Making $$ with AI Agents
Greg Isenberg · 2026-04-29
💡 Quick Take
1. AI agents are poised to disrupt nearly every industry, with the potential market size being far greater than the estimated trillion dollars.
2. Software engineering is currently the most penetrated domain by AI agents, but even this is an underestimate of the true shift happening.
3. The current state of AI models, particularly with advancements like GPT-4.5, makes them capable of performing complex tasks autonomously, akin to human engineers.
4. The true power of AI agents is best understood through hands-on experience; superficial use won't reveal their full potential.
5. The cost of AI agents should be viewed in terms of value and human equivalent cost, not just token price, as they can drastically reduce time and opportunity costs.
6. Enterprise adoption of AI agents is experiencing the fastest growth curve in history.
7. Building a valuable business with AI can be achieved through Product-Led Growth (PLG) or a top-down, enterprise-sales approach.
8. The future of work involves companies operating with fleets of specialized agents that map to human job roles, rather than a single, all-powerful AI.
9. HyperAgent is designed as a user-friendly, "Mac" version of AI agent builders, focusing on intuitive UX and visual interaction.
10. HyperAgent goes beyond just app building; it acts as a "founder" by researching business context and then building informed applications.
11. HyperAgent offers powerful out-of-the-box tools and the ability for agents to learn and integrate with new APIs.
12. Skills are a crucial concept for AI agents, allowing them to be trained on specific tasks and domains, making them composable and interactive.
13. HyperAgent differentiates itself from competitors by offering a superior UX, more powerful built-in tools, and a focus on scalability and deployability into team settings.
14. Rubrics in HyperAgent enable automated evaluation of agent performance, ensuring consistent quality and allowing for cost optimization.
15. Building successful AI agents requires consistent effort and "coaching" rather than giving up after initial attempts; it's about investing time to optimize.
16. The AI revolution requires a fundamental reset in how businesses operate, similar to the early days of the internet and SEM.
17. Milestones like making the first internet dollar and reaching $10k/month are crucial for solopreneurs to build confidence and commit to AI-driven businesses.
18. Consistent daily practice with AI agent tools is key to becoming a top builder and achieving outsized returns through compounding.
19. HyperAgent can help users identify business opportunities by analyzing their personal context and suggesting relevant use cases.
20. HyperAgent aims to be the "iPhone" of agent builders, offering both extreme accessibility (low floor) and powerful scalability (high ceiling) with a strong focus on UX.
21. HyperAgent provides a visual, "desk-like" interface that appeals to users who prefer visual interaction over command-line interfaces.
22. HyperAgent offers a generous credit giveaway for early adopters to encourage adoption and make frontier models more accessible.
23. Solopreneurs and small businesses are ideal for rapid AI innovation due to their agility and willingness to deploy agents extensively.
📊 Detailed Explanation
1. AI agents are poised to disrupt nearly every industry, with the potential market size being far greater than the estimated trillion dollars. Howie Lou believes the market opportunity for AI agents is immense, significantly exceeding the commonly cited trillion-dollar figure. He suggests that even with current AI capabilities, deploying agents across various industries like software engineering, back office, marketing, sales, and CRM could achieve near-universal adoption. This implies that the TAM (Total Addressable Market) is not just a trillion dollars but potentially the entire GDP of white-collar labor, which is in the tens of trillions. This is because AI agents can automate and augment tasks that were previously exclusively human-domain.
2. Software engineering is currently the most penetrated domain by AI agents, but even this is an underestimate of the true shift happening. The transcript highlights a Sequoia chart showing software engineering at almost 50% deployment for AI agents. However, Howie argues this is an overestimate because the *way* software is being developed has fundamentally changed. It's no longer just about AI autocompletion (like GitHub Copilot) but a paradigm shift where the IDE might not even be needed. Developers are using autonomous agents with multiple cloud code instances, collaborating with other agents on PRs. This "frontier mode" of development, where AI is the primary driver and humans are reviewers, is still adopted by a minority, meaning the true disruption is deeper than the current penetration numbers suggest.
3. The current state of AI models, particularly with advancements like GPT-4.5, makes them capable of performing complex tasks autonomously, akin to human engineers. Howie emphasizes that recent breakthroughs, like GPT-4.5, have reached a "high water mark" where AI models truly feel like capable software engineers. They can autonomously tackle tasks that would take humans hours or days and deliver clean, production-ready code. This intelligence isn't limited to coding; these models can handle complex subject matter, provide expert-level answers in fields like management consulting, and coherently execute multi-term tasks with various tools and context. This intelligence is the foundation for deploying agents across all domains.
4. The true power of AI agents is best understood through hands-on experience; superficial use won't reveal their full potential. A key insight is that "using is believing." Many people are not fully grasping the power of frontier AI agents because they are experimenting superficially. Asking simple chatbot-like questions doesn't unlock the autonomy and capability of these agents. To truly understand, one needs to spend significant time (like a full weekend) with ambitious prompts and tasks, experiencing how these agents can structurally enable multi-billion revenue businesses with minimal human input. This hands-on engagement is crucial for extrapolating the types of companies that can now be built.
5. The cost of AI agents should be viewed in terms of value and human equivalent cost, not just token price, as they can drastically reduce time and opportunity costs. There's a common misconception about the cost of AI models. While frontier models can be expensive per token, Howie stresses reframing this cost. Instead of comparing it to traditional software subscriptions, one should consider the human equivalent time cost. For instance, generating a board memo that would take significant human effort can be done by an agent for a fraction of the cost in terms of time and potentially even monetary expense. The opportunity cost of a human's time is often far greater than the token cost of an agent performing the task. This reframe is essential for understanding the economic viability of agent-based businesses.
6. Enterprise adoption of AI agents is experiencing the fastest growth curve in history. The chart showing the percentage of enterprise apps with embedded AI agents reflects an incredibly rapid adoption rate. Howie isn't surprised, noting that this pace is also constrained by the ability of incumbents to integrate AI. The true indicator of this profound growth is the aggregate revenue of AI companies like OpenAI and Anthropic, which has surged from nearly zero to tens of billions in just a few years. This suggests a category-defining shift, with AI revenue growth outpacing historical software adoption curves.
7. Building a valuable business with AI can be achieved through Product-Led Growth (PLG) or a top-down, enterprise-sales approach. Two primary angles for building AI businesses are discussed. The first is PLG, where products like OpenAI's ChatGPT gain massive adoption through ease of use and accessibility, leading to significant token consumption. The second is a top-down approach, where companies pitch to enterprise boards and CEOs, offering to solve their AI problems for substantial fees (e.g., $100 million+). This is driven by existential risk mitigation for CEOs; not investing in AI is a guaranteed path to being fired, making the investment a strategic imperative, regardless of the ultimate transformation achieved.
8. The future of work involves companies operating with fleets of specialized agents that map to human job roles, rather than a single, all-powerful AI. The vision for the future is not a single omnipotent AI but rather a "fleet of agents." These agents will be specialized and map to intuitive human job roles (e.g., content marketer, market researcher). This partitioning is necessary due to fundamental limitations like context windows. Just as humans are specialized to manage complexity, agents will be too. This creates an "agent command center" where a company, potentially even a one-person company, manages numerous agents, each with a defined purpose and expertise.
9. HyperAgent is designed as a user-friendly, "Mac" version of AI agent builders, focusing on intuitive UX and visual interaction. Howie positions HyperAgent as the "Mac" to other agent builders' "Linux." The core philosophy is applying the same design principles and UX obsession that made Airtable successful to the agent space. This means making agent building intuitive, visual, and easy to grasp immediately, moving away from raw, technical interfaces. The goal is to democratize agent building, making it accessible to a wider audience beyond just highly technical users.
10. HyperAgent goes beyond just app building; it acts as a "founder" by researching business context and then building informed applications. A key differentiator for HyperAgent is its ability to act as a "founder." It doesn't just build an app; it first researches the business context, validates market needs (e.g., by finding user feedback on Reddit), analyzes competitors, and then builds an application informed by this comprehensive understanding. App building is seen as a commoditized feature; the real value is in the agent's ability to perform end-to-end business research and strategy before development.
11. HyperAgent offers powerful out-of-the-box tools and the ability for agents to learn and integrate with new APIs. HyperAgent comes equipped with powerful tools like Google Maps and AI image generation. More importantly, it allows agents to learn and integrate with new APIs. If a connector doesn't exist, an agent can research API documentation and build a custom skill to interact with it, requiring only API credentials. This empowers agents to perform virtually any task by giving them access to the right context and the ability to learn new capabilities.
12. Skills are a crucial concept for AI agents, allowing them to be trained on specific tasks and domains, making them composable and interactive. Skills are presented as the most important primitive in frontier agents. They are how agents are trained on specific domains or tasks. Skills are composable and can be interactively created. For example, an agent can be tasked with creating a "Greg Eisenberg AI content" skill by researching Greg's style, preferred platforms, and content topics. This skill can then be pinned to an agent or used on demand, effectively creating a specialized "avatar" for a specific persona or function.
13. HyperAgent differentiates itself from competitors by offering a superior UX, more powerful built-in tools, and a focus on scalability and deployability into team settings. Compared to competitors like Perplexity Computer and Manis, HyperAgent emphasizes a more visual and interactive UX. It also boasts more powerful out-of-the-box tools. Crucially, HyperAgent is designed from the ground up for scalability and deployability. This includes features like a "command center" for overseeing a fleet of agents, the ability to deploy agents into Slack channels for team collaboration, and automatic self-improvement loops where agents suggest new skills or prompt tweaks based on their performance.
14. Rubrics in HyperAgent enable automated evaluation of agent performance, ensuring consistent quality and allowing for cost optimization. Rubrics are an evaluation tool within HyperAgent that define what "good" looks like for a specific task. By pinning a rubric to an agent, users can have its output automatically scored by a separate LLM. This provides observability into agent performance over time, allowing for identification of areas for improvement. It also enables cost optimization, as users can see if reducing model quality (e.g., from Opus to Sonnet) impacts the score significantly, allowing for a five-fold cost reduction without a major dip in quality.
15. Building successful AI agents requires consistent effort and "coaching" rather than giving up after initial attempts; it's about investing time to optimize. The key to unlocking the full potential of AI agents is not a one-shot effort but sustained investment in "coaching" and curation. Users need to treat agent development like learning a skill, going through a "messy middle" to achieve great outputs. The arbitrage opportunity lies with those who are willing to invest this time and effort, as 99% of people don't. This consistent practice and optimization are what lead to significant leverage and valuable, always-on digital employees.
16. The AI revolution requires a fundamental reset in how businesses operate, similar to the early days of the internet and SEM. The analogy of two friends selling knives in the early 2000s illustrates the need for a radical business reset with AI. One friend supplements their door-to-door sales with Google AdWords, while the other fully commits to building an internet business. The latter, despite initial zero revenue, eventually builds a multi-billion dollar e-commerce empire. This highlights that what might seem like experimentation (investing time in AI) can be the most profound way to create long-term business leverage, even within a six-month timeframe.
17. Milestones like making the first internet dollar and reaching $10k/month are crucial for solopreneurs to build confidence and commit to AI-driven businesses. Confidence is a game-changer for solopreneurs. Making the first dollar from an AI-driven idea "rewires the brain." Reaching $10k/month is often a tipping point where individuals quit their jobs and go all-in, recognizing the tangible potential of their venture. These milestones provide the validation needed to fully commit to building a business around AI agents.
18. Consistent daily practice with AI agent tools is key to becoming a top builder and achieving outsized returns through compounding. To become a top 1% AI agent builder, consistent daily practice is essential. Committing to a set amount of time each day (e.g., 30 minutes) makes using the product a part of one's workflow. This daily engagement leads to compounding returns, similar to how writers improve by writing pages daily. This habituation and continuous learning are what drive outsized results.
19. HyperAgent can help users identify business opportunities by analyzing their personal context and suggesting relevant use cases. HyperAgent's onboarding flow is designed to overcome the "blank slate" problem. By connecting to a user's personal data (Gmail, Slack, Notion, etc.), it can analyze their context and suggest relevant use cases. For example, if a user is a VC dealing with deal flow, HyperAgent can suggest an agent to summarize and research investment pitches, or even draft replies. This personalized approach helps users discover how AI can solve their specific problems.
20. HyperAgent aims to be the "iPhone" of agent builders, offering both extreme accessibility (low floor) and powerful scalability (high ceiling) with a strong focus on UX. The vision for HyperAgent is to be the dominant platform in the agent builder space, much like the iPhone is for smartphones. This means providing an exceptionally intuitive user experience (low floor) that allows anyone to get started easily, while also offering deep functionality and scalability (high ceiling) for serious business operations and enterprise use. This combination of accessibility and power, driven by a relentless focus on UX, is HyperAgent's core differentiator.
21. HyperAgent provides a visual, "desk-like" interface that appeals to users who prefer visual interaction over command-line interfaces. The visual design of HyperAgent is likened to a "desk." It presents a tangible, organized workspace where users can see their tools, notes, and projects laid out. This contrasts with purely command-line interfaces and appeals to users who think and work visually, making the complex process of agent building feel more intuitive and manageable.
22. HyperAgent offers a generous credit giveaway for early adopters to encourage adoption and make frontier models more accessible. Leveraging Airtable's financial strength, HyperAgent is providing significant credits to early adopters. This generosity is aimed at driving widespread adoption and making powerful frontier models, like Opus, more accessible and affordable. The goal is to subsidize usage and allow users to experiment and build without being immediately constrained by token costs, fostering a robust ecosystem.
23. Solopreneurs and small businesses are ideal for rapid AI innovation due to their agility and willingness to deploy agents extensively. Howie highlights that solopreneurs and small businesses are where AI innovation will happen fastest. Their agility allows them to pivot and deploy agents across their operations much quicker than large, incumbent companies. These small shops are already becoming highly sophisticated and adopting AI in game-changing ways, demonstrating the power of agents for agile entities.
🎯 Expert Opinion
The conversation with Howie Lou is a masterclass in understanding the current AI agent landscape and its future trajectory. From an expert perspective, several key themes stand out, reinforcing and expanding upon the points raised in the transcript:
The "Agent-First" Business Model is Not Just a Trend, It's a Paradigm Shift: What Howie describes with HyperAgent isn't just about automating tasks; it's about fundamentally rethinking business structure. The concept of a "fleet of agents" mapping to job roles is the logical evolution of the digital workforce. We're moving from augmenting human jobs to creating entirely new roles for AI agents. This will lead to unprecedented lean organizations, where a handful of humans can manage operations that previously required hundreds. The implication for entrepreneurship is immense: the barrier to entry for complex businesses is plummeting.
The UX/UI Barrier is the Real Bottleneck, Not AI Capability: Howie's emphasis on HyperAgent's UX is spot on. While the underlying AI models are becoming incredibly powerful, their true potential remains locked behind complex interfaces or technical hurdles. Platforms that democratize access through intuitive, visual design will win. This mirrors the early days of personal computing where Apple's GUI made computers accessible. The "Mac" experience for AI agents is crucial for mass adoption and will unlock innovation beyond the technically elite.
The "Messy Middle" of Agent Optimization is Where Value is Created: The analogy of learning tennis or writing is perfect. The current arbitrage opportunity in AI agents lies precisely in the willingness to invest time in the "messy middle" – the iterative process of training, refining, and optimizing agents. Most users will likely scratch the surface. Those who commit to this optimization will build defensible, high-value businesses. Rubrics and automated evaluation tools within platforms like HyperAgent are essential for managing this optimization at scale, moving beyond subjective human judgment to objective performance metrics.
The Economic Reframe is Critical for Business Viability: The discussion on cost is vital. Businesses that continue to anchor AI costs to traditional SaaS models will miss the boat. The real economic advantage comes from the drastic reduction in human labor costs, opportunity costs, and increased speed-to-market. Agents aren't just cheaper; they enable fundamentally different business models and profit margins. This economic shift will be the primary driver for enterprise adoption and the success of agent-first startups.
The Future of Work is Orchestration, Not Just Automation: While automation is a component, the true future lies in orchestration. Managing fleets of agents, ensuring they collaborate effectively, and overseeing their performance through tools like HyperAgent's command center and rubrics is the new frontier. This requires a shift in management thinking, from managing people to managing intelligent systems. The ability to build and deploy these orchestrated systems will define the next generation of successful companies.
Prediction: The "Agent Ecosystem" Will Fragment and Consolidate: Howie touches on this by comparing HyperAgent to other platforms. We'll see a fragmentation into specialized tools (e.g., pure coding agents, pure research agents) and then a consolidation around platforms that offer comprehensive, integrated solutions like HyperAgent. The key differentiator will be the ability to manage, deploy, and optimize agents at scale, with a strong emphasis on user experience and robust tooling for oversight and improvement. The "iPhone" analogy is apt – a platform that integrates hardware (AI models), software (agent building), and a robust app store (skills/tools) will dominate.
The Generosity Model for Adoption is Smart: The credit giveaway is a brilliant move. In a rapidly evolving space, getting users hands-on with powerful tools, especially frontier models, is the fastest way to drive adoption and identify killer use cases. By subsidizing these costs, HyperAgent is effectively investing in its future ecosystem and accelerating the learning curve for its user base. This aligns with a PLG strategy that prioritizes user value and organic growth.
Kanal: Greg Isenberg