Google’s Cloud AI leads on the three frontiers of model capability

The Three Frontiers of Model Capability: Insights from Google's Cloud AI

As the field of artificial intelligence continues to evolve at breakneck speed, it's easy to get caught up in the hype surrounding the latest advancements. However, beneath the surface of flashy demos and impressive tech specs lies a more nuanced reality. According to Michael Gerstenhaber, product VP at Google Cloud, the development of agentic AI systems is facing a unique set of challenges that are holding back their widespread adoption.

The Three Boundaries of Model Capability

Gerstenhaber identifies three key boundaries that models like Gemini Pro are pushing against: raw intelligence, response time, and cost. These boundaries are not mutually exclusive, and the optimal solution will depend on the specific use case.

Raw Intelligence

In some cases, the primary goal is to achieve the highest level of raw intelligence possible. This is often the case in software development, where the best code is the most important factor. Gerstenhaber notes that in these situations, the model's intelligence is the primary concern, and latency is secondary.

Response Time

However, in other cases, response time becomes a critical factor. For example, in customer support, the answer must be provided quickly, or the customer will become frustrated and hang up the phone. In these situations, the model's intelligence is still important, but the response time is the primary constraint.

Cost

Finally, there is the issue of cost. In some cases, the model's cost becomes a significant factor, particularly when dealing with large-scale applications. Gerstenhaber notes that in these situations, the model's intelligence is still important, but the cost of deployment is the primary concern.

The Missing Infrastructure

Despite the progress made in developing agentic AI systems, Gerstenhaber notes that there is still a significant amount of missing infrastructure. This includes patterns for auditing what the agents are doing, authorization of data to an agent, and other essential components.

The Software Development Lifecycle

Gerstenhaber notes that the software development lifecycle is a key factor in the adoption of agentic AI systems. The process of writing code at Google requires two people to audit that code and both affirm that it's good enough to put Google's brand behind and give to customers. This human-in-the-loop process makes the implementation exceptionally low-risk.

Practical Implications

The insights provided by Gerstenhaber have significant practical implications for the development and deployment of agentic AI systems. By understanding the three boundaries of model capability, developers can design systems that are optimized for their specific use case. Additionally, the missing infrastructure and software development lifecycle are key areas of focus for the development of agentic AI systems.

Forward-Looking Thoughts

As the field of artificial intelligence continues to evolve, it's clear that agentic AI systems will play a critical role in the future of technology. By understanding the challenges and opportunities presented by these systems, developers can create solutions that are optimized for their specific use case. The insights provided by Gerstenhaber offer a unique perspective on the development of agentic AI systems and highlight the importance of considering the three boundaries of model capability, missing infrastructure, and software development lifecycle.

Conclusion

The development of agentic AI systems is a complex and multifaceted challenge. By understanding the three boundaries of model capability, missing infrastructure, and software development lifecycle, developers can create solutions that are optimized for their specific use case. The insights provided by Gerstenhaber offer a unique perspective on the development of agentic AI systems and highlight the importance of considering these key factors. As the field of artificial intelligence continues to evolve, it's clear that agentic AI systems will play a critical role in the future of technology.

Source: https://techcrunch.com/2026/02/23/googles-cloud-ai-lead-on-the-three-frontiers-of-model-capability/