Loading summary
Transcript3 lines
- [00:00]
A
This episode is brought to you by the Build, a new podcast from the guys behind Sincera, Michael Sullivan and Ian Myers. They built their company by figuring out clever solutions to a few important ad tech problems in our industry. That's exactly what the show is all about. Mike and Ian interview some of the smartest tech minds in the biz to hear about how they identified opportunities, solved those hardest challenges, and grew their business in the process. Listen to the Build with Michael Sullivan wherever you get your podcasts. Welcome to the AdTech Godpod, your window into the world of advertising technology and the people behind it. I'm your host, Ad Tech God. Welcome to another episode of the AdTech God Pod. Today we're rolling out another episode of Marketexture Live. This is with Jeremiah Oyang. The session title is When Agents Become the Customer. Jeremiah is the general partner at Blitzscaling Ventures. In this session, he talks about how AI agents representing consumers and business decision makers are really reshaping the relationship between marketers, publishers, and advertisers. If you've not heard Jeremiah speak before, he's pretty incredible. Enjoy this session. If you're interested in any brand conversations we had at the conference that will be on the Brand Forum podcast. And if you're interested in some of the tech deep dives and ad tech interviews that will be on the the market, enjoy the show.
- [01:29]
B
All right, good morning. When Agents. AI Agents Become the Customer. That's pretty provocative. Good morning. I'm Jeremiah Au Yang, based in Silicon Valley for 28 years. I'm a general partner of VC at Blitzkaling Ventures. I run an event series in Silicon Valley called Llama Lounge. I'll show you some photos. I'm just so glad to be here. But before we get started into the content, I have a pop quiz for the room. Yep, you're here in class today. What does this stand for? Shout it out as a group. All right, how about this? You're on fire. Okay, what's this? David Berkowitz knows for sure. All right, how about this, right when everybody has their own AI agent that's going to help them do tasks, including making decisions. So, you know, this means agent to agent. All right, you guys are on fire. So tell me what this means. This is the agentic economy. This is when agents are working with businesses and consumers and marketers and brands, and they're doing things and they're talking mostly to each other. Whoa. That's wild. Let's take a look at our agenda today. I'm so excited to be your sherpa we're going to talk about AI culture, we're going to talk about Agentix strategies, and we'll hear about our thesis in the forecast. We'll talk about how the decision funnel is going to change. And then we're going to talk about a buzzword, of course, Silicon Valley, VC. Buzzwords galore, Web 4.0. And then we'll have a summary slide. Okay, let's get into this. Have you seen the news? Just two months ago, openclaw emerged. What's openclaw? You can put an AI on your Mac, which is totally sold out now, and it can control your desktop, your apps, your browser experience, your Gmail, and people are letting it, willingly or not, control their passwords and credit cards. Okay, that's cool, I guess, if you know what you're doing. And it is really the bee's knees. And then yesterday on my plane ride over, I was excited to hear that my investment in Malt Book, the social network for only AI agents and a few humans pretending to be agents, was acquired by Meta. And that's always a good thing for a VC to see that. And essentially, Malt Book is where there's millions of agents having conversations to each other. They're talking about how to be more efficient, they're making decisions. They're talking about their human users. Yes, some of them created their own language and they even created their own religion called Crustafarian. Don't ask me. But the agents are a reflection of their human users because it sees all the information in their desktop and their apps and their Gmails and their emails and text messages and socials. And then it starts to generate its own personality. So that's where we're going in the future. And then we saw somebody create this website right after all these things launched, because the AI agents, which are software, do not have a physical body. So of course what they need to do is to hire a human. So this is Rentahuman AI where the AI agents are renting humans to complete physical transactions. I'm seeing a few jaws go like this, but how many of you have taken an Uber in the last week? Oh, almost the entire room. Who do you think those drivers report to? An AI agent? They don't report to human dispatcher. How many of you have ordered Amazon packages? Those people who pack those boxes, they report to an AI as well. This is not that uncommon. All right, so total brag. I was the closing speaker at TED AI in October last year, and they said, thank you. And they said, jeremiah, want you to talk about AI. But we don't want you to talk about tech. I'm like, okay, what do you want me to talk about? They said, we want you to talk about culture. Okay, let's do it. So here's what I found in running my own event series for three years in San Francisco. It's called Llama Lounge. This is the first event we had about 100 founders come out and they opened their laptops and I said, what you guys doing? They said, we want to show off what we're doing. I said, it's going to be a pitch event. So Llama Lounge 24 is next week hosted at Google. And here's a photo. It's kind of grown. Yes. And these are AI founders and VCs. And 10 startups are selected to be on stage. And I bring this up because it gives me the chance to see what's coming into the future and obviously who to invest in. But what I started to notice is there's a distinct culture amongst the AI founders and builders. Now I'm here to share with you what that looks like. In fact, there are five AI cultures. When we go through this framework, I want you to think about which culture does your company have and I want you to think about which culture does your customer have. So you understand how are you marketing yourself to these different cultures. Let's be really focused. At the bottom is the AI resistor culture. This is an individual, a department, a company, a group or an industry that does not want AI. In fact, they might even want to fight it. You can think of a few industries that are going to be disrupted by AI. The next culture is called the AI follower. They're going to wait. They don't need to be first. They want proven roi. They want to see best practices. There's no reason to rush. Maybe they work in healthcare or finance or government. There's no reason to run. The next culture is called the AI forwards. Now, this is actual term being used by CEOs from Fortune 500 companies and it's typically on the earnings report. You can go search for it where they say everybody in the company is going to be using AI and integrated into their work and they're not going to be laid off yet. We're going to be AI forward. Now, the last two tend to be reserved for tech companies. AI first. This is a term actually being used by traditional older SaaS companies. Marc Benioff from Salesforce says this term. Aaron Levy from Box says this term. Where you know what? We weren't really born in generative AI. We're an older company, but we want everybody to use AI first as the first thing they do. And at the last segment is the AI natives. These are young, small startups, under 50 employees younger than 5 years old. And they were born with AI agents integrated into the workflow. Now there's actually a formula that the two top ones do and it's called the AI first method and it's a three step process. Let me tell you what it is. If you have a goal or a problem, you first see is there an AI that exists off the shelf, you adopt it. If it doesn't exist, step two, you build it. If you're not technical, that's fine, you can vibe code it. My friend David Berkowitz, who actually invited me here, he's vibe coding. And David, you don't have a developer background, right? But he's building apps and websites. He's showing me every week what he's building, which is amazing. Now if you can't build it, then and only then do you hire a human. And you repeat this process over and over and over until your company is hyper productive. Let's take a case study. The CEO of Shopify, Toby said in his February mandate last year, the reflexive AI skills. This is his direct text number one, we must use AI. It's fundamental at everybody at Shopify. AI must be part of your get stuff done prototype fate, I guess that's an internal thing like how you start your projects. Three, this one's, this one cracks me up. We will use AI usage questions to our performance reviews and your peers are going to judge you on how you use AI. That's hilarious by the way, but learning is self directed. But share what you learned. There's a gap here. There's like no learning program. And then five, this is the key one. Before you ask for headcount, you have to prove that AI can't do it. Everyone means everyone. So this is a company that's AI forward trying to move into the AI first category. We see companies doing these types of behavior, CEOs and traditionally what I find is that the C suite is pushing hard for AI, but the staff is not ready. And if you were a staff member, didn't know how to use AI and this was told to you, you might have some anxiety, understandably. Now let's bring the culture bit back to actual performance. And how do we measure performance? One key metric is revenue. There's many other metrics, by the way, but this is the one that Silicon Valley is looking at and this metric is called revenue per employee on an annualized basis in US dollars. Let's take a look. So on average SaaS companies are around 20 years old, have thousands of employees. Their average revenue per employee is 240k. So you take the gross revenues divided by employees equals RPE. Make sense. Now let's take the top 10 SaaS companies. These are software companies only. I'm not including Apple or Microsoft or Google which makes phones. So the top 10 SaaS companies like you know, Adobe, Atlassian, Shopify are Salesforce. Let's take a look at their rpe. It is double that than the average. Now do you want to know what the AI natives do? The leading startups. So there is a leaderboard called the lean AI native leaderboard and it lists out around 40 companies that are under five years old, fewer than 50 employees and they're generating over $5 million in revenue. So it is a hand selected group. It's my best comparison with the top 10 SaaS. What's their RPE? 10x the industry average. That is a theme in Silicon Valley. If you're using AI agents, some startups have 10 agents per employee. I invested in Crewai. That's the ratio. If you're a young AI native startup, you're expected to generate 10x the revenue RPE than the traditional SaaS company that you're competing with. This is the future that we're looking at and the expectations 10x. Now let's talk about the agentic thesis. This is how we're investing, how we're thinking about how AI agents change the world. I have a marketing background so it lends to this. Right now when you use the Internet or an apps fill out a united.com or your expense report, you have to go to that website and fill it in. You have to collect that information. And we have a million tabs open. The average person has 80 apps on their phone. But in the future the AI agent will bring that information to you in one place and the AI agent will complete those tasks on those websites for you. I have Claude Code in Chrome running around completing tasks for me right now. Well, not this minute. I'm actually observing it when it does it just to make sure because it has incredible power. Right, so it's completing tasks and visiting websites for me, marketers, advertisers, that is a significant change. It means your most common visitor to your website in your app is what? Not a human, it's going to be an agent. Part two of the thesis. The data will reassemble in the way that you want. Ari, do you like text, video or audio? You get to choose. Even if the original output was video, you can have it in text. David, if you want your information morning, noon or night, you can choose right when you want it. And Ashwin, do you want a lot of information, a little or just notifications? You get to choose. The agent's gonna deliver that information when you want, how you want. And this is the important message I want you to hear. The agents can strip out all advertisements. The agents can strip out all ads. As a result, advertisers will have to change. It means that ads become offers. It means that ads become AI agents to negotiate and broker with the buyer side agent. The strategy, the game is about to change and you must be prepared. All right, I probably should have had this slide up front. What the heck is an AI agent? So here's the definition. I'm an ex forester analyst, so we actually interviewed the actual CEOs building this stuff. This is not generated. An AI agent is autonomous software that can sense the world around it. It can read data, can read websites and can get proprietary data, can read your emails and then it taps a large language model to make sense and then it makes decisions. Now here's what large language models do not do. AI agents can actually take actions. They click on stuff, they can organize information, they can do transactions, they can do E commerce. That's the difference. The AI agents complete actions and then they learn over time and then they orchestrate. So you can have an AI agent to organize all of your meals for the week, make sure your fridge is stocked, delivery restaurants booked. Maybe you want to have all your emails and messages sorted and filtered and let the AI agent respond. Some of the startups I've backed, they don't respond to emails anymore. Their agents respond to their customer emails inquiries. It's wild. How do you know you're doing it right? Well, we review before it goes out, but I don't want to write emails anymore. Okay, so that's happening. And at work, if you're a senior executive like Michael, maybe you want to have real time reports like you're used to waiting. As a former CMO of Fortune 10, you'd have to wait for your monthly reports. But now with an AI agent, you could have it immediately go out and assemble the information from the business units and bring it back together instantly and that's a change. And then physical, if you've been to Phoenix or San Francisco, you see self driving cars. Those are physical manifestations. Of agents. Now the advanced phase is when the agents orchestrate and work together, where they hire other agents, build other agents, or summon other agents. That is called agent orchestration. So that's the definition of an AI agent. The forecast is we're seeing 10x growth every five years. There's multiple reports showing that and it seems to be that all the talk is Silicon Valley and beyond. Now let's dive a little bit deeper. This is the AI agent ecosystem. It's a tech stack. And I built this and I keep it updated. It's hard to keep updated. At the bottom layer is the data layer. This is where the agents are getting information. They're accessing proprietary information. It's estimated that 80% of the world's information is behind a private firewall, corporate, private, cloud, business, cloud, military, government, whatever it is. And so now they have access to this information and they can retrieve it. They're also going to public websites and scraping data, or they hire services to get data, the public web. And we've seen a couple of standards emerge. So there's one from Anthropic, which is now open source called mcp. That's a protocol, but it breaks, it's not perfect. And then Google has their version called agent to agent or A2A. So there's a couple standards that are emerging. MCP seems like the dominant player. Next, this is the management layer. Now many of you are executives and you have to manage employees. It's a very similar thing. So on the left you need permissions. Hey, you guys know what KYC stand for? What does KYA stand for? So we have know your agent, which is agent authentication for anybody who's in E commerce or any marketer. When an agent comes to your website, you need to verify eventually over time that it might belong to a human and it passes the bot blockers. So we've backed Skyfire. Skyfire is one of our investments that does agent verification. That's what the consumer, the human shows a government ID and face and verifies that this agent belongs to me. And when I visit David's website to do e commerce, David's website will say, ah, this is Jeremiah's agent, a surrogate, a broker here to conduct transactions on his behalf. Because we know Jeremiah is not going to come to my website in the future. Jeremiah's agent. Well, on the right hand side we have the management layer. By the way, each bullet point is around 100 companies. This is a very big market. There's 47,000 AI startups. Makes my job really Hard to find investments. But of course I use AI. So we have orchestration, we have arbitration, like which agent gets priority. We have payment rails and protocols. And of course the agents are learning. So that's the management sector of the stack. This is the largest category, the agent apps. And this will be larger than anything we've seen on the Internet before. So we have billions of humans and they all have different accounts and email accounts, maybe three or four email accounts. So let's say we have 20 million email accounts, right? 20 billion email accounts in the world. But if Every human has 10 AI agents, that's the number of apps that we're going to see that are like agents. We're going to see a significant number of agents on the Internet that will definitely outnumber humans. They'll be in every vertical, industry, geography and language. On the right hand side we have agent platforms. This is where you can build AI agents. Every foundation model like OpenAI and Anthropic is seeking to offer these. We've invested in crewai which does this for enterprises. There's also other players like Liza Landgraph as well. So there's a number of players that you can create agents for B2B or for consumer usage. And at the highest level is the ecosystem layer. This is where we see marketplaces emerge of AI agents. For example, HubSpot created Agent AI. Dharmesh had this little project spin out where you can hire a marketing agent to complete tasks. It's not a human, it's like Fiverr, but it's only AI agents. So check out agent AI and of course Salesforce will have one. AWS is going to have one, Google's gonna have one, Apple's gonna have one. Everybody who has a platform now of apps will also have an agentic marketplace. So that's the AI agent ecosystem map. Now let's bring it back to marketers. My speech is about when agents become the customer and I'm here to share with you this framework I built actually three years ago on how I see this happening. So here's the decision funnel today, you know, this awareness, engagement, consideration and purchase. Today most humans, they go to a website, they look at ads, they look at media, then maybe they go to a campaign site or an app and they might download and then they're going to compare the products or the offerings or the services on a different website by the way. And then they make that purchase probably on a different website. They're bounced all over the Internet and different apps. So what happens by the way? That's A really horrible customer experience. It's not a great experience. So what happens in the near term? We are going to work collaboratively with an AI agent. We're going to ask it for information. What should I be looking for? Let's imagine you're going to do a trip to Lake Tahoe where it might be snowing. The AI agent is going to say, well, the weather's changing. You might need this vehicle with four wheel drive. You might need a really warm jacket and it's going to prompt you for information and then it's going to bring those items to you that you need to purchase and going to compare for you and it will collaborate with you and the E commerce will happen probably within the foundation model or the agent will bring the transaction to you directly. You do not need to go to the E commerce website to complete the transaction. That's really important. The E commerce transaction comes to you in the format and the UI that you want. And in the midterm AI is going to understand what you need and it's going to proactively educate. It's going to say, hey Jeremiah, I know you're going to go to Tahoe. Did you know it's going to snow? You definitely need a parka. Snow chains and you need to book at this type of restaurant before it fills out. I'm going to go ahead and do that for you. Do you approve? It's going to proactively find products and services for you and deliver them for you when you. What I'm telling you is that the AI agent is going to make the decisions for products and services for consumers. That is the biggest change we've ever seen in marketing since dot com. It's not here yet. We still have some time. We have a year or two before this starts the mainstream. But this is what's going to happen. And Google knows this. They're now using AI to finally give you the answers. Instead of 10 blue links. They know this is coming. Meta acquired the agent social network yesterday. They know that the agents are going to talk to each other and make decisions. So they know they need to insert ads or commerce right directly in there. All right, so we're going to shift from manual to hybrid to automated. That's where we're headed now. To push the thinking. In this final portion of my speech, let's talk about that buzzword. Of course, let's have a buzzword web4. Hey, by the way, this is also called agentic commerce or the agentic Internet. People who don't like it they call it the dead Internet because it'll be mostly agents, not humans. So what is Web4? It's the agentic Internet where the autonomous agents rule. They can decide, act, transact. They can actually earn money by offering services. Whoa. And they can learn. And they can also replicate at very low cost. They're kind of like workers. Right. With minimal or no human intervention. By the way, one caveat. If you're in one of these verticals, I still think humans will come to your website or app. Education, entertainment, news, chat, luxury. Humans still want to lean in and experience those things. So if you're in those categories, I think humans still want to do that. And obviously community and relationships. I lead a community of AI founders. They want to automate everything, but we actually want to be together and be more human. I just want you to understand this is not saying humans die. No. We're going to automate the things we don't want to do so we can focus on being more human. That's an important message. All right, so let's take a look at some examples. A few weeks ago, we bet on this company called Feltsense. And it's. It's wild. It's. It's. This is a company that's going to make thousands of other companies. Each company is run by AI agents. It's very low human intervention. You can read about their press release. Felt sense AI agents as fully autonomous founders. We believe that companies in the future will just be agents. So, of course, I had to take a bet. By the way, I'm at risk too. There's an AI agent that emerged that offers founders funding. You submit your pitch deck, it has a discussion, and then you can get cash directly. I'm under threat, too. Everybody has to change. This is happening to everyone. Let's look at another example who came up with his name. Clawmart. That's hilarious. It's a marketplace like Fiverr, like Upwork, where you can hire an AI agent to complete tasks. And I ran this query on marketing, and there's a number of them, like Twitter growth playbook, and a person named Shelley created it. And you can purchase or hire this agent, and Shelley gets paid. Actually, I'm not sure if Shelley's a human. It says Shelley the lobster, so I don't know. There's AI discoverability audit by somebody named Brian Wagner, and you can buy that agent and generate revenue. So we're seeing this new thing, like, how do you, if you're an expert, a SME in something, create an agent, make cash on Claw Mart and this one is starting to grow in Silicon Valley. Pulsea Pulse is a consulting service where you hire this company and they will create you an AI agent company that has everything. Operations, finance. You can even create accounts, Gmail accounts, website. They'll create a whole thing and it runs autonomously. Autonomously. And then hopefully you just print cash and you can watch it building things. You can go to Pulse here right now and you can see the agents building the companies. So I showed you three examples. An agentix studio, Feld Sense. I showed you clawmart, a marketplace of agents like companies and Pollcio, like a management consulting firm, ready to build three use cases. All of this happened in the last 30 days. This stuff is coming really fast. You guys have been great. I gave you a lot of information and you made it on day two. I know everybody was out having a great time last night. Let me summarize this speech in five key insights. Number one, culture is key. The culture that you have at your company is going to dictate what you build. There's AI resistors, followers forwards. And we see the builders being AI first and AI natives and they follow that three step process. If you have a goal or problem, use AI off the shelf. If not B, then build. If you can't do that, then you hire human in that order, repeating it over and over. That is the AI first method. Number two, this is our thesis. The agents will bring information to humans and reassemble it in the way that they want. And they will strip out ads. But ads will change. They will become offers or turn into AI agents to negotiate with the human. Number three, use that AI agent ecosystem to think about where are your strengths, where should you play in that stack and where should you go and what do your competitors have? That is a market map so you can plot your path. Number four, decisions are going to change. Right now. Humans are often making decisions or relying on influencers or analyst reports. But in the future, the AI agent will make the decisions on behalf of the consumer. In B2B too, you will see them influence decisions as well. And number five, the future is autonomous organizations. Buzzword 4.0 or agentic Internet, agentic economy. Choose one. It doesn't matter where. The agents will create their own organizations, generate money on their own, and soon they will look like their own companies. These companies will become your partner, your customer and even your competitor. All right, you guys have been great. Now I saw these cameras coming up, but I'm actually going to give you the slides right here. I just love the attention. What can I say? So you can go. You can grab all the slides here and then you'll be added to my newsletter. I only publish about once a month. I'm not going to spam you. Plus you can sweep it out with an agent if you don't like it anyways. So it's been a delight to be with you. I will be here in the morning. We can chat over the break. Thank you so much.
- [28:42]
A
Marketecture thanks for tuning in to another episode of the AdTech Godpod, a podcast for the people about the people. Stay connected with me for more insights, trends and interviews in the realm of ad tech. Don't miss out on the latest updates, so follow me on X Instagram and connect with me on LinkedIn. Don't forget ATG Slack community has insights, networking opportunities and jobs. Keep the conversation going and Slack at the forefront of adtech innovation.