At HubSpot for Startups' annual Gen AI Summit in San Francisco, four prominent venture capital partners came together to discuss the evolving landscape of AI investing. The panel featured Jake Saper from Emergence Capital, Bhaskar Ghosh from 8VC, Jennifer Li from Andreessen Horowitz, and Alex Ren from Fellows Fund, with Charles Elkan (CEO of Thick.ai and former Global Head of Machine Learning at Goldman Sachs) moderating the discussion.

Each Firm's AI Investment Focus

The panelists outlined their distinct approaches to AI investing, revealing how different funds are positioning themselves in this rapidly evolving market.

Emergence Capital takes a B2B-focused approach, concentrating on how AI can augment workers rather than replace them. Jake Saper emphasized their interest in integrating AI into applications that help workers perform more effectively, citing Zoom's Sales IQ product as an example of AI-powered real-time coaching.

8VC casts a wider net with investments spanning from biotech to logistics, where AI appears as automation mechanisms. Bhaskar Ghosh noted their particular focus on infrastructure and platforms, including data and AI plumbing, middleware, and departmental applications accelerated by foundation models.

Andreessen Horowitz concentrates on early-stage B2B enterprise and infrastructure investments that enable developers to build intelligent workflows. Jennifer Li highlighted their broad approach across biotech, gaming, and consumer experiences, with investments ranging from vector databases to foundation models.

Fellows Fund brings a unique model with 25 fellows who are industry leaders at top tech companies. Alex Ren explained their strategy of seeking the best AI company from each vertical, including accounting, insurance, healthcare, and enterprise sectors.

Building Moats in the Age of APIs

With many companies building on readily available APIs like OpenAI, the question of defensibility becomes crucial for both founders and investors.

Key strategies for creating sustainable competitive advantages include:

• Comprehensive workflow solutions: Going beyond basic AI functionality to solve complete jobs-to-be-done
• Business outcomes data: Gathering proprietary data from actual workflow usage to create closed-loop improvements
• Domain-specific fine-tuning: Focusing on specialized applications rather than general solutions
• Enterprise integration: Building the "boring SaaS scaffolding" necessary for real business applications

Jake Saper emphasized using Clayton Christensen's "jobs to be done" framework, noting that if most of a job can be accomplished by going directly to ChatGPT, the startup likely lacks defensibility. The key is building comprehensive solutions that require much more than just the model.

Evolving Data Strategies for AI Startups

The panelists revealed a significant shift in thinking about proprietary data requirements for AI startups.

The Four-Year Evolution: Jake Saper noted how his perspective has changed dramatically, explaining that four years ago, he wouldn't have believed general models could pass the bar exam or obtain medical licenses. Today's general models are so capable that startups don't necessarily need extensive proprietary data upfront.

Practical approaches for startups include:

• Starting with workflow products that naturally generate proprietary data over time
• Pursuing creative business development deals with legacy companies for data access
• Focusing on purposeful, high-quality data rather than massive datasets
• Fine-tuning models for specific use cases rather than building from scratch

Jennifer Li provided a compelling example of a company focused solely on texture generation for game design and product photography, demonstrating how startups can succeed by being highly purposeful with their data strategy.

Regulation and Compliance: Opportunity or Obstacle?

The panelists largely viewed increasing regulation and compliance requirements as positive developments for the AI industry.

Why regulation benefits the ecosystem:

• Prevents potential "FTX moments" that could derail enterprise adoption
• Creates market demand for compliance solutions
• Establishes necessary guardrails for enterprise sales
• Addresses critical privacy and data governance concerns

Bhaskar Ghosh outlined specific B2B compliance challenges that represent opportunities for startups, including industrial privacy concerns, copyright compliance in model training, and data-as-a-service legal frameworks. Jake Saper warned that without proper guardrails, a major incident could cause enterprises to stop purchasing AI solutions entirely.

AI's Impact on Physical Industries

While much AI focus centers on knowledge work, the panelists discussed opportunities in physical sectors like construction, manufacturing, and logistics.

Current applications showing promise:

• Drone-based crop estimation and construction variance detection
• Computer vision for manufacturing assembly guidance
• Next-generation robotic process automation (RPA 3.0)
• Document processing for logistics and transactions

However, Alex Ren noted that language model innovations remain primarily text-focused, making immediate applications in physical industries more challenging compared to content generation and enterprise applications where startups are already profitable.

The Content Generation Productivity Debate

Addressing whether AI-powered content generation creates real productivity gains versus zero-sum competition, the panelists emphasized that AI enables people to accomplish tasks they previously couldn't perform.

Alex Ren shared a personal example of creating a beautiful poster in three minutes using AI tools, a task that would have previously required hiring a designer for several days. This democratization of capabilities—enabling non-experts to produce professional-quality work—represents genuine productivity enhancement rather than mere competitive redistribution.

What VCs Look for in AI Founders

The ideal founder profile for AI startups differs significantly between application and infrastructure companies.

Essential founder qualities:

• Speed: Exceptional rate of learning and execution in a fast-moving field
• Problem experience: Having personally felt the specific pain point being solved
• Domain expertise: Deep understanding of the target market and user workflows
• Product instinct: Ability to productize technology for adoption and integration

For application companies: Domain expertise trumps AI expertise, with founders who understand user personas and workflows having significant advantages.

For infrastructure companies: Deep technical AI background becomes essential, particularly for founders tackling foundation model operations and AI lifecycle management.

Alex Ren emphasized that while there are hundreds of thousands of AI engineers, only a few hundred people can truly build viral businesses, making product market fit and go-to-market execution the primary differentiators.

The panel concluded with the insight that in a world where anyone can build something with OpenAI over a weekend, founders solving problems they've personally experienced maintain an unfair advantage in an increasingly crowded market.

AI Disclaimer: The insights shared in this video or audio were initially distilled through advanced AI summarization technologies, with subsequent refinements made by the writer and our editorial team to ensure clarity and veracity.