Episode Categories:
Resources:
Join the discussion: Got something to say? Let us know on LinkedIn and network with other AI leaders
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
AI Agents in 2025: Practical Business Considerations and the Enterprise Path Forward
The AI landscape is evolving with unprecedented velocity, particularly in the development and deployment of AI agents. Enterprises face a new reality: questions around “Will there be more AI agents than humans?” aren’t simply theoretical but guide investment decisions and operational strategy for the coming year. This article distills the critical insights from a recent industry discussion on the present and future of agentic systems, drawing out tangible implications for business decision-makers.
AI Agent Proliferation: Current Numbers and Definitions
Current estimates suggest that active AI agents may already be rivaling or exceeding global human population figures, depending on definition. For instance, agentic versions of large language models such as the widely-used chat platforms tally up to a billion monthly active sessions, each capable of acting as a stand-alone agent. Moreover, as advanced features—like autonomous scheduling, email integration, and personalized reporting—roll out into mainstream products, agentic processes are no longer an expert-only affair. The definition continues to expand, encompassing everything from scheduled background workflows in enterprise SaaS to personalized digest emails that integrate with calendars and productivity software.
Enterprise AI Agent Infrastructure: Building at Scale
Leading technology providers have adapted existing infrastructure to support planetary-scale agent deployment. Network companies hosting web applications have repurposed distributed server networks, originally designed for web security and reliability, to run agentic applications close to physical users—with latency as low as 20 milliseconds. Platforms now offer agent software development kits (SDKs) that let teams quickly build AI-driven tools to automate business communication, reporting, and integration across silos (e.g., connecting Slack, Gmail, GitHub, and internal wikis into actionable, autonomous reporting agents). Weekly download numbers exceeding 150,000 on developer package registries underscore not just hype, but surging enterprise experimentation.
Business Process Automation: From Subject Matter Expert Tools to Everyday Use
Business process automation remains the primary use case for agentic systems. These agents are not futuristic, sentient companions; they are practical, low-level digital interns capable of amplifying business process efficiency and scale. Organizations leveraging agentic tools automate reporting, communications, and repetitive workflows across platforms—moving beyond isolated scripting and into integrated, multi-platform orchestration. The capability leap is significant: agents now autonomously coordinate across ERP, CRM, and communications suites, handling tasks that previously required manual intervention from specialized personnel.
Digital HR for Agents: Managing Multi-Agentic Environments
Rapid agent deployment presents unique operational challenges. Unlike traditional software, deploying dozens or hundreds of agents per human operator requires “digital HR”—systems for tracking agent performance, observability, and process fidelity. Enterprise leaders must evaluate and implement management frameworks that monitor goal completion, error rates, and security boundaries. The scalability of agentic systems makes digital oversight a necessity; it is now an essential consideration in technical architecture and workforce planning.
Adoption Strategy: Experimentation and Familiarization as a Business Imperative
The uncertain but inevitable improvement of agentic technologies means that hesitancy exacts a real cost. Business leaders should prioritize their familiarity and hands-on experience with agent systems, using existing mainstream tools for automating low-stakes business activities. From personal productivity (auto-generated reports, meeting scheduling, communication digests) to more ambitious internal process automation, active experimentation enables organizations to identify valuable, scalable use cases amid fast-moving technological progress.
Future-Proofing: Social and Technical Trust in Agentic Systems
Despite ease of deployment, trust and transparency remain central to successful enterprise adoption. Organizations must invest not only in agent capabilities but also in standards for secure agent orchestration, clear boundaries for automated action, and robust onboarding and education for staff. The process mirrors the historical rollout of website builders and database tools—technology races ahead, but sustainable value depends on trust, socialization, and responsible operationalization.
Conclusion: Practical Steps for 2026 and Beyond
As AI agent capabilities become default features across software ecosystems, competitive enterprises will shift their focus from questioning utility to discovering untapped opportunities for efficiency, scale, and automation. Early, active engagement with agentic platforms—especially for automating mundane and repetitive business processes—positions organizations to lead rather than lag. The strategic priority is clear: ensure organizational literacy, build robust management infrastructure for agent oversight, and remain vigilant as both capabilities and potential risks evolve.
Direct, hands-on experimentation and process adaptation are the way forward. In the next year, the conversation will not be about what agents are, but about what remains to be optimized, and how they can further drive measurable business value.
Topics Covered in This Episode:
- Combining Multiple Features in Large Language Models
- Visualizing Data in ChatGPT, Gemini, and Claude
- Creating Custom GPTs, Gems, and Projects
- Uploading Files for Automated Data Dashboards
- Comparing ChatGPT Canvas, Gemini Canvas, and Claude Artifacts
- Using Agentic Capabilities for Problem Solving
- Visualizing Meeting Transcripts and Unstructured Data
- One-Shot Mini App Creation with AI
Episode Transcript
Jordan Wilson [00:00:17]:
I had kind of a bold ish prediction a year ago when it came to AI agents. I said that by the end of 2025, there would probably be more AI agents than humans. So here we are at the end of 2025. And although there is probably no definitive answer to that question, I think it's safe to say that it's a valid question. Are there more AI agents than humans? Well, I don't know. But also as we're planning, and looking into 2026 and looking back at maybe agentic efforts from 2025, I think it's important to kind of take a look back and see what's been working across the industry, what hasn't, and what should we be looking forward to in 2026 when it comes to the future of AI agents, in the enterprise. So I'm excited to tackle that on today's episode of Everyday AI. What's going on y'all? Welcome.
Jordan Wilson [00:01:14]:
My name is Jordan Wilson. I'm the host, and this is for you. It's your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me make sense of all that's happening in the world of AI. My gosh. AI agents are changing literally by the hour it seems like. So every day AI, we help you make sense of all of this and, grab the actionable insights out of everything to grow our companies in our careers. So if that's what you're trying to do, awesome. It starts here with the unedited, unscripted live stream podcast.
Jordan Wilson [00:01:42]:
But to take it to the next level, make sure you go to your everydayai.com. Go sign up for our free daily newsletter. We're gonna be recapping the highlights from today's show as well as keeping you up to date with all of the other AI news that matters. Alright. I am excited, for today's guest. It's gonna be a good conversation. Whether you care, about AI agents, whether you're using one, a 100, or a thousand, I think today's conversation is one that you're going to need to pay attention to. So, live stream audience, please help me welcome to the show, Sunil Pai, the tech lead of AI agents at CloudFlare.
Jordan Wilson [00:02:17]:
Sunil, thank you so much for joining the Everyday AI Show.
Sunil Pai [00:02:21]:
Hey. Thank you so much for, having me. I'm very excited to be here. I also like that my name itself has the word AI in it, and I'm kind of realizing it.
Jordan Wilson [00:02:30]:
Yeah. It is. That's that's, mine doesn't. Right? But, some like, my my friend said I should, adopt the nickname of Air Jordan, but have the AI capitalized with a lowercase r, but, you know, just just never never really stuck. But, Sunil, can you tell us a little bit about what you do in your role at Cloudflare?
Sunil Pai [00:02:50]:
First of all, like I said, thank you so much for having me. I love talking shop about these things all the time. If you follow me on Twitter, you know this. I'm the tech lead of, of the AI agents effort inside CloudFlare, which is a, network infrastructure hosting company, that's basically spread across the planet. The roughly, like, one in two websites are backed by CloudFlare in some way, which is kind of an insane thing for me to understand. So I've been focusing on, like, building out AI agents for ourselves and for our customers over the last year. I also have them build out the developer CLI. Previous to this, I was on, like, the React JS coding.
Sunil Pai [00:03:31]:
I've done, like, a bunch of things. But, yeah, the focus is the topic. I wanna say topic du jour, but, honestly, I think we are gonna be exploring this for a few years, Yeah. Agent's one.
Jordan Wilson [00:03:41]:
Yeah. And and and talk a little bit, you know, I think most people know CloudFlare. Right? One of the, the the largest, organizations in the world in terms of, you know, having speed, security, reliability, you know, for online, you know, applications and websites. But can you explain a little bit about what you have on the agentic side because you do have a platform for building AI agents. Before we dive into all the fun stuff, tell us a little bit about what CloudFlare is doing on the agentic side.
Sunil Pai [00:04:09]:
So the story actually starts closer to seven, maybe eight years ago, where we've built out this network that helps protect your website and your servers from DDoS attacks, hacker attacks, any kind of, well, like, attacks of any sort. And then we decided to run JavaScript on all these servers across well, now I think it's something like 300 plus cities, 15,000 points of presence. The statistic that I really love is that I think close to 90% of the world's population is about twenty milliseconds away from a CloudFlareBot. So at one point, we are like, we're gonna run JavaScript on all of these things so that you can build out all your web applications, APIs, things that, you know, mobile apps connect to, etcetera. And, we've been building it out for a few years, and when LLMs took off a couple of years ago, because CloudFlare is such a, for lack of a better word, very innovative company, we noticed that some tech that we had built out, namely, like, durable objects and the network itself, were extremely good ways of running these little agentic applications. And when I say an agentic application, I mean something that that you can connect to an LLM, run this agentic loop where you're like, oh, has the goal been reached yet? Can I talk to these tools? Can I, you know, connect to my Gmail and my Salesforce and so on? And, keep doing it in the background without a browser connected to it or anything until it reaches that goal. So we took that seed of an idea, and we built out a library and supporting infrastructure in the platform to build and deploy these. So you can go in like, if you use the agent software development kit of which, for which which I build, like, every day now with my team, it's as simple as, yep, I point my little coding agent to our example which, like, oh, here's a Slack bot that I wanna take.
Sunil Pai [00:06:10]:
I wanna connect it to my Gmail, etcetera. And I want I also want to give it scheduling ability, so I want to be able to say in Slack, hey, little bot. You can give it a name, by the way. I sometimes say Freddy, because I I I like to choose the name of a friend I don't really have. Otherwise, it's just weird. I I I can actually say something like, hey. Every Friday at 9PM, I need you to go through my email, all the code commits that I've made on GitHub, all the edits I've made on the Wiki, and I want you to compile them and send an email to my manager because he keeps wanting to know what I'm cooking. And I I don't really wanna waste time on a Friday evening.
Sunil Pai [00:06:52]:
I have better things to do on a Friday evening. And it turns out that these capabilities, scheduling, the ability to talk to inference, use Webhooks, we have, like, incredible WebSocket support. All of these things are so easy to build out and deploy in a planetary way. And the best thing about our AI agents is that they can run super close to a user, which is to say twenty milliseconds away from you or close to other geographical locations that you want. So, it's been a wonderful year building this out. I think we are at over, like, a 150,000 downloads on the npm registry, the JavaScript package registry, every week. And, it's been incredible talking to customers. And like I said, like, because customer because CloudFlare has, like, every customer, I think we do something like 40% of, like, Fortune 500 and thousands, millions everywhere else.
Sunil Pai [00:07:45]:
It means we have seen, like, the spread. I really enjoy that.
Jordan Wilson [00:07:50]:
Yeah. So, I kinda wanna skip to the end here, and and and then we'll circle back. But, are there do you think there's more AI agents than than humans right now? Right? Like, there's no definitive number. I've seen estimates anywhere from, you know, 600,000,000 to 3,000,000,000, but who knows? What's what's your take? Do you think there's more AI agents than humans right now, or will there be soon?
Sunil Pai [00:08:14]:
Okay. So I I can use simple numbers to explain this, and it boils down to what you think is an agent. Like, if you ex like, it's almost like the tuning test was solved, like, a year and a half ago, and we were like, yep. Cool. Awesome. Yeah. Move
Jordan Wilson [00:08:30]:
on. Yeah.
Sunil Pai [00:08:30]:
Think. Yeah. So chat GPT right now does roughly 1,000,000,000, monthly active users. Okay. So if you think if you can tell yourself that each chat GPT instance instance is an agent because it can do things in the background, you can actually ask it to schedule stuff and talk to tools and connect it to your Gmail, etcetera. Well, you're at 1,000,000,000 agents right there. Okay? So we are, like, eight x away. If you consider every running session to be an agent, now that starts multiply, maybe two or three.
Sunil Pai [00:09:09]:
ChargePD themselves, I think they're working on some tech called Pulse, I forget what it's called, where it's going to start being proactive. Based on everything it knows about you, it's going to start doing things at the background whether you tell it or not. We're like, hey. Next week's your, wife's birthday. Flowers, chocolate, movie tickets. Oh my god. The advertising potential there. And now that's just chat GPT.
Sunil Pai [00:09:31]:
There are so many in house systems. There are people who are deploying agents for all their customers. So you take somebody like an Atlassian, which has another few million users, etcetera etcetera, you quickly start getting to, like, 8,000,000,000. And that's, like, today. Like, one of my pieces is even if there's, like, no model improvements, which is not true, by the way, this I I suspect somebody is going to do a Christmas surprise and say, yep. We beat all the benchmarks. Like, these things are now happening on a weekly basis. I go I go to bed, I wake up, and, like, the world has changed.
Sunil Pai [00:10:03]:
K. So if if you even if you freeze that and all you do is you focus on getting this technology into the hands of regular people, just regular business owners, consumers, I don't know, teenagers, etcetera, you can cross 8,000,000,000, like, millions. And that's, like, today. I'm say like, give it another year, and that number is blown out of the water. You're kind of
Jordan Wilson [00:10:30]:
Yeah. I agree.
Sunil Pai [00:10:30]:
For any definition of AGL. This is this is
Jordan Wilson [00:10:35]:
Yeah. And and, you bring up a great point. Right? You said, you know, something about chat g p d pulse, you know, something they rolled out to, you know, their pro tier subscribers a couple of months ago. I read it almost every day. Right? I get it delivered, and I almost forget, like, oh, wait. That is an agent. Right? It connects to my calendar. It connects to my email.
Jordan Wilson [00:10:52]:
And every morning, it delivers a customized personalized digest, and I can give it feedback, and then it just goes out and does the work. You know, it seems like even just the the the very definition of what an agent is, what it does has changed drastically. Because I think you just gave a great example there of an agent that many people already have access to, and we're probably not even thinking of it as like, oh, this is an AI agent because it's just there, and it's easy, and it's accessible. Could you give me maybe your definition of an AI agent? And also maybe talk a little bit maybe on how that's changed or maybe it hasn't changed over the past year or so.
Sunil Pai [00:11:28]:
I mean, the biggest thing is a year ago, I didn't think we'd move this fast and be here. Like, I need to, like, revisit my timelines all the time. I just saw the new video model, released by Runway today, and it's indistinguishable from reality. It's crazy. Yeah. It's scary. Anyway, so what is my definition of an agent? So, I I, of course, work, in, I'm a software engineer, so I think of things in terms of software. So I like thinking of it as an entity that can take inputs.
Sunil Pai [00:11:57]:
So inputs can be as low level as something like HTTP requests, web requests, or we can think of it higher level. I can send emails to it. I can make phone calls to it. I can, send text messages to it. I can connect it to chat apps, etcetera. So there are all these inputs. I expect it to be able to do outputs, which means I expect it to be able to send emails to, say, make phone calls to come back to the chat. So the inputs and outputs for me are very important.
Sunil Pai [00:12:27]:
And then I like thinking about the agentic loop. You start giving it goals. These goals can be something that ends very quickly where you're like, hey. Can you find me, the top 10 pizza places in, The US? From what I hear, they're all in New Haven. That's what I understand. I had some New Haven style pizza, very nice, very crispy. But some of them can be open ended, which is until the day I die, you're going to keep looking at my health markers, which I will feed you via inputs with my Apple Watch, and you're going to tell me how to make myself healthier. What do I eat? How do I exercise? How do I sleep? That's it.
Sunil Pai [00:13:05]:
And that's really it. The combination of these input outputs and and this loop where it keeps trying to go towards goals powered by extremely smart models, that's honestly all it takes. Like, you can you can talk about, oh, it needs capability to do code execution or it needs the ability to do planning. Well, this is just enhancements on the same model itself, and I think, Yeah. I think the of course, the big thing is that you're now giving this magical entity the, intelligence and I like considering it hardware, but the ability to interact with the real world and get things done. This was forget about unheard of, like, two years ago. We this was only in the realm of science fiction. This was Star Trek, Battlestar Galactica, all of that.
Sunil Pai [00:13:55]:
Like but it's real. Like, you can use it right now.
Jordan Wilson [00:13:58]:
Yeah. And and I even remember going back, to, you know, let's just say, this time last year, right, when I made this, you know, kinda crazy prediction. At the time, I think it was maybe only a handful of the biggest companies that had, you know, true agent builders. I think it was November, 2024. Actually, no. Because it was my birthday, and I was at the keynote in Chicago, where Sadia Nadella, you you know, announced Copilot Studio. Right? But at this time last year, there were very few kind of low code or no code true agent builders. Right? And so now we already talked about what you do at Cloudflare.
Jordan Wilson [00:14:37]:
We kind of already referenced, you know, OpenAI. They have their agent builder. You know, Google just released their sales Salesforce. Everyone, it seems like, has a no code, low code agent builder no matter what platform you use. So what does that mean? Like, how should we be thinking about agents? Should should we be trying to deploy them, across any software we use? Because it seems like, right, if my website's hosted on CloudFlare and I use, Gmail, but I have a Windows computer, it's like, okay. Well, should I have agents running in all of those? How should we be thinking about where and when and how to deploy agents since they're everywhere?
Sunil Pai [00:15:14]:
Look. The the way I think about agents is I think about is exactly the way I think about software. Soft the first software company ever was actually like Microsoft, and I still think it's a terrible name for a software company. And it's been a while. It's they're not going to change it. Software, the number one use case has not been for, let's say, like, for sentient girlfriends, even though that's a thing that happens today. The biggest number one use case has been the most boring three word phrase uttered, which is business process automation. It has been a tool to amplify, people's capabilities, and, their impact on the planet.
Sunil Pai [00:16:01]:
It started off with, I don't know, like little shell scripts that researchers would use, but then it entered spreadsheets, data processing systems, ecommerce systems, to the point where, like, things that were the purview of very hardcore subject matter experts who paid a lot of money in the seventies and eighties to regular people nowadays. You you can't really make the joke anymore that, oh, it used to be that, like, my parents were bad at Google, like, using Google. But, like, it that that doesn't make sense in mobile because, like, everybody on the planet, young, old people, interact with software on a daily basis. Smartphones did that for everyone. Okay. So agents take that, and they give the ability to take a very enthusiastic, slightly stupid intern level intelligence with and help you amplify, like, your impact on, your business, your life as well. But the the big thing that I always think about are the machines that drive commerce and capitalism, like, across the world. That, for me, is, like, the number one thing.
Sunil Pai [00:17:08]:
So yeah. Absolutely. I think you should be, like, running these on your Windows machine and on CloudFlare and on wherever else good software is, is run. Yeah. And and you talked
Jordan Wilson [00:17:21]:
a little bit just kind of, you know, business process automation, right, which for the most part, that's where we've been for, I don't know, ten to twenty years at least. Right? Like, we've had software that can automate certain, you know, manual knowledge work tasks yet, you know, something I'm always grappling with. And luckily, I get to talk to a ton of smart people such as yourself. Right? But if AI agents are very capable of doing the exact things that many of us are spending the majority of our time doing, how should number one. Okay. Well, how can we even plan to deploy them? But number two, what should we be planning in 2026 and beyond for ourselves and, you you know, our human counterparts to doing? Because clearly, it has to be something more than just overseeing and orchestrating multi agentic systems.
Sunil Pai [00:18:18]:
When databases started becoming a bigger thing in the nineties, I remember that there used to be a real job called I mean, there still is, but it used to be a very big deal of database administrator, DDA. You would have to spend a lot of time learning about these things and running them on these expensive machines and mainframes and so on. And it was the job of, it it it many people took it upon themselves to just keep working at that problem for years and, like, decades to the point where it went from the purview of subject matter experts to, becoming like, you don't really you had to be a programming expert to have a blog in 2001 as as, as late as 2001. You don't think about that anymore. You make a Facebook account, you make a Twitter account, you're good to go. And those systems scale to, thousand orders of magnitude bigger than the kind of scale you used to have on the Internet back then. So it's actually, like, a harder problem to scale that now, but regular consumers don't see it. I think we are in a similar phase with AI agents.
Sunil Pai [00:19:28]:
If you're a subject matter expert or you're a developer, like, someone like me, our job right now is to basically figure out the process where we commodify making this regular technology for regular people. That's what we are doing right now. We don't want you to have to worry about your servers going down, etcetera, which takes me to the, next part of your question, which is, well, what happens when you start deploying these in, large numbers? And the phrase that I like throwing around is you need to have digital HR. Like, if you imagine a world where one human being has 20, let's say 20 agents, running at a time, That's like having 20 stupid employees, and that's not the easiest thing to, like, manage. So how do you look into how well they're doing, whether they're, getting their goals? How are they not, for lack of better phrasing, going out in the middle of the night and getting digital drugs? You know what I mean? And showing up, high to work the next so what is the equivalent of this? And you can't tell regular people, well, you have to set up observability systems and have metrics and Prometheus. No. So what is the commodification process of that? And that has always been a hard problem no matter how good technology is. You wanna make sure that it's, like, rolled out in a safe way, but still in a way that amplify you you don't want to hold back the capabilities, until, well, I don't claim to be, like, an expert in running business systems, so I need to give people the freedom to be able to take this and use them, like, safely.
Sunil Pai [00:20:59]:
And once these things happen, what do humans do? Man, I live in the North Of England. I'd like to go to the beach where it's warm and chill without worrying about how I'm paying, like, rent next month. This is all a little bit handwavy. It's why we have AI safety systems and the smartest people, not just in technology, but, like, in government wondering what people will do when a lot of this work is being run by I don't know. I just I'd like to be on a warm beach myself, I think. I would recommend that.
Jordan Wilson [00:21:29]:
Yeah. Absolutely. And anything an AI agent can do to help me spend more time on the warm beach, it's like, yes. Let's let's prioritize that. But, you you know, Sunil, you you you brought up an interesting point, you know, kind of talking about the history of of development and, you know, even, you know, how over time anyone could, you know, go and create a website. Right? I like to think sometimes, just the ease of creating a website. I mean, it seems like it took a really long time. Right? Even though, you you know, WYSIWYG editors, you know, what you see is what you get, they were around ten, fifteen years ago, but it really took a long time for them to be truly accessible, to the masses.
Jordan Wilson [00:22:12]:
What do you see on the agentic side? Because in my opinion, it seems so easy. Right? All all the big players have essentially the version or the equivalent of a WYSIWYG agent builder. Some of them, you just speak with your voice and nothing else. Right. So how should we be building these agents, and how can we prepare for the future when maybe our agents is that gonna be the only way to get our work done because everything is just going to be agentic by default? I think a lot
Sunil Pai [00:22:47]:
of this is speculation, on my side. I think we're going to have to, like, rethink. A, a, I think, as technologists, we're going to have to rethink the entire software stack where they can be run by these agents. Like, for example, just in Cloudflare, for example, there are so many checks and balances in place to make sure, a human is responsible for certain decisions that are made, configuration changes, code deploys, etcetera. Like, if we get it wrong, we occasionally take half the Internet, which is a little weird. We don't we don't want to do that. But now not only do we need to make these systems where we can let these agents learn blues, but we need to develop a sense of trust for these software systems, which kind of takes you back also to, like, fifteen, twenty years ago. These these concerns about the Internet taking over, what what they are going to do to the economy, how are how will the easy availability of information where anyone can say anything truthfully or they can lie through their teeth.
Sunil Pai [00:23:53]:
How do you develop a trust in technology systems? I actually don't think, like, we have, like, good answers for that yet. Like, my uncle and aunt, they still are highly distrusting of AI systems, and it's probably because no one has sat them down and explained to them how they work, what are the bound what is it that it can do, what is it it can't do. And every day, the list of the things it can't do gets shorter, which is also a thing. And which is why I think it still is going to be let's call it like a slow takeoff. No matter if the technology is good enough, building out just the social systems for people to understand how these things work, what they're good for, I suspect I I don't think there are good answers for it yet, but just like we as humanity, as a group of people, we figured it out over the last twenty, fifty, hundred, ever since the industrial, era. I think we'll figure that out. It's probably education. It'll happen.
Jordan Wilson [00:24:52]:
Yeah. Yeah. A 100%. And, you know, education is huge because, you know, probably just in the time that we've had this conversation, I'm guessing that there's been a lot of, even in this, you know, twenty some minutes, there's been new developments that, you know, even if you try to stay up every single day, that's what I do. You can't. There's always new capabilities. Right? And that's something I'm always having to look back and be like, I can't believe how far we've come in, you know, one one year, one quarter, one month, one week. Right? So as we look forward, alright, what You know what? I don't know.
Jordan Wilson [00:25:25]:
We wanna have this conversation next year. What might we be talking about?
Sunil Pai [00:25:29]:
Well, here's here's my first recommendation. There should be someone named Jordan who makes it his job where every day he has a podcast where he explains this to everyone who's willing to come and listen to it. That goes, like, a a long way towards making this kind of technology. This technology paradigm shift more accessible. So if you know someone like that, you should let them know. In one year, what can we look forward to? I think oh, here's the thing. Models will get smarter. They will be able to run for longer periods of time and do longer tasks.
Sunil Pai [00:26:08]:
It used to be a year ago, you could barely get a minute of work out forget about a minute, like, fifteen, twenty seconds of work out of any out of chat GBD. But now it can if you pay for the higher tier, it can go off and do things for, like, twenty minutes, half an hour just researching, coming up with ideas. Dude, it's coming up with, like, new drugs to fight cancer and diabetes. Huge. Like, oh my god. That's the thing I want technology to do for us. I suspect all these AI companies with a lot of money in the bank, they are going to work very, very hard to get this technology into people's hands for whatever purpose because they need to prove their valuation, sure, but because they actually believe this technology is going to make people's lives better. If you know anyone who works at these companies, you know they're not like, they actually care deeply enough about affecting, like, social economic change in on the planet.
Sunil Pai [00:27:02]:
So I suspect the conversation in a year is not going to be what is AI, but, what do I not know about AI that I can use to make my life better? Because by this time, I'll already have these agents working for me in my phone, in the house. Turns out Spotify is now going to let you prompt what music you want. I'm really looking forward to that, because I like music from, the before the year 2000, and I would like I don't like music after the year 2024. Like, I I've I've lost touch. I'm too old. Same. Right? I just, like, just I I need guitars. I need drums and bass.
Sunil Pai [00:27:40]:
Like, just play me music only with that. So, I think the goal of humanity has always been, like, radiating outwards from identity. How do I make my life better? How do I make my family's lives better? And how do I impact the outer world to, to be a good place to, like, live in. And in a year, we're going to be looking at how do I use technology to do all these things. It has always been the story with technology, and I think we will be continuing that part of the conversation.
Jordan Wilson [00:28:09]:
Mhmm. Alright. So, Sunil, we've covered a lot in today's conversation just talking about the future of AI agents, everything from challenges and opportunities to observability, safeguards, accessibility, etcetera. What's your one, kind of biggest takeaway for business leaders that maybe are still grappling, right between the huge upside of AI agents and their capabilities with the maybe sometimes unknown or unforeseen, pitfalls. What's your biggest takeaway for them? Oh,
Sunil Pai [00:28:42]:
it doesn't matter if the technology isn't good enough right now, whether the models are not smart enough. We have almost a guarantee that they will get better, which means your job right now is to become familiar with these systems and just to, like, get experienced in using them so that you know what they're good at, what they're not good at, and how they could benefit you. I guarantee you, most people are either afraid or they don't have the time or even worse, some of them have been told by other people that this is just bad technology. Like, there's, there's, there are rumors and there are people who don't understand it and are influencing other people. Your job as a as an entrepreneur, as a technologist, right now is to find familiarity in the systems so that when the opportunity does come, you to you, you know what next step you can take. We were like, oh my god, Sunil Pai was talking that I could build a little AI agent that can answer my phone calls for me and my and I can have an email agent that I can send all my invoices to so that in a year when the models are good, I can say, can you just calculate my taxes for me? Which is something, that I hear is, which is just something so stressful. Because if you get it from, you go to jail. Right? Like so your job right now is just to get familiar.
Sunil Pai [00:30:08]:
Pay $20 a month for chat GBP or Claude or one of these things. Talk to people in the space. Find out, like, what they're doing. You will get out of the 10 things you attempt, nine of them will be useless, but one of them will be useful. But you can't know which one of those 10 will be useful unless, like, you all try all 10. I I highly recommend people just get their hands in there right now. Do it for low stakes things. Don't run, like, your bank account on it.
Sunil Pai [00:30:36]:
Do it, just to automate some boring stuff in your life. You will be in the 1% of people who are ahead of the game when the opportunity presents itself. We see this time and time again without customers. Same thing.
Jordan Wilson [00:30:49]:
Fantastic, parting words. And thank you so much, Sunil, for taking time out of your day to join the Everyday AI Show. We really appreciate it. Thank you, Jordan. So grateful to be here. Appreciate it. Alright, y'all. And if you miss anything, don't worry.
Jordan Wilson [00:31:03]:
We're gonna be recapping all of that in today's newsletter. So if you haven't already, please go to youreverydayai.com. Sign up for that free daily newsletter. Thanks for tuning in today. We'll see you back tomorrow and every day for more everyday AI. Thanks, y'all.