Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
5 Stages of AI Agents & How Marketers Can Prepare Now
276 Plays20 hours ago
AI agents are generating a lot of buzz right now, but how close are we to fully autonomous AI? In this episode of The AI-Driven Marketer, host Dan Sanchez (Dances) cuts through the hype to break down the five stages of AI agents and what marketers need to do to stay ahead. From simple automation to full AI autonomy, Dan lays out a roadmap for how AI will evolve in the coming years and how marketers can start preparing now. Whether you're just getting started with AI automation or you're ready to build intelligent workflows, this episode is packed with actionable insights to future-proof your marketing strategy.
Resources Mentioned:
- AI Fundamentals Course: https://danchez.com/courses/ai-fundamentals/
- N8N (AI Automation Platform): https://n8n.io/
Timestamps:
00:00 - Introduction: The AI Agent Hype
02:00 - What Is an AI Agent? Defining the Term
04:30 - The Five Levels of AI (From Chatbots to Full Autonomy)
07:45 - Where Are We Now? AI's Progress in 2024
10:15 - The 5 Stages of AI Agents
12:00 - Stage 1: Automation with AI (Simple AI-Powered Workflows)
14:45 - Stage 2: Intelligent Automation (AI Begins to Think for You)
17:30 - Stage 3: AI-Assisted Automation (AI Helps Build Campaigns)
20:15 - Stage 4: AI-Led Automation (AI Makes Key Decisions)
23:00 - Stage 5: Fully Autonomous AI Agents (The Future of AI in Marketing)
26:45 - Predictions: How Fast Will AI Agents Advance?
29:30 - Practical Steps to Prepare for AI Agents Today
32:00 - Why Most Marketers Are Still Behind (And How to Get Ahead)
34:45 - Closing Thoughts: The AI Learning Curve and Next Steps
Key Takeaways:
- AI agents are not fully autonomous yet, but they are rapidly evolving.
- Most "AI agents" today are actually just automated workflows with AI enhancements.
- The transition from AI-assisted automation to AI-led automation will be a game-changer for marketers.
- Marketers should start experimenting with AI-powered automation tools now to stay competitive.
- Full AI autonomy is likely less than two years away—prepare by learning automation now.
Subscribe & Review:If you enjoyed this episode, please subscribe and leave a review on your favorite podcast platform. Your feedback helps us bring more AI-driven marketing insights to you!
Recommended
Transcript
The Hype and Reality of AI Agents
00:00:00
Speaker
There is a ton of hype around AI agents out there right now. And if you're like me, you're looking at all the headlines wondering like, are are they a thing? Are they here? But I think it's mostly companies trying to get ahead of the market, trying to get the hype out there, trying to get people interested in buying stuff you don't need right now.
00:00:17
Speaker
As a marketer, I've been surveying the land and trying to figure out what's what and how it's going to pan out because I think it is going to come and it's going to come quickly, but it's not here yet. So in this video, I want to present the five stages of AI agents and how to make the most of it right now.
Understanding AI Agent Stages
00:00:34
Speaker
Welcome back to the AI-driven marketer. I'm Dan Sanchez, my friends call me Danchez. And today i really wanna unpack this topic because I feel like there is just so much confusion around what an AI agent is, what it's capable of, what we have now, what's gonna come in the future.
00:00:49
Speaker
And so I thought I'd make it clear. And I spent a lot of time thinking about what these stages would look like, what we currently have, what we're starting to see and what some of the I'm hypothesizing what some of the next steps will be before we actually have full blown agents. And I'm going to define it.
00:01:04
Speaker
We're going to break it down real simple. And I'm also going to show like examples of every single stage, even the ones that I think are a little bit more hypothetical right now so that we can actually start making the most of what we have now to be prepared for what's to come.
00:01:18
Speaker
So with that in mind, let's review the big picture of where we're at with AI and then talk about where agents fits into it so that we can take advantage of it. Now going to open up my deck right here. And if you're watching it following along with me visually, it's just going to be plain to you.
00:01:33
Speaker
But if you're listening to this, just to remember that I'm going to be explaining everything we go through it here. And you won't be missing much if you're not following along with the video. But if you do want to follow along, check out the YouTube channel for this channel or open it up in Apple Podcasts. If you open it up this episode full width, you can actually watch the video right in Apple Podcasts.
00:01:53
Speaker
So let's go ahead and look at this. I have here an upside down pyramid with five different levels with one at the bottom and level five at the top. And it is actually the five different levels that we're going to be talking about that OpenAI, Sam Altman put out this summer, that became a really helpful tool for me to understand where we've been, where we are and where we're
Autonomous Project Handling by AI
00:02:13
Speaker
going. So let's review the five levels of AI right now.
00:02:16
Speaker
We start off with level one, which is chatbots. This is what landed two ah ah a little over two years ago in November of 2022 when ChatGPT kind of hit the market.
00:02:27
Speaker
that Everybody became aware of where AI chatbots were. like That kind of opened up this level one. The next level is reasoners. That's where we finally got 01 Preview this last summer. And now we have 01 Pro and 03 Mini High just came out a few weeks ago and more coming out soon.
00:02:49
Speaker
Reasoning models are the next level of AI. We have Agents. This is what we're talking about in this video, right? These are... These are AIs that can go and do multiple things for you and act independently in order to accomplish not just tasks, but whole projects.
00:03:05
Speaker
And we're going to dive more into that in a bit, but let's continue on. We also have innovators, AI that can do innovative things, and not just follow a task, but actually create original stuff.
00:03:16
Speaker
And then we have organizations, AI that's actually able to lead a whole organization, right? So you have an AI leading a lot of AIs to accomplish the work of an organization. It's pretty, pretty intense.
00:03:28
Speaker
So we are not at that level yet. Right now we are in between, we are at level two and we are on our way to level three.
00:03:39
Speaker
So I have to ask the question, where is this all going, right? That's where we are. But I wanted to break down step by step for how I think the future will go. And this can get hard, but I think over over time here, AI is starting to show some predictability in how this thing plays out.
00:03:54
Speaker
So I'm going to give my best shot at trying to bridge the future here. So for starters, it took us two years right here in order to get from chat bots to level two reasoners, right?
00:04:07
Speaker
It was near just a little under two years by the time we had chat GPT 3.5 launched and when we got 01 preview, and when we got ah ah one preview Two years before we got 01, right? It came off a preview.
00:04:19
Speaker
So two years, it took a long time. i remember thinking right after four came out, everybody was asking, well, when's five come out? Maybe next fall, we were saying no, five still hasn't come out, right? So but it took a long time to get to that next one.
AI Development Timelines and Predictions
00:04:31
Speaker
Now, Sam Altman has said, now that we have reasoning models, the path to AGI will take a lot less time. I'm going to say less than two years. right We don't know how long it's going to take, but Sam Altman's confident, and probably because he's seen the models on the back end, that it's going to take less than two years to have full agents.
00:04:50
Speaker
So based on the progress curve we see, this was something OpenAI put out just this last December, just a few months ago. You can see the rate of innovation just based on these Arc AGI semi-private V1 scores over time. And I'm looking at a graph here of like, it's like flat line. We have ChatGPT 2, 3 barely goes up.
00:05:09
Speaker
ChatGPT 4, 4, oh, it's starting to peak up a little bit just above the, you know, the X curve or the X line on this graph. And then we have, it just starts to sky go jump up a wall, right? When it comes to time over the scores on this, this graph, it goes from chat GPT 4.0 to scoring a 20% benchmark with O1 Preview to like a 30% benchmark with O1 High, a 50% benchmark with O1 Pro, and then O3 Low, which is unreleased. Actually, I think it's what we have now, maybe with Mini. I don't know. I've heard Mini slants somewhere, but like just behind O1 Pro. So it's like right here.
00:05:45
Speaker
But O3 that they're currently working on will be coming out soon scores just under the 80% and just probably around 87% on its highest fine best model yet.
00:05:57
Speaker
And it's yeah remember it's going from chat GPT for oh, which is maybe scoring at like a eight. 8% correct on this test. So it just and this is just one test. There's many different ways to test AI. There's many different places out in the wild where it goes where it does well or doesn't do well.
00:06:14
Speaker
This is just one graph, but it really starts to show the wall that we're climbing here when it comes to rate of innovation, the rate of improvement. So it makes you start to look at this. Hey, well, how fast will this come?
00:06:25
Speaker
Well, we know it's probably going to be less than two years, but is it going to be six months, eight months, one year, maybe 18 months? We don't know, but it's coming fast.
00:06:36
Speaker
And I've actually, I thought this was going to be a year away. It might end up being summer when we see what I'm predicting is going to be like a full autonomous agent, but it might be coming a lot faster. So let's break down the five stages and actually try to think about what's actually coming with these agents.
Stages of AI Automation
00:06:53
Speaker
So again, I have a line on this graph with level two reasoning slash problem solver, because that's the stage we're at right now in the five stages of AI. And then we have level three AI agents, which I'm calling calling AI agents, the fully baked agent,
00:07:10
Speaker
that can go and do your bidding, uh, no matter what, like with, within parameters. Right. so let's break down step-by-step what the five stages of AI agents will look like.
00:07:21
Speaker
Stage one automation with AI. We can already do this. Like I've been doing like automation with AI since this last summer when I made the AI fundamentals course, take it if you haven't,
00:07:35
Speaker
It uses marketing automation within the high level CRM and just runs out to ChatGPT to go and fetch some data, plug it into a cell or a field for the contact record to it then inject into an email that goes out to them to customize the course on the AI fundamentals.
00:07:52
Speaker
You can go to danchez.com slash courses in order to find that course. to test it out to see what it's like. But automation with AI, most people, what even people are calling these agents right now. And I'm like, like it's not an agent.
00:08:06
Speaker
It's just automation that has a little bit of AI in it. And let me give you an example. As I just created this, awesome automation to automate the distribution of this podcast. So it goes from being the podcast RSS feed to YouTube, and then it creates a blog post, publishes it to my WordPress site, takes a featured image, crops it, gets it up there, it creates LinkedIn posts.
00:08:28
Speaker
I'm still working on this automation, by the way, and I'm hoping to preview this whole automation with you very soon. It's using an awesome program that I only just found out about from my friend Audra Carpenter, who was a guest on this show just just a few months ago.
00:08:42
Speaker
And it's an awesome automation program, N8N. Kind of a developer, not great name, but just Google it, you'll find it. So I'm going to be doing a full breakdown of this in a future episode. So stay tuned for that.
00:08:54
Speaker
But this this isn't an agent. I've had a lot of people, i I posted the same picture to LinkedIn. People are like, oh, look at this agent. I'm like, this this ain't an agent, ladies and gentlemen. I had to painstakingly tell this sequence what to do, when to do it, where to get the information, where to put the information. like it And it was painful. like I had to run into a lot of errors, had to figure out how to do a bunch of different things, had to work through. And it took me a long time for every little block you see on this diagram.
00:09:20
Speaker
And we're looking at a diagram. If you can't, if you're listening to this, it's like this, like, it looks like a typical automation sequence where a bunch of blocks are linked to each other. And you can see the journey that this automation takes. Nothing really complex or nothing really fancy to look at. It's just like, it looks like a flow chart or a diagram.
00:09:39
Speaker
But this automation isn't an agent. it's It's simply an automation where a few times into it, I make a call to Google Gemini or I make a call to ChatGPT and be like, hey, here's some information. Give me back this information.
00:09:52
Speaker
I stick the information. It gets back to me in a field and then on the automation it goes. That is not... what I would call agentic that is not acting, but it is certainly a step towards agents and what I would call the first step.
00:10:06
Speaker
So this has been possible, like even before, like it chat GPT came out, like its API was available for people to start injecting AI into automation. In fact, most AI before chat GPT, this is what it was, it was just a little bit of AI baked into very robust automations.
00:10:24
Speaker
But let's talk about the next level. And that's what some people are calling intelligent automation. Now, there's a difference with intelligent automation. It's not just doing a little bit of AI in the automation.
00:10:37
Speaker
It's actually doing much more significant. And we can do this now in N-A-N. Man, horrible name. They actually have a tool that I haven't started using yet, but as soon as I saw it, I started realizing its potential.
00:10:49
Speaker
They're calling it an agent. You can see even here, I have a screenshot of one of these agent nodes in NAN within an automation sequence. And you can see it's a little bit more sophisticated than the ones we were looking at before, where it's just like a rectangle and a little, you can attach some tools to it. This one's more, because not only can you determine the the chat model that's attached to this or the agent,
00:11:11
Speaker
the AI model that's attached to this, but you can actually give it memory to different kinds of databases. You can give it tools to go and work with. And all of a sudden, yeah, you're going to load it with instructions, probably put some documents into it, kind of like a custom GPT.
00:11:25
Speaker
It's now a much more robust situation where it has to think through with the model you've attached to it how to use these tools and access real live data. Like you can hook up your WooCommerce store to it. You can hook up your air Airtable database.
00:11:41
Speaker
Now imagine for a second, if like I wanna continue on with this, podcast distribution. It's writing a blog post for the podcast and it's embedding the YouTube video into it from the podcast. That's great. It's already doing that.
00:11:54
Speaker
But what if I went a step farther and hooked up an AI agent to it and said, hey, here's the blog post. I want you to add some compelling links to previous blog posts I've written on this topic.
00:12:06
Speaker
Go and search the database with this blog post in mind. And now I want you to rewrite the blog post and add a few different links back to two the back to sorry one to three different articles that I've written before podcast episodes that I've had before on this topic so keep it in mind find the links find the topics that has been written about this on and then go and add some links to it and now it has to think through much more complex array of information it's got to search a database it's got to take multiple steps probably going want to hook up a reasoning model to this because it actually has to reason through multiple steps
00:12:41
Speaker
in order to accomplish this objective and so that my blog post can have a lot of internal link links baked in and it's fully automated. It's like I don't have to touch it again. I might have to review it, but it's a lot more complex than being like, here's some information, go write a blog post, get the blog post back, move it on, right?
00:12:57
Speaker
It has to walk through and think through a number of different steps in order to accomplish this one specific task. So this is definitely a level up from just AI with automation and what I'm calling intelligent automation.
00:13:10
Speaker
But it's still just another step forward in our path to a agents. So let's go ahead and take a look at the next step or stage the next stage.
00:13:21
Speaker
So let's go ahead and take a look at the next stage in a agents. I am calling it the AI assisted automation. Yes. So we're going from AI with automation, intelligent automation in the AI assisted automation. Now, again, i'm a little I'm a little bit forecasting here, but I have actually seen an example of this.
00:13:41
Speaker
And I think we'll start seeing this a lot more. over the next year like i wouldn't be surprised if you start seeing this a lot more over the next couple of months because i've already seen it and it's i've seen this even a year and a half ago there is a crm called element 451 i worked there for a short time and they were doing this early i think they're way ahead of the market in fact of all the tech companies that i've seen element 451 which is this like car an industry that's really lagging behind which is higher education they have probably one of the most cutting edge like ai platforms that i've ever seen
00:14:13
Speaker
One of the cool things that it does is what I'm calling AI assisted automation. They have a bunch of different marketing automation tools inside this platform, kind of like in a HubSpot where you have your flow and workflow builders.
00:14:27
Speaker
And like many automation systems, they have a bunch of prebuilt templates. But I don't know about you. like I like building it from scratch, but a lot of people get like to use these templates, and then they have to go and fill it in with all their information. But what Element 451 does is they have AI assistance. I would say it's getting on the lines of agents.
00:14:46
Speaker
You can just say, hey, I'd love to build one that looks like this. And it would go and look through the library of templates, pick a template, and then begin filling in the template based on all the basic information it already knows, all the photos that it has of the university, all the basic data that it knows to fill in, and then actually fills in the entire automation. right And that's a bit different.
00:15:07
Speaker
Because before, remember I said I had to painstakingly go and drag each node onto the sequence with NAN in order to get this sequence to work, right? But with something like element 451, it's just going to pull the template. And it's not just going to pull the template for me to go and then and customize it. No, it's customizing it for me.
00:15:25
Speaker
That's a whole different level of automation that AI is going and using. And it might even have some AI elements in it where it's going to chat GPT and getting back some custom information. I don't think element 451 has some AI elements in it where it's pulling that, but maybe it does. It's been a while since I've actually looked at this.
00:15:42
Speaker
So that's starting to get to what I'm calling AI assisted automation, but that's not it. We have two more stages. So stage four is AI led automation.
The Future of AI-led Automation
00:15:52
Speaker
Okay, so let's go and revisit. I have not seen this one yet, but I imagine it will come very soon.
00:15:59
Speaker
Imagine that we have element 451, right? And it's not just pulling off from a bunch of different templates of automated campaigns, but it's actually thinking and building the automation for you from scratch. And it a even has some very smart looking intelligent automation pieces in it.
00:16:17
Speaker
where AI is plugged into the automation, is actually making decisions about should the user go down this way or that way. If they have this, then that. But it's not using basic search features like, oh, if they have if they selected this degree, send them down this path. If they selected this degree, send them down this path.
00:16:36
Speaker
No, it's actually using AI to assess a broader situation in order to guide them through an automation. But it's still an automation. But it's not pulling from a template. It's actually thinking through what needs to happen where. So it's looking through all the different nodes. It's looking through all the different engagement pieces in the automation and then architecting it out for you really quickly, customizing it.
00:16:56
Speaker
All you're doing is giving it instructions, looking at it, and maybe even being like, oh, I don't like how this one works. Like, let's make some changes here. And kind of like you make changes to a blog post with ChatGPT, you're having a conversation with it and it's actively building the automation for you.
00:17:10
Speaker
with some really smart looking intelligent automation into it. So that's a whole that's that's a lead automation, right? It's not just assisted. It's a lead a leading the automation. A is leading this kinds of this stuff.
00:17:25
Speaker
I actually think this stage will exist and will be really helpful. But I think we will breeze over this particular stage really quickly, because by the time it's able to do this, We are literally just a moment away from actually getting to what I'm calling the net last stage, the fifth stage of AI agents, which is the fully full autonomy or fully autonomous agents.
00:17:48
Speaker
And that's what I mean when I'm talking about AI agents. I want to be able to give it an objective and it just builds the whole automation campaign for me, even if it's only used one time. But it might be something that gets used a couple of times or be modified ongoing. Or you have to build a whole new automation campaign every single time because it has to think through all the different objectives, all the different tools it needs, the sequence to be built off of.
00:18:12
Speaker
Maybe I am running a university as a marketing director and I want to build a new campus preview weekend because we're to inject some new special campus preview weekend to come visit the campus. I need to build a whole campaign for it, but I only need to do it one time. This is a special opportunity that we're doing.
00:18:27
Speaker
And so it needs to build the whole sequence of who to invite, when to send it what to send with when they get here after they get here. And it's a lot of campaigns to build out. But an agent will be able to just be like, but build it out, you execute it, and then it's done. Or even just I want to send out a campaign that's today.
00:18:42
Speaker
And it's multiple steps across multiple channels. And the AI agent will be able to go and collect everything, set it, schedule it, and then start working on it immediately. That's an agent. I believe it's going to go a lot faster because one prediction that I had is that it will be really hard for AI agents to do this because it needs access to all these credentials and APIs.
00:19:02
Speaker
But now that they're starting to kind of unleash AI agents just directly in the browser, if you have if you just give it permissions and passwords and maybe give it like some credentials and some logins like you would an employee, well, you don't really need all those APIs anymore, even though the API me...
00:19:18
Speaker
Maybe more efficient for the AI to deal with for itself, right? Because clicking around and looking at contact records is kind of a pain. But if an AI agent could just go search through an API of your CRM real quick, you'd probably find the information a lot faster itself.
00:19:33
Speaker
So again, this is less than two years away, what we're calling AI agents, even fully autonomous agents. This is less than two years away. We don't know how long it'll take to get all the way from level two to the complete level of level three.
00:19:50
Speaker
But it's going to be less than two years based on what OpenAI is putting out there. And based on the rate of progress that I'm currently seeing, even though we're early on in this race from level two to level three, it it I think it's going to be less than two years. I think it might even be one year.
00:20:06
Speaker
We might even start seeing signs of a full autonomy. In fact, we already are. We are already starting to see some signs of full autonomy now with ChatGPT's operator.
Preparing for AI Integration in Workflows
00:20:18
Speaker
It's not very good, but it's starting to show signs that we're getting into that direction very quickly. And we will see all of these stages along the way.
00:20:28
Speaker
So as you're seeing this, you might be overwhelmed thinking like, oh my gosh, I haven't even started building any automations at all. I want to encourage you with a few different things. One, if you're listening to this and you're already in chat GPT, you're already using AI tools. so Let me tell you, the first thing you need to do to start preparing for this reality is building custom GPTs or what I call specific custom GPTs, ones that walk through multiple steps.
00:20:51
Speaker
I promise that building things with custom GPTs first is a great precursor to warm you up for the automations that you're probably going to be building and working on with AI and AI agents very soon. It's the best way to put in the reps of thinking through how AI can execute something step by step in your workflow.
00:21:08
Speaker
If you haven't done it before, I highly recommend taking my AI fundamentals course that will take you from beginner to advanced very quickly. It's a five-day video course, and it even has a customized email that perfects and customizes the lesson just for you and your role and your audience that you're marketing to.
00:21:27
Speaker
The next thing you want to do to prepare, if you've already built some custom GPTs, you're like, I'm ready to roll up my sleeves for more and get ready for this agent thing, is to actually start working in order of the stages I've just laid out.
00:21:38
Speaker
Start with doing some automation and then pulling OpenAI or Gemini or Claude or whatever into the automation. How can you inject AI into the automation in little places in order to make your workflows and your process as a marketer just go more smoothly?
00:21:52
Speaker
Even if it's doing something that maybe only saves you a little bit of time, just having the work of building it and getting it gives you the kind of repetitions you need in order to get good at this skill. And then go on to step stage two, where you're starting to not actually just inject a little AI into it. AI is starting to think with you and actually you start to work through some tools in order to make more intelligent decisions for intelligent automations.
00:22:14
Speaker
Soon we'll have some AI tools that'll be able to like draft campaigns for you. and even lead the charge right now. I haven't seen a lot of those tools unless you're working with element 451. Yeah.
00:22:25
Speaker
But as soon as those tools are out, I'm sure to be covering them in this very show. So just keep listening to the show. I'll will announce when those are out for now. Just start working through stage one in stage two. I am and I haven't even gotten to stage two yet. Hopefully soon.
00:22:39
Speaker
And the last thing I wanted to bring up is just to remember, it takes most companies and individuals a very long time to actually grab hold of new technology and new ways of doing things and actually implementing it in their business.
00:22:51
Speaker
I know like all the hype and all the influencers on LinkedIn and all the different people talking about it on X and all that stuff makes it sound like everybody's doing it. I promise nobody's doing it. like very, very few people are actually doing this stuff.
00:23:03
Speaker
And they very few will continue doing this stuff for another year or two. We will literally have AI agents out and people will be like, Oh, I haven't logged into chat gbt yet. the this We were really early right now.
00:23:14
Speaker
And if there's one thing I know is that most companies are just really slow. Most companies still don't implement inbound marketing right. And that came out in 2006, 2007. Like, come on. They still don't know how to do content marketing properly in order to build a list, draw people in, and stop blasting them with like a traditional marketing approach. That's still a thing.
00:23:34
Speaker
So this is going to take a long time for people to catch up. If you're listening to this, you are way ahead. Start getting your hands dirty by building some custom GPTs. Take my free course and start thinking about how to build an automation. Again, episodes coming soon about how I'm using and NAN in order to build the automation that powers the distribution of this very podcast.