inputting [00:34:00] keyboards. It's only gonna be limited to your imagination and your ability to actually communicate your ideas effectively through voice.

I really believe that that's gonna be one of the big things. Let's pivot to our last question for today. In your opinion, Pat, what do businesses need to do to make real progress with AI in 2026? What's the very first thing that you would advise?

Pat Barry: Don't just buy AI for your company. One, I would try it out. Whatever workspace you're using, whether it's Google or Microsoft, just experiment with it there first, see if it can do what you need it to do. Because I think there's a lot of clients that come to us, like when we're doing stuff in partnership, it's always like, Hey, I bought, mid Journey and I've got chat GPT and we're looking at perplexity too. And my first question is what's your office suite? The bulk of our clients are on Google Workspace, which is great. And I'm usually like, okay, you don't need midjourney. You can get it if you want. If you're a creative agency, maybe it's got some other features but [00:35:00] if you're average Joe making, regular marketing materials like nano bananas, fantastic.

I would say stay within your office suite. Google's much more advanced right now than Microsoft, but Microsoft is a beast. They always will be. They've got a lot of money. They will come back and so I would caution people about, willy-nilly buying tools. Chat GPT is pretty common. But even too, like. As you and I have seen, google's catching up pretty quick, a lot of Google workspace companies that come to me, I'm like, I wouldn't buy anything. You can do a lot within ecosystem.

I think it's one of those things like don't rush to judgment, but you also need to, if you haven't done anything yet, you need to start now, because your competitors are doing it. If they figure out efficiencies for some odd reason, that can undercut your costs, doesn't matter what business you're in. People will pay less money for a service if they deem it will be the same quality that they would get from you.

Erik Martinez: Or better.

Pat Barry: Especially the companies that have started to figure it out, they probably have better relationships with their clients because they're people, that's where [00:36:00] humans should be spending time communicating with humans building those relationships, understanding the needs of the client, being able to communicate that back to your agent. So it can either, build a new logistics plan, build a new creative opportunity. If you haven't done anything, do something like take a poll, take a survey of your company, just send a survey around what do you like about AI? What tools are you using in your personal life?

Figure out, what people in your company have already done. See if you can surface some of that to bring out to the masses. Maybe, joe in accounting has some process he's using with chat GPT that shaves two hours off of his week. That's great maybe you can take that process and evangelize it to other business units or talk to other people. Get people talking about it, do something with it, and bring more efficiencies to the table. I think that's the biggest thing. Try to get over the fear. That'd be my answer.

What about you? What are you thinking about? In terms of businesses?

Erik Martinez: I agree with everything you said.

There are a lot of tools out there outside of the core AI models, [00:37:00] outside of Chat, GPT and Gemini that are wrapping AI into their product.

Pat Barry: Yeah.

Erik Martinez: It is not an equal experience folks. Some of it is really good. I'll give you a good example. I've been playing around with, agent inside of, zapier, which is an automation tool. It's really pretty dang good, but it's not perfect. And I had a scenario where I was trying to connect something and I kept saying, yeah, yeah, here I can do it. I can do it. And it came up with strategy and it didn't work. And then finally we figured out the piece of data we were looking for wasn't even available in the API. And that should have been one of the first things to check like, oh, I can't get that piece of data. It's a really, good tool. There are a lot of not very good tools out there that don't add any practical value.

So my advice is the same as Pat's.

Start with Google or start with. Open AI or start with copilot. Copilot isn't [00:38:00] as well integrated into the Microsoft products as you would like. It will help you solve problems within there, but it is not like the experience that Google is now starting to deliver. Google, I think in the last month or so has leapt ahead of the rest of the field with what they have released.

I have a love hate relationship with Google on the advertising side, but in this particular case, I think they are doing really good work. They're integrating these tools into their products smartly. They actually work most of the time and it's just a few dollars more a month on your license if you're already using Google products. If you're not, there's some free versions you can use. You'll have some limited capabilities, but it's better than nothing. So I definitely think that, starting there. I really believe that starting some basic prompt training and AI literacy training within your organization, you know, we do that type of stuff. You don't have to come [00:39:00] to us, there's other people who are doing it, but you really should invest in some basic training.

I think that's a good thing. Pat mentioned the survey, we'll put out in the show notes. I've got a free survey that asks some of these questions. Happy to share the results with you. If you fill that out, we can put that in the show notes and make that available. It's a 10 minute questionnaire that just says, Hey, are you using AI? What tools are you using? Is your organization providing training? Do you have a roadmap? I know 90% of the answers will be no, if you take 10 minutes, you can learn a lot about, what you guys are doing. We also have a simple assessment that we'll get up online and put in the show notes. Start documenting your people's daily work. What is it that they're doing as part of their function? This is a critical starting point, and then there's a process to break it down by understanding that you can start seeing how the pieces get put together in your organization. Then you can start figuring out, you know, we could solve this one problem and it [00:40:00] frees up all these other things

We talked about training, picking one of the frontier model tools. They're great and not super expensive. Don't go buy a bunch of 10,000 other tools 'cause they're probably overkill for what you need. And then really start documenting, what's going on in your organization so you can figure out where the pain points are and start solving those problems. What I would do.

Pat Barry: I have nothing to add. That was pretty comprehensive.

Erik Martinez: We appreciate the time in listening. This was a little bit of a different one, but wanted to give you some food for thought for 2026. And, provide some learnings that you might be able to take back to your business and think about. Feel free to reach out to Pat or myself, to ask questions.

Pat, what's the best way to get ahold of you if somebody wants to reach out?

Pat Barry: Just via email, pat.barry@aiconsultingpartnersllc.com.

Erik Martinez: It is B-A-R-R-Y,

Pat Barry: That's right, the right way

Erik Martinez: [00:41:00] pat.barry@aiconsultingpartnersllc.com.

And over here you can reach me at Erik@digitalvelocityagency.com.

Happy to, talk about anything, AI and brainstorm. Happy to talk to people. I hope this gives you a good start for your 2026. Some ideas and ways that you can get AI more embedded in your organization.

Pat, any last thoughts that you'd like to share?

Pat Barry: Just keep trying stuff. Things will get more streamlined in the future. Promise. AI is here to stay, adopt now and get over your fears as best you can.

Erik Martinez: There's no one right answer on this particular topic. So, thank you very much for your time, folks for listening.

Narrator: