The artificial intelligence landscape is rapidly evolving, but progress isn’t simply about building more intelligent models. According to Michael Gerstenhaber, Vice President of Product at Google Cloud and head of Vertex AI, the future of AI development hinges on simultaneously addressing three critical frontiers: raw intelligence, speed, and cost-effective scalability. This framework offers a nuanced perspective on the challenges and opportunities facing organizations deploying AI at scale.
Gerstenhaber, who previously held a role at Anthropic, oversees Google’s developer platform for AI, Vertex AI. His position provides a unique vantage point into how companies are leveraging AI and the hurdles they face. He argues that a holistic approach, considering not just capability but also practical limitations, is essential for unlocking the full potential of agentic AI – systems capable of autonomous action.
“I see three boundaries,” Gerstenhaber explained. “Models like Gemini Pro are tuned for raw intelligence…Then there’s this other boundary with latency…And then there’s this last bucket, where somebody like Reddit or Meta wants to moderate the entire internet. They have large budgets, but they can’t take an enterprise risk on something if they don’t know how it scales. And for that, cost becomes very, very important.”
The Three Frontiers of AI Model Capability
Gerstenhaber’s framework identifies three distinct areas where AI models are currently being pushed to their limits. The first, raw intelligence, focuses on the model’s ability to perform complex tasks, such as code generation or data analysis. In scenarios where quality is paramount and time is less of a concern, developers can prioritize the most intelligent models, even if they require significant processing time. He cited code writing as an example, where developers prioritize the best possible outcome, even if it takes 45 minutes to generate.
The second frontier is response time, or latency. For applications requiring real-time interactions, such as customer support, speed is critical. A highly accurate response is useless if it arrives after the user has lost patience. In these cases, developers must balance intelligence with the need for quick turnaround times. As Gerstenhaber noted, “more intelligence no longer matters once that person gets bored and hangs up the phone.”
The third, and perhaps most overlooked, frontier is cost-effective scalability. Organizations like Reddit and Meta, tasked with moderating vast amounts of user-generated content, require models that can operate at massive scale without incurring prohibitive costs. This necessitates a trade-off between intelligence and affordability, ensuring that the model can handle unpredictable workloads without exceeding budgetary constraints. According to a report from AI Chief, this focus on cost is becoming increasingly important as AI deployments expand.
The Slow Adoption of Agentic Systems
Despite advancements in AI models, the widespread adoption of agentic systems – AI systems capable of performing tasks autonomously – has been slower than anticipated. Gerstenhaber attributes this to the relative immaturity of the underlying infrastructure. “This technology is basically two years old, and there’s still a lot of missing infrastructure,” he stated. Specifically, he highlighted the lack of established patterns for auditing agent actions and authorizing data access, critical components for ensuring responsible and secure AI deployments.
However, Gerstenhaber pointed to software engineering as an area where agentic systems are gaining traction, due to the existing development lifecycle and robust human-in-the-loop processes. “We have a dev environment in which it’s safe to break things, and then we promote from the dev environment to the test environment,” he explained. He emphasized the need to replicate these patterns in other professions to accelerate the adoption of agentic AI across various industries.
Google’s vertically integrated approach, encompassing everything from data centers and chips to models and APIs, positions the company uniquely to address these challenges. Gerstenhaber highlighted Google’s ability to control the entire AI stack, from infrastructure to interface, as a key competitive advantage. Recent updates to Google Cloud’s generative media models, including Gemini 2.5 Flash Image and Veo, demonstrate the company’s commitment to innovation in this space.
Looking Ahead
The three frontiers outlined by Gerstenhaber – intelligence, speed, and cost – provide a valuable framework for understanding the complexities of AI development. As organizations continue to explore the potential of AI, a balanced approach that considers all three factors will be crucial for success. The ongoing development of infrastructure and standardized patterns for auditing and authorization will be key to unlocking the full potential of agentic systems and driving broader adoption.
What are your thoughts on the challenges and opportunities facing AI development? Share your insights in the comments below.
