How Enterprise is Leading AI
I’ve been on a leave of absence and have been keeping away from “work” as much as possible to get a perspective on life. Asking myself questions like what does Life 3.0 really mean for ordinary people? How will Superagency or Agentic AI impact our society at the most fundamental levels like food, family, or shelter? What can history tell us about how to navigate this momentous time and are our institutions up for the challenge?
There has been a lot to think about… so let’s start with where we left off.
In Search of the AI Application Layer
My last post summarized an interview with Databricks CEO, Ali Ghodsi. The standout point was that we're just at the beginning of the AI Era, still building the infrastructure. The real winners - the Googles and Facebooks of AI - that'll dominate the Application Layer, are yet to emerge. Today, we're seeing some promising productivity applications, but truly consumer-facing "Agentic AI" is still a bit elusive.
This focus on enterprise applications isn't surprising. Many innovators are zeroing in on the enterprise sector for a few key reasons:
Immediate ROI – new productivity tools typically drive cost savings
Corporate Procurement – big budgets and focus on overall cost more than price
Controlled Environment – defined processes, end user and integration requirements
Security and Compliance – tend to operate within a regulatory framework
Desire to Innovate – many are actively seeking cost advantages through innovation
Hierarchies – top-down and centralized decisions generally mean lock-in
For these reasons, most innovations initially target enterprise applications. We've seen this pattern before, from Lotus Notes to Microsoft Office, and even Google Search (which was first licensed to web portals like Yahoo). Moreover, the underlying infrastructure layer - including hardware, software, protocols, standards, and legacy systems - often represents a sunk cost for the same enterprises that innovators are trying to sell to. This dynamic creates an interesting situation: not all enterprises will rush to adopt new innovations. This hesitancy actually provides an opportunity for innovators to demonstrate success with early adopters and potentially control pricing later on as the technology gains wider acceptance.
From the perspective of a small innovative team with limited talent and runway, focusing on enterprise use cases offers several advantages. It provides clearly defined use cases and markets, the potential for a substantial payout from a small number of customers with lock-in potential, and the legitimacy that comes from partnering with trusted, regulated corporate entities.
However, many innovators are now opting for an even more efficient approach to market entry. This strategy, pioneered by companies like AWS, Github, and Stripe, targets individual contributors or small teams with constrained budgets as a pathway to eventual enterprise sales. The maturity of internet and mobile technologies has not only made this kind of diffusion possible but is actually accelerating its effectiveness. This approach allows startups to gain traction and prove their value at a smaller scale before tackling the complexities of enterprise sales.
The Diffusion-Layer: Rapidly Accelerating the Adoption of Technology
This software diffusion-layer allows productivity tools on the application layer to quickly embed themselves within influential teams or individuals. Value is demonstrated rapidly as these early adopters spread the word within their organizations, often leading to corporate-level adoption and even investment.
AI tools like ChatGPT or Perplexity have shown that this enabling diffusion-layer has truly come of age. Meanwhile services like Threads have demonstrated that it can be particularly powerful if you're an established platform provider launching new solutions into your own ecosystem. While ChatGPT's achievement of reaching 1 million users (a long-standing benchmark) in 5 days was widely celebrated, Threads hit that mark in just 1 hour.
Both Gemini and ChatGPT leveraged ecosystems to reach this benchmark in less than a week, providing a worthwhile comparison. Meanwhile, Threads has shown that diffusion once measured in weeks will soon be counted in hours. Gemini launched into Google's own ecosystem, integrating with Search while bundling with Workspace and Google One. ChatGPT, on the other hand, proved that the underlying diffusion-layer was robust enough to rapidly gain adoption without an established ecosystem. The key factor for each was value, and I'd argue that the vast majority of that value lies in productivity enhancement.
Following the rapid adoption of these consumer-facing productivity tools, a range of competitors quickly turned their attention to enterprise customers. This includes AI innovators like OpenAI and Anthropic, as well as tech giants such as Google and Microsoft. Their goal is to collaborate on new large-scale AI platforms, while refining their systems to become more Agentic in the near future.
So where does this leave the Application Layer? In short, it's predominantly in the enterprise sector, and it's likely to stay there for some time. The enterprise market provides the resources and complex challenges needed to drive AI innovation forward, making it the primary focus for AI development and application in the immediate future.
Consumer-facing AI services like ChatGPT, Gemini, Claude, and Perplexity will continue to provide significant value to users, especially for tasks relying on publicly available information. These tools are proving incredibly helpful for casual inquiries, copywriting, image generation, video production, and research. Integrating these services into work processes can significantly enhance productivity.
However, elevating these services to truly support or potentially replace top performers in specialized roles will require fine-tuning within controlled environments. These environments need to offer structured data and cooperative employees. The Application Layer is likely to focus on use cases that can deliver Agentic AI systems, capable of more complex, context-aware, and autonomous operations. This aligns with the current focus on the enterprise sector, where AI can be pushed to its limits in data-rich environments, potentially paving the way for more advanced systems that could eventually expand to broader consumer applications.
What is an Agentic AI?
Agentic AI is an advanced AI platform that orchestrates multiple AI agents to accomplish diverse tasks aimed at achieving specific goals. These goals, tasks, intermediate outcomes, and final outputs are all communicated in human language. For instance, a customer service agent could handle L1-3 tasks seamlessly, resolving issues or escalating them to human supervisors when necessary. Real-world examples of agentic AI in action include real-time automated trading systems and self-driving cars.
Agentic AI platforms are characterized by several distinctive features. Unlike narrow AI, they have a broad scope, capable of handling a wide range of tasks and scenarios. They are goal-oriented, driven by specific objectives rather than predefined tasks. These systems exhibit high autonomy, operating with minimal human intervention, and can learn and adapt their strategies to achieve their goals.
A critical aspect of agentic AI is its alignment with human-desired outcomes or goals. This alignment requirement is a primary source of concern, as it can be challenging to achieve. To function effectively, an agentic AI platform requires a robust environment with an established library of tools and clearly defined capabilities for well-known tasks. Human operators must provide an exceptional definition of intended goals and verify that there is indeed an aligned understanding between themselves and the AI system.
Once deployed, the system must operate in an information-rich environment where automation is needed. Crucially, it must be able to communicate its plans, intentions, and progress to humans honestly and effectively. This level of transparency and interaction is essential for maintaining control and ensuring that the agentic AI continues to operate within its intended parameters while pursuing its goals.
Once an Agentic AI system has analyzed and planned its approach to achieve its goal, it must execute its plan and engage in self-learning to ensure success in both current and future instances. This process of execution and adaptation is crucial for the system's ongoing effectiveness and improvement.
While this level of AI capability might seem like the stuff of science fiction, we are actually on the cusp of achieving these types of systems for human institutions. Innovators with the most advanced AI infrastructure are eager to progress with Agentic AI and are actively seeking out partners to help realize this potential. As these systems continue to evolve, they promise to revolutionize how organizations operate and solve complex problems, potentially reshaping entire industries in the process.
Agentic AI: Proactive, Broad Scope, Goal-Oriented platform for Well-defined Goals
Discovering the AI Application Layer
The emerging Application Layer for Agentic AI is likely to be dominated by industries that have already undergone significant digital transformation and possess the necessary infrastructure to support advanced AI systems. While there's a clear need for AI solutions across various sectors, the initial adoption of Agentic AI will be concentrated in industries that meet specific criteria. Prime candidates for early Agentic AI adoption include telecommunications, healthcare, gaming, and Over-the-Top (OTT) services.
Telecommunications companies, with their vast customer bases and extensive digital infrastructure, are well-positioned to leverage Agentic AI for network optimization, customer service, and personalized offerings. The healthcare sector, as an early adopter of AI, is poised to benefit significantly in areas such as personalized treatment plans, operational efficiency, and enhanced patient care. The gaming industry's digital-native nature makes it an ideal candidate for Agentic AI implementation, particularly in creating more dynamic and personalized gaming experiences. OTT services like DoorDash, Netflix, Coursera, and Ally Bank, which already operate at scale and have a digital-first approach, are well-positioned to integrate Agentic AI into their platforms.
While industries like Education and Agriculture have significant potential for AI integration, they face unique challenges. These sectors often lack the structured, centralized data necessary for effective Agentic AI implementation, have complex and diverse value chains making uniform AI adoption challenging, and many organizations are still in the early stages of digital transformation.
Key factors driving Agentic AI adoption include robust digital infrastructure, access to large volumes of customer data, executive buy-in for AI initiatives, and operational scale. Companies with these elements in place are better equipped to integrate and leverage Agentic AI effectively.
As Agentic AI continues to evolve, we can expect to see its application expand beyond these initial sectors. However, the industries mentioned above are likely to lead the way in harnessing the full potential of this transformative technology, setting the stage for broader adoption across other sectors in the future.
Enterprise Scale: Offer both Structure Data and Established Workflows
These initial use cases for Agentic AI in the Application Layer, while not directly consumer-facing, will significantly impact consumer experiences. When implemented effectively, they should enhance customer interactions in several key ways:
Improved customer service through more efficient and personalized support
Highly tailored content recommendations based on individual preferences and behaviors
Lower prices for consumers while maintaining or even improving service quality
This development appears to be advantageous for incumbent companies with established customer bases and robust data infrastructure. However, it also presents opportunities for disruptors and visionaries in the tech space.
Innovative startups and forward-thinking companies can potentially set a new standard for service quality by leveraging unique datasets or novel ways of presenting information that are challenging to replicate. We're already seeing examples of this in productivity applications like Cartwheel, which offers a fresh approach to animation, and Perplexity, which is potentially disrupting the traditional search engine model.
These disruptors demonstrate that while established players may have advantages in implementing Agentic AI, there's still ample room for innovation and differentiation in the market. The key to success for these newcomers lies in their ability to offer unique value propositions that leverage AI in ways that larger, more established companies might not have considered or be agile enough to implement quickly.
Looking Ahead
As we've explored, the AI Application Layer is currently flowering in enterprise environments before it will fully blossom in the consumer space. This mirrors historical technology adoption patterns but with dramatically accelerated timeframes, thanks to the maturity of our digital infrastructure.
The emerging winners in this space will likely be those who can effectively harness data at scale, deliver meaningful customer experiences, and navigate the complex integration challenges that come with building truly Agentic AI systems. For incumbents with digital transformation already underway, this represents both enormous opportunity and existential threat. For disruptors, narrow windows exist to create compelling value propositions that established players cannot easily replicate.
What does this mean for society more broadly? The questions I asked myself during my leave of absence remain relevant. Life 3.0, Superagency, and fundamental societal impacts are still unfolding. But perhaps the answer lies in observing how industries like Telecom, Healthcare, and Gaming are implementing these technologies now - their successes and failures will inform the broader adoption patterns.
As consumers, our first meaningful interactions with Agentic AI will likely come through the services we already use daily, gradually becoming more capable and personalized. The true revolution may not be as visible as ChatGPT's explosive growth, but rather in the quiet transformation of experiences we've come to take for granted.
The infrastructure layer continues to develop, and the application layer is finding its footing in enterprise use cases. The next chapter in this story will be how these enterprise applications evolve to reshape consumer experiences in ways we've only begun to imagine.
So while I return from my leave with more questions than answers, one thing is certain: we are witnessing the foundation-building of something truly transformative. The AI Era is just beginning.