Episode Transcript
[00:00:30] Speaker A: Welcome back to AI Today. I'm your host, Dr. Alan Bideau. And this week we've got an exciting topic for you. You've probably heard it just about everywhere. You can hear about AI talking about Magentic and autonomous agents and whatever other buzzwords we decide that we're gonna throw out there as engineers, like, we're really good at. That's. I know it's frustrating a lot of people. I've gotten a lot of emails this week on it simply because it's just been all over the place. It seems like everybody's got something that is now agentic or now an autonomous, you know, agent or whatever. And, you know, I'm going to provide some clarity this week for folks. We're going to go over some keywords, some terminology that I think is going to be really important for you all to, to better understand, because then it's one of those things where you can start to pull it into your business. Like we, like we always talk about if it fits and not every single type of platform is going to fit what you're trying to do. So we're going to talk a little bit about that in the next segment. But, you know, really, let's just, you know, let's just take, you know, just a quick step back because, as you know, we went into 2025, you know, we knew that there were going to be some economic challenges. There were, you know, a lot of companies that were starting to pull back on their investment dollars and, you know, they started looking at different types of, you know, AI systems to see which one is really going to, you know, really impact what my business can do. Still hit some of my growth targets that I have, still save some money in the back office, things like that. And what, what we found was, is that, you know, a lot of companies have just actually just rebranded some of the things that they've, they've been doing. You know, these, these, these, you know, autonomous agents and these, you know, agentic platforms that are out there. These are not new. So I don't want folks to get scared that there's some new type of AI that has been released. Okay? It's, it's, it's just something that, you know, they're trying to get more excitement around it. They're trying to do some new things. There are some, some new capabilities that, you know, when you integrate them appropriately with large language models and some other AI, then you can really start to impact what your, your work looks like, what's your, you know, your workflow, the things that you're trying to accomplish. So that's really, you know, important. And so I want you to stick around and pay attention because I think it's going to be able to help you. You know, as 2025 rolled in, right, we knew transformative period for AI. We were really in one of those areas where large language models had started to kind of back off on releases and, well, other than the 12 days of Christmas that of course, that OpenAI gave us. But others had started to back off, you know, because the cost of building some of these models has just become really prohibitive in, you know, their ability to just keep scaling and scaling and scaling. And I know that folks have seen some of the things that, you know, Elon Musk has said about, you know, it's really the, you know, it's peaked out, it can't do anymore. We'll see. But, you know, those are, those are just, you know, some of the sentiments that are, that are going on right now. And so folks started to look at things and they started to say, you know, I don't know if I want to invest in really trying to do new AI research until we get a better understanding of what sort of impact it's going to have on people. How can we commercialize it?
What's the delivery medium going to look like? Are we going to continue as a chat type platform or what's that interface going to look like? There was a study by Forbes, and I believe it was Stanford University when they saw that, you know, inference, computing large, you know, these large language models were really reshaping how we were interacting with technology and what it was doing in the workplace. And, you know, that was important because it was really an eye opener for a lot of companies to start to say, you know, our strategy, our strategy is a little shaky. We're trying to figure out what we're going to do, but now we're going to, you know, we have a better understanding where the technology is going. And so now we are going to dive in and do this, whatever this is for, for some companies. And so they started to look at that and they started to say, okay, we don't have the resources necessarily to do, you know, what we really want to do. We have bits and pieces of that, what technology is going to fit in best with us. And you know, really taking these autonomous agents and layering them in with your, your staff was an easier approach for, for some folks to, to be able to understand and be able to apply it to a real problem that they were potentially Having, and that started to, to impact how we were looking at, you know, different types of AI. And I think that's, that's really the key. You know, as you look at things, as you look at this technology and you're saying to yourself, geez, there's something new released all the time, there's a new model, there's a new something. What am I going to do? I just invested in another one. You know, how am I going to keep up? Take a step back, look at your, what would you always say? Look at your strategic plan, make sure it aligns and then make the decisions that you have to, that you have to make. Because, you know, we have, you know, April has been, you know, really, it's, it's, it's, it's very interesting. It's really started to become an inflection point where, you know, for, for AI development, for that focus on the technology landscape and that it seems that the, the tides and the winds are blowing in a direction where autonomous agents are going to get really a lot more funding, a lot more investment, you know, a lot more applications and, you know, that usability piece. And so it started to change the landscape of AI and you know, that's one of the bigger drivers that we're going to have to, that we're going to have to pay attention to. You've got to be, you've got to be proactive. You've got to look at some of the tools that are out there and you've got to say to yourself, does it fit or does it not fit?
Then you've got to figure out what kind you're going to get. And by what kind, I mean, there's generally, you know, five or six different types of these platforms.
You know, the first one is really just your normal, everyday boom. It's just, you know, it's all autonomous, no decisions, going from point A to point B, moving, you know, numbers from one spreadsheet over to another spreadsheet, for example, very easy, very automated type, you know, bots. No, no intelligence there whatsoever. And that, you know, starts to lend itself towards, you know, how am I going to transition the work that I have to an agent or the work that I have to an intelligent agent or, you know, a certain type of platform.
And it starts to complicate things really fast. You know, we talk about the lack of resources, the lack of human capital that we have already to deploy these things, and it really starts to scare some folks to think, geez, now I'm going to have to not only Manage my people. I've got to figure out how I'm going to manage these agents and these, these, you know, platforms that are out there. I don't know how to do this. I don't know where I, where even to start. And so it's frustrating a lot of people. And I think, you know, this is a great topic for, for us to start to look at because, you know, we've touched on it a little bit in the past. I'm not even going to talk about my, you know, cognitive Persona platform that we have, which is two or three steps above even the agent, you know, agentic platforms. But, you know, it's starting to think about ways that you can break problems down, really complex problems into smaller chunks and have these agents go focus on certain pieces of that problem so that they can solve it and then you can integrate that solution into the greater problem that you're trying to, to tackle.
That's really at a high level, what these platforms are meant to do. And of course it's at different levels and it can be at different skill sets and those sorts of things. But you've got to start to think about not only your strategic path that you have as a company and as a business, but also what resources you're going to align to that strategic path. You're going to solve one problem with one set of agents, potentially another problem with a different set of agents, potentially. How are you going to take the information and communicate that to humans? And you know, do you want the platforms to communicate back and forth to each other? You know, a lot of, a lot of different decisions that have to be made in order for it to work the right way to, to really do what you want it to do. And we haven't even talked about the right tool yet, so you know that that poses a little bit of a challenge for us. And so I want you to stick around though, because in the next segment we're going to talk about, you know, some terminology that is going to be very important for you to learn, understand, do research around it, get, get your, your favorite large language model out and ask it questions so you can understand it better. But stick around. We'll be right back after a couple of short messages from our sponsors.
Foreign welcome back to AI Today. I'm your host, Dr. Alan Badot. And this week we're talking about agentic systems. Sounds. Oh, it sounds so important, doesn't it? Sounds so. Oh, yeah, yeah, like it can help drive more business. Right? Doesn't it? When you say it doesn't it make you feel like, like you can tackle the world now, right, because you've got your agentic platform right behind you. Well, no, it doesn't work that way. Sorry. Don't want to burst anybody's bubble. But too bad.
You've seen it all over the place, right? Copilot. Everybody's probably heard of Copilot.
You know, of course, you know, the, the things that are going on in open AI, you know, every single large language model seems to have some sort of, you know, agentic platform that is available. You know, they're, they're really all over the place. And so, you know, I'm gonna, I'm gonna be nice.
Some of them are shrink wrapped from five years ago, quite honestly, and some of them are offering some new things to, to pay attention to. It's not my job to go out and tell you exactly which ones those are. However, you can do some research and you can see some functionality that has been available for a while in some of these platforms and that really should give you an idea. Okay, so that's all I'm going to say about that for a few minutes. But you know, I want folks to not be afraid of some of the things that they're hearing. Not be afraid that it's going to take over your business. Not be afraid that, oh, if your, your competitor gets a hold of it, that now you're, now you're really in trouble because they're gonna, they're gonna, you know, you know, start to start to smoke you on sales and stuff. Right? The, the reality is from, you know, three months ago till today, my message has always been the same. If you're not using AI somewhere, then you're really missing out on the possibilities, the advantages that you can get. Now just because the terminology has started to shift a little bit and the marketing has started to shift a little bit. That doesn't mean necessarily though that the impacts to you are that big.
You've just got to pay attention because these are the types of things that we were talking about and have been talking about for over a year. You know, I, I did agent, you know, autonomous agents and stuff. Five, six years ago, they were different, but still the same principle, still getting, you know, autonomous, you know, different pieces, different chunks of the workflow, all working together, communicating and trying to solve a bigger problem.
And so that's where terminology, familiarity with the terminology is going to become so important because, you know, you'll see that as you learn more about these things, you understand them better, then you'll Be able to, you know, really, you know, sift through all the, the garbage that's out there and get to something that is meaningful to you and going to be helpful. Okay, so that's what we're going to do. Now, there's so much data. We know that you have so much data as, you know, business, you know, enterprises and entrepreneurs and stuff, you have a ton of data based on your customers that you have. Every data is going to be different. In some cases it could be text, other cases it could be images, maybe it's sensors from phones or different information that you're getting from, you know, these devices that are, you know, in the field of, you know, the Iot, the cloud and the, you know, the sensors on the cloud and those kind of things.
But, you know, at the end of the day, what are we trying to do? We're trying to take that data, trying to better understand it, we're trying to apply that understanding and then we're trying to free up time, make more money, whatever our goal is.
And the decision on what type of platform that you're going to use or what type of, you know, you know, agent that you may deploy that could be part of a large language model or something different becomes very important.
Because once you make those decisions and you start to integrate those technologies into your business, there can be some pretty expensive ramifications to that. These technology, these tools are not cheap. Technology is not cheap.
So you've got to just make sure that it applies to what you're trying to do. It's flexible enough so that it can grow with your needs and, you know, hopefully it scales with you because you guys are going to start to make more money and generate more revenue and do all those things that we're trying to do.
So just keep that in mind. Now, I'm going to talk about some things, you know, chain of thought. We'll start there. Let's just chuck it out there. Chain of thought.
What does chain of thought mean? Well, it means exactly what it sounds like. You've got these agents and some of you, you know, if you have, you know, the, the new version of Chat, GPT or some of the other ones, you can click a button and you can see what it's quote unquote thinking. Right? Well, that's what chain of thought is.
It's, it's looking at, you know, allowing these agents to explain, you know, almost step by step. But it's not really step by step because we can't track every single step that they're going through, but it's almost like it's going step by step, you know, of the pro, you know, through the process that it's trying to use to solve a problem.
Or, you know, in some cases, it might get stuck with something and it'll say, oh, I'm gonna try this other avenue because the user wants this, and this, this one doesn't work. Right? So it's. It's trying to get these agents to think a little bit more, really, with the end goal in mind, as opposed to us saying, you know, go do this.
You know, we're telling it, this is what we want to accomplish. And we're trying to let the agents, you know, give them some more leeway so that they can. They can do what they're really good at. And that's when, you know, you couple them with these large language models. It makes a big difference.
Now, we've talked about this, so if anybody doesn't understand what this is, either you're a new viewer or you were sleeping through the show, and I hope you weren't sleeping. Human in the loop. Very important.
Why? Because we never want our AI to be able to operate by itself, unwatched, and have it bite us in the butt. That's what it royals down boils down to, right? We want to make sure that we put a person wherever our comfort level is, wherever there is a tough decision to be made, especially when it comes to, you know, certain types of markets where there are some pretty big ramifications that, that come with messing stuff up. Like the lending market, the jobs market, where, you know, maybe a bot has, has messed up and is filtering certain.
[00:19:17] Speaker B: Types of resumes out, right?
[00:19:19] Speaker A: So that's why we always want a human in the loop. We want to make sure somebody's watching.
[00:19:25] Speaker B: The cookie jar while the AI is.
[00:19:26] Speaker A: Just, you know, going through and running around. Okay?
Very, very important. You know, I stress this all the time.
Gotta keep a human in the loop.
Now, tool use and integration, right? I'm not gonna, I'm not gonna go too much into that because we talk about that every single week. But it's picking the right tool for the right problem.
Now, what does that mean? Well, it means exactly what it's. You know, I said, if you don't have a problem, don't buy a tool.
If you have a problem that you're trying to solve, pick a smaller part of that problem to test. Let's see if it works. Let's see if we do, you know, one solution over the other. Which one gives me a Better result.
Now, not always is, you know, your pilot that you have going to scale. However, it will give you a little bit of confidence at least to say, you know what, what I did was pretty sound. The results that I got back were pretty good.
So that gives you the confidence to take that next step, right? Because that's all we're trying to do now.
Software developers, you, you guys know this. I don't have to, I don't have to explain this one to you, but APIs, what in the world is an API?
Well, simple. I'm not even going to tell you what it stands for. You can go look that up. An API is your ability to plug your AI into another tool.
That other tool could be Windows Office 365, for example. It could be PowerPoint, it could be your HR system, it could be your CRM.
If you want to look at more advanced things, it could be some other technology or research that's taking place that allows you to take your AI, plug it in, communicate with that other system so that they exchange data, and then you take that data and you apply it to your, your problem.
[00:21:34] Speaker B: That's it.
[00:21:36] Speaker A: There's a lot of different platforms out there and, and actually they're really easy to write on your own now because you can actually ask a lot of different AI tools, these large language models, how do I take my AI tool and connect it to Gmail? It'll walk you through the steps. You have just built an API. If you did that, it may not tell you that, but you really have very simple.
[00:22:04] Speaker B: This allows two applications to talk to each other. That's not frightening.
[00:22:11] Speaker A: There are, I will give you a couple more. Right. But you know, cognitive Personas, what my.
[00:22:19] Speaker B: Company does and what we're building and.
[00:22:22] Speaker A: Have built, we're putting personalities into it.
So then it can flex, then it can understand how better to use the tool. You hear me talk about this all the time.
[00:22:35] Speaker B: It's impossible to take the bias out of any AI model that you have today, these large language models.
[00:22:43] Speaker A: It's impossible.
[00:22:45] Speaker B: You have to assume that there's bias in there.
[00:22:47] Speaker A: So what you can do is, is you just can try to shape it.
[00:22:51] Speaker B: Maneuver it better, understand it, so that.
[00:22:54] Speaker A: At least when you're applying your technologies to, you know, your problems that you have, you're going in with an open.
[00:23:02] Speaker B: Mind and you know that there's bias in there and you have to account for it.
[00:23:06] Speaker A: Transparency, it's all we ask for. Right? You can't get it anywhere else. So the, our, our Personas allow you to understand that better and apply, you know, to harder problems, that's really going to be key, is as you move forward, what type do I need?
[00:23:25] Speaker B: Do I need one that's just autonomous.
[00:23:27] Speaker A: And running, or do I need one that's got some personality to it, that's.
[00:23:30] Speaker B: Going to, you know, give me feedback.
[00:23:34] Speaker A: Negative feedback, maybe on performance or those, you know, other things that it can, can help impact? And so that's it.
There's no other real terminology that you have to really be concerned about.
[00:23:50] Speaker B: Think about the basics.
[00:23:52] Speaker A: And if you do that, you're not.
[00:23:54] Speaker B: Going to be blown away by when somebody says, oh, I've got my GenTech platform that's running on the fifth level and doing all these other things. Forget that just means that he's got some bots, they're doing some train of thought, they're using some large language models and they're able to execute some stuff and they're applying it to their problem. Just that simple.
[00:24:15] Speaker A: All right, so stick around. We're going to do a demo in the, the next segment.
We'll see if it works. We always, we always walk the edge sometimes with, with this show when we do these kind of things live. So stick with us though. We'll be right back after a few messages from our.
Welcome back to AI Today. I'm your host, Dr. Alan Badot, and this week we're talking about agentic platforms. We're having a fun. I love saying that word, authentic agentic.
Oh, yeah. Give me shivers every time I say it. So I'm trying to get folks to just not worry about some new terminology.
We're gonna, we're gonna do a demo this time and I think it's something that, you know, we, we, we always try to do them and we're gonna, we're gonna see how this one goes this week because, you know, I don't, I don't try to use the power of TV and go from one oven to the other and magically done. You're gonna see, you're gonna see exactly what I go through as we, as we try to build some of these things live. Because it's you to see, it's okay if you mess up, start over. It's not that big of a deal. Okay, so what we're going to do is we're going to take our friend Claude. I love Claude. And Claude is, you know, probably my 1A or 1B favorite platform.
Very good, very good. You know, examples that they have. The interface is nice and clean, you know, does a heck of A job programming, does some really good job, you know, with some other things, sometimes gets too excited and we'll just start writing stuff that I didn't ask it to do. So you've gotta, you've gotta take the good with the bad sometimes, you know. Of course. You know, I love OpenAI's love perplexity for things. I just, I use them for different types of things. And for, for Claude, I just want to show you some basic ways that you can interface your, you know, your AI with, you know, some different parts of your, you know, business or your, your personal life. And in essence, we're building a low level agentic platform.
So just bear with us as we, as we go through this. Right. And so, you know, let me just walk you through a couple of different things. So you've got some different projects you can see here. You can start new chats, you can open your own projects. I've got a whole bunch of them in there. So we're gonna, we're gonna keep that in a couple of customer ones in there. So we're gonna keep that download for now. But you know, if you look down here, down here in the bottom right underneath you can see this new feature, Connect apps. Now we were talking about APIs.
You know, Claude, in a lot of cases can do a lot of things for you and it's pretty, it's pretty slick. So if you want it to, let's do this. I don't think I have a Google Calendar. I can't do something that's going to.
[00:27:53] Speaker B: Get me in trouble. Well, I can, but we'll see what happens.
[00:27:56] Speaker A: So if you wanted to connect to your Google, your Google, you know, your calendar.
Yeah, we're gonna, we're gonna do that. Hopefully it lets me do it without. Oh, I gotta sign in and do a whole bunch of other things.
[00:28:08] Speaker B: Well, stick with me.
[00:28:09] Speaker A: We're gonna, we're gonna just keep hitting, continue until we get there.
All right, well, that wasn't as bad as I thought it was gonna be. Usually I got to put in a password or do 30 other things. So that's good, you know. So we have just done our first.
[00:28:25] Speaker B: API boy checking make sure that nobody's watching this and gonna prank us.
[00:28:34] Speaker A: But it's not that hard, especially today. It just did it automatically for us. I pointed it at my calendar and then boom, click the button and I just launched, you know, an API that connects those two pieces. It's pretty easy. Now the great thing is, is Claude has just, you know, is given, trying to give Us some ideas, you know, what's my calendar? You know, what are some of the meetings that I have? You know, are they optimized? What's the, what's that going to look like? Well, you know I want to, I want to do just. We're going to do a couple other things with it. We're going to say you can still see it's, it's still connected down here. Right? You know, so I want to, I got an idea and we're going to look at coding with our calendar and we're going to say can you create a react code?
Well, let's say it then tell it the name react 19. Very easy things for folks to be able to look at. React 19 code to manage my calendar via my phone.
Something simple. We want to be able to push meetings that we have right to our Google Calendar. So I can push meetings from my phone to my Google Calendar. See what it says, we're releasing the agents.
No, it's telling you, it's walking you through some of the steps now just sit back and watch it's going through. It's creating an app for you.
It's doing this automatically.
It's really, this is where the agentic platform stuff starts to come into play.
[00:30:34] Speaker B: Claude is acting as an agent, an.
[00:30:37] Speaker A: Agent that I asked it to do some software development for. It's writing the software and now when it's done it's going to show us what some of the results are.
Before it's non trivial to develop something like that. You can see the code and the styling that it's doing. Well, it just did its styling now and it's thinking about it and it's looking at it and now it's going to tie into the API service that Google has and oh amazing.
You haven't had to do anything around that.
[00:31:17] Speaker B: You can see that it's trying to.
[00:31:19] Speaker A: Talk to Google, trying to connect, trying to make sure that it's able to.
[00:31:24] Speaker B: Pull that information in.
[00:31:25] Speaker A: And you can see that it's multi step.
It's a multi step workflow that it's.
[00:31:30] Speaker B: Trying to go through.
[00:31:32] Speaker A: It's thinking.
[00:31:36] Speaker B: I want to solve the first problem which is he asked me to create an app that will generate something.
[00:31:46] Speaker A: Then I've got to connect to an API. I'm going to write code to do that. Now it just gave me a message. It's going to say use the calendar app. You need to set up a few things.
[00:31:58] Speaker B: Create a Google Cloud project and enable.
[00:32:01] Speaker A: The Google Calendar API.
Configure OAuth security oauth equals security consent screen to create your credentials because you don't want somebody else messing with your calendar and install some required dependencies. Now, we're not going to go through.
[00:32:19] Speaker B: All that because we don't have time, right?
[00:32:21] Speaker A: We just, we just don't, however, say, for instance, you're clueless about it. It's gone all these steps you've asked it to do some things it has delivered for you, and now it's told you, here's some steps that you have to go, go through to finish being.
[00:32:38] Speaker B: Able to do it.
[00:32:39] Speaker A: Well, I don't have a clue. Just thinking about it, right? I mean, you may not have a clue what your. How do I set up a Google Cloud project? Enable the Google Calendar API? Well, let's ask how do I do? Number one, I like to go dot, dot, dot. Don't ask me why. It's a habit.
Some folks have noticed that. But we're asking it now. You got to help us keep solving this because we're not, we're not as.
[00:33:12] Speaker B: Familiar with it as we would like.
[00:33:16] Speaker A: To be, so we need some additional help. Well, it just pulled up and it says, well, here's how. You go to Google Cloud Project and enable the API.
There's the console, there's the create project. It walks you through the steps, it's.
[00:33:31] Speaker B: Telling you exactly the APIs and the services and where to go, how to.
[00:33:37] Speaker A: Enable it, and boom, you're ready to.
[00:33:40] Speaker B: Go to the next step.
[00:33:44] Speaker A: Well, that's great. Well, let's assume that we forgot the other pieces too. Oauth, boy, let's see.
How in the world do I configure OAuth ou OAuth, whatever that is, right? It's a security standard for pretty much Google. And, you know, a lot of folks, you know, adhere to, but pretend you don't know that.
Well, you go there, you get your credentials, you sign up here, you've got a, you know, select your, your OAuth client ID.
[00:34:32] Speaker B: You set that up, you set your device up. It tells you how you can, you.
[00:34:37] Speaker A: Can get encryption for it, which is really. This ssha is really just an encryption mechanism that really gives you your digital.
[00:34:46] Speaker B: Fingerprints for, you know, this app being able to recognize you.
[00:34:51] Speaker A: You enter it and boom, you're off and running.
Even tells you how to do it for, you know, your, your iPhone versus your, your Android device. Device. Pretty slick, Pretty easy so far. Right now, you can do this with any other app that you have as long as they have an ability to connect to it.
[00:35:12] Speaker B: Some Don't. Okay, you're going to run into some.
[00:35:15] Speaker A: Applications that they just don't have an.
[00:35:17] Speaker B: API for it that's, that's free. They may have one you got to.
[00:35:21] Speaker A: Pay for, but be careful. They're usually, they're usually costed per transaction. So if you got to pay for one and you don't know how many.
[00:35:30] Speaker B: Transactions you're going to run into during.
[00:35:32] Speaker A: The day or whatever that time period is, can get expensive.
[00:35:36] Speaker B: So pay attention.
[00:35:37] Speaker A: Okay, but then the last piece is install required dependencies in your react Native project.
[00:35:47] Speaker B: Well, maybe you don't know what react native is.
[00:35:50] Speaker A: It's actually just a mobile capability, but.
[00:35:52] Speaker B: Maybe you don't know what that is. So then we're going to ask it.
[00:35:55] Speaker A: What is React Native Native and why.
[00:36:02] Speaker B: In the world do I need that?
[00:36:06] Speaker A: Let's ask. It tells you what the, that it's a framework, right?
[00:36:16] Speaker B: Helps you walk through the process.
[00:36:19] Speaker A: So in 10 minutes we were able to tell it, you want an app that will help you manage your calendar.
[00:36:29] Speaker B: So you can send it from your iPhone device.
It went out and said, you know what, I'm going to write code to accomplish that. I also know that that code has got to talk to Google, which is where my calendar is.
[00:36:48] Speaker A: So it's got to talk to Google. So I'm going to do the API.
[00:36:51] Speaker B: Piece for the human and then I'm.
[00:36:55] Speaker A: Going to tell it the steps that.
[00:36:56] Speaker B: It has to go through in order to accomplish the rest of this. And we were like, whoa, wait a minute, we don't understand some of those things. Let's ask now what we've done is, is we've shown how easy it is to create a very low level agentic.
I think I need some music to play every time I say that.
Agentic platform, very low level, very simple, but still, still agentic.
[00:37:28] Speaker A: We've demonstrated a human in the loop. Oh yeah, that's right. It didn't go do all this on its own.
[00:37:36] Speaker B: It was checking with us, telling us things that we needed to do.
[00:37:40] Speaker A: We were asking questions, we were interfacing it.
[00:37:43] Speaker B: Human in the loop.
[00:37:45] Speaker A: And at the same time, between the.
[00:37:47] Speaker B: Two of us, it was a chain of thought, right? We were trying to just get through things, thinking together.
[00:37:52] Speaker A: We were doing most of the thinking.
[00:37:54] Speaker B: It was doing most of the hard work. So that is all these platforms are.
[00:38:03] Speaker A: So don't be afraid, stick around. We're going to be back after a few messages from our sponsors and we're gonna, we're gonna talk a little bit more about this. And then we're gonna, we're gonna bring it on home, so stay with us. We'll welcome back to AI Today. I'm your host, Dr. Alan Badot. And we have been diving deep into agentic platforms.
So last segment, I showed you the basics, a basic version of how you can use Claude as an agent.
[00:39:09] Speaker B: Right.
[00:39:10] Speaker A: Showed you how.
We gave it some, some directions.
We didn't specify the steps that it had to take. We just gave it some, you know, we told it we wanted to be.
[00:39:21] Speaker B: Able to interface with our calendar and control it via our phone.
[00:39:25] Speaker A: Claude went out and developed the software for us. Then it told us the steps to connect it. We didn't understand the steps, so we.
[00:39:32] Speaker B: Wanted Claude to tell us some more. And Claude walked us through the process.
[00:39:37] Speaker A: That's pretty good.
And we've created a human in the loop, low level agentic platform.
And it's not scary. It's not that hard.
You, you know, again, taking that first step, doing a deep dive, understanding it, that's the important piece. Right? And now we're gonna, we're gonna use Claude for a few more minutes just.
[00:40:04] Speaker B: To kind of give some folks some ideas on, you know, what they should look for and, you know, really how you can use some of these tools, not necessarily for more complex problems, but.
[00:40:17] Speaker A: Really how you can just help yourself understand them a little bit better.
So I'm gonna say this, I'm gonna say I am a.
Oh, it's tax season. Let's use. Yes, I am an accountant and accountant who has over 75 clients that I do their taxes for. Scheduling this time of year. Oh, year is an issue.
Of course. Don't forget to do your taxes. Yeah.
[00:40:59] Speaker B: Today, if you're watching the show and.
[00:41:02] Speaker A: You haven't done them, it's too late. Get an extension. So scheduling this time of year is an issue.
I would like some. Oops. Some help in reminding my clients yearly or throughout the year to keep their receipts and venture prepare and then schedule accordingly.
It's terrible spelling. You guys know I'm terrible speller, but I know how to spell that. But I'm typing in a hurry.
What type of agentic platform is right for me? Okay, very simple.
[00:42:00] Speaker B: We're just gonna ask it to give us some advice.
[00:42:03] Speaker A: What kind of platform do we need? What. What makes the most amount of sense? Now I'm gonna take Claude off concise mode. We're gonna put Claude.
We want Claude to be a little bit more wordy so that we can get some. A little bit better understanding. That's a great thing.
[00:42:19] Speaker B: About Claude, you can play with it and tweak it and you know, change, change some of its settings with that. It's really, it's really a nice tool.
[00:42:27] Speaker A: If you haven't tried Claude and I.
[00:42:28] Speaker B: Haven'T told you, shame on me. But I really, I really do like.
[00:42:32] Speaker A: It'S great to interface with and it.
[00:42:35] Speaker B: Puts up with me. So we're going to ask it to.
[00:42:37] Speaker A: Give us some ideas on what types.
[00:42:39] Speaker B: We should be looking for.
[00:42:41] Speaker A: And so it's talking about client management.
[00:42:46] Speaker B: Systems, tax specific practice management software, you know, CRM software with some automation capabilities. You see, it's just, it's just going off and it's, you know, picked a whole bunch of different things and it's looking at it from different perspectives, right, because each of these are a little bit different. You know, now you've got a tax management software versus a CRM system which is really your client facing, you know, your client type of relationship software versus, you know, a client management system with automation in there.
Three different tools, three different costs, more.
[00:43:34] Speaker A: Likely at each one at a little bit different of a level. But you know, it's looking at some.
[00:43:41] Speaker B: Of the capabilities, right? And it's also telling you, well this is a potential implementation approach.
[00:43:49] Speaker A: Now it's hard to, you know, take.
[00:43:54] Speaker B: A step back and say, okay, if I'm going to do this and I want it year round, what's my implementation approach going to be? That is not trivial.
[00:44:06] Speaker A: So let's ask it, what's.
[00:44:11] Speaker B: What type.
[00:44:13] Speaker A: Of schedule should I think about for deploying a.
[00:44:22] Speaker B: Say you've chosen one, let's choose one. Let's choose, well, it is tax season. Let's choose the tax specific practice management software.
And we won't, we won't go into any of these specifically. We won't pick a specific one because I haven't used them, I don't know.
[00:44:39] Speaker A: If I like them or not.
[00:44:39] Speaker B: So I'm not going to tell you if I like them or one way or the other. So I'm not going to pick one. We're going to just still be general. What type of schedule should I think about for deploying a tax specific practice management software using, you know, an agentic platform? We're going to keep that general also because Claude is doing pretty well.
[00:45:02] Speaker A: So it is now I'm going to.
[00:45:04] Speaker B: Say, okay, let's break it up. Phase one, we've got a couple of, you know, milestones that we want to hit. Here is our, our software gives you four to six weeks to choose what software package that you want to do it's telling you you need three to five candidates. Makes sense, right? You don't always pick the first one, you know, unless they just absolutely blow you away. And then, you know, that's, that's okay because, you know, relationships matter, those kind of things matter. But still, no, no issue with that in the implementation phase.
Now, remember, and this is something that you've got to pay attention to. These agent platforms do not come out of the box. They tell you they do, but they don't. They don't because your requirements for, you know, attacks, you know, how you do folks's taxes, is very different than somebody else's. If you're in a big business, that's one thing, but they have their own way of doing it. If you're doing it yourself, and this is your business and your, how you run your business is going to be different. It may not work. It probably will not work. Well, no, most definitely it's not going to work. It's going to require some customization that comes with that. There is a cost to that. There's an implementation fee or there's some kind of fee that goes along with that. Don't be surprised by that. It's, it should be, you know, just the nature of the game and you want to pay it. Why do you want to pay it? Because it can be done faster with an expert than it can be done with you.
You can do a lot of things, but if it's very complicated, if it's touching a bunch of different APIs or other applications, it can get tricky and, and it can get dangerous really quick. It's just easier to pay it because if you don't, if you go back and you look at the amount of time that you tried to do it, you're gonna see one. You spent way more time than you needed to to. The costs associated with that are usually two to three times more if you try to do it yourself. So keep that in mind.
Now talks about, you know, the phase two and make sure you do it in the off season. That's good advice.
Client onboarding and what the initial development is going to look like, meaning how is it going to interface with your clients? How are you going to be able to push that over to them? How are you going to remind them, hey, don't lose your receipts.
Let's not dig through receipts for, you know, six hours trying to find one that was really expensive because it's not worth it. So you've got to keep that kind of thing in mind. As you're going through, making sure that you've got the campaign set up right, everything is in play, then we'll, you'll.
[00:47:59] Speaker A: Be fine doing it.
[00:48:00] Speaker B: All right, so look at things, look at the cycles, ask Claude for help, and then you'll be able to do it just fine. Now, of course, the last piece is really around. What do you consider success?
It's going to help you with that. It's going to help you map out what those kind of deals are going to look like. How is it going to impact you, what other tools do you might need?
So it's giving you some advice that's going to really make a big difference. Okay. Now if you want agents to be able to facilitate that, we would have to dive deeper into this outline to say what agents do. I want solving which parts of the problem.
We don't have time this week to do that. But you see, you're breaking the problem up into multiple phases, multiple steps, and then when you get to a comfortable point where the agents can easily do those jobs or, you know, maybe it's got to be a more complex agent to do it, that's fine. But that's where you're using these platforms to automate, to make some smaller decisions, to accomplish some things, to interface with some apps and really make things a lot easier in, in the long run with you. Now that's it for this week's show. I hope everybody feels more comfortable being able to, to not be afraid when they hear agentic.
Yeah, it's, it's, it's, it's just again, repackaging, trying to, trying to stay ahead of, you know, the, the marketing, you know, spiel that's, that's going to be coming out. It's not groundbreaking.
People have been doing this for a long time now. We've got large language models.
Our ability to interface with those and do some of these things quicker, get some answers quicker, allows us to really, at the end of the day, deploy some of these quicker. That's it.
Nothing to be afraid of. It's just like all the other things that we talk about.
You know, it's another way, another tool that you have in order to make faster decisions, save time in your life and get an advantage over your competitors. That's really all it is.
So I hope everybody learned something. Look forward to the emails for, for any questions that you had. Got a lot of emails this week. I hope I was able to answer, you know, most of those emails around agentic and those kind of things. So I appreciate you all. Thank you again for watching. And we'll be back next week with another great show.
[00:50:48] Speaker A: This has been a NOW Media Network's feature presentation. All rights reserved.
