AI Progress and Impact on Ecosystem Players with CapitalG’s Jill Chase | AI Basics with Google Cloud

Key Concepts

• AI investment landscape: chips, language models, services, and AI-powered startup scaling.
• Platform shifts: Internet, mobile, cloud, and AI.
• Investment categories: base layer (foundation models), infrastructure layer, and AI applications.
• "On the business" vs. "in the business" AI application.
• Durability and differentiation in AI startups.
• Co-pilots vs. agents in AI applications.

Investment Thesis and Mental Model

Jill Chase from Capital G (Alphabet's independent growth fund) discusses their investment approach in the AI space. Capital G focuses on post-product market fit companies (around Series B) and provides long-term capital and support. Their investment strategy is thematic and thesis-driven, requiring constant re-evaluation (every six months) due to the rapid pace of AI development. The core question is: "Where will the value accrue?" as AI represents the next major platform shift.

Investment Categories in AI

Capital G categorizes AI investments into three main areas:

1. Base Layer (Foundation Models): Investing in companies like Anthropic, OpenAI, Cohere, and DeepSeek.
2. Infrastructure Layer: Investing in the infrastructure that makes these models accessible and usable by businesses and consumers. This layer is attractive because it doesn't require betting on a specific model or application, but rather on the overall growth of AI usage.
3. AI Applications: Investing in software companies that leverage AI to enhance their offerings or improve their internal operations. AI is becoming integral to all software companies.

The Messy Middle and Application-Level Opportunities

The period between the introduction of a technology and its widespread adoption is challenging for investors. It's difficult to predict which applications will become successful. While early smartphone applications were basic (calculator, flashlight), later applications like Uber, Coinbase, and Robinhood became massive successes.

Jill notes that AI is particularly challenging because the "model brain" takes over some of the value creation. Model companies can directly address consumer use cases, potentially disrupting companies building on top of their APIs.

Example: Coding Applications

Coding is an obvious use case for generative AI. Companies like Cursor have built successful businesses by leveraging off-the-shelf models and adding technical advantages (e.g., RAG) and compelling UIs. However, as foundation models improve, model providers may directly compete by building their own UIs (e.g., cloud code).

"On the Business" vs. "In the Business"

Jill introduces the concept of using AI "on the business" and "in the business":

• "On the business": Using AI to improve business operations, such as customer support, sales, and marketing. This can lead to increased efficiency and faster growth. Startups can now accomplish tasks previously requiring larger teams.
• "In the business": Integrating AI into the product itself to create new features and enhance the customer experience. This lowers the barrier to finding product-market fit, as startups can quickly test new ideas using off-the-shelf APIs.

Rapid Testing and Iteration

AI enables startups to run tests and iterate much faster. A test that previously took 50 hours can now be done in 2 hours, allowing startups to explore more ideas and find valuable features.

Example: Testing a Meditation App

Instead of spending days building a website and running ads to test interest in a meditation app, a startup can quickly create a landing page and use AI to generate content and target potential users.

The Shadow Side: Durability and Differentiation

While AI democratizes the ability to build and test products, it also increases competition. Startups need to focus on building durable differentiation to avoid being easily copied by competitors or incumbents.

Strategies for Building Durability

Jill highlights two examples from her portfolio:

1. Building Something Hard: Motif, a next-generation Autodesk, is building complex 3D CAD software in the browser. The technical difficulty of the project creates a barrier to entry.
2. Landing and Expanding: Abridge, a healthcare AI company, starts with an automated scribe product. While the core technology may become commoditized, Abridge can leverage its relationships with healthcare organizations to build additional AI-powered products and services.

The Importance of Trust and Focus

Building trust and a strong brand are crucial for long-term success. Even if a product can be easily copied, customers may be reluctant to switch if they trust the original provider. Founders need to stay focused on innovation and customer satisfaction.

Example: Google Search

Google's obsession with improving search speed (even by milliseconds) led to increased usage and customer loyalty.

Co-pilots vs. Agents

The discussion touches on the transition from co-pilots to agents in AI applications. While fully autonomous agents may not be feasible in the near future, the trend is towards gradually increasing the autonomy of AI systems. Co-pilots can assist humans with specific tasks, and as models improve, these co-pilots can evolve into fully autonomous agents.

Conclusion

The AI landscape presents both tremendous opportunities and challenges for startups. By focusing on building durable differentiation, delivering exceptional customer experiences, and staying focused on innovation, founders can create successful and sustainable businesses in the age of AI. The ability to rapidly test and iterate, combined with the power of AI, makes this an exciting time to be a founder.