Bessemer’s Talia Goldberg and Google Cloud’s COO Francis deSouza were kind enough to join us at SaaStr AI Summit 2025 on what’s driving the fastest technology adoption in decades.
The New Rules of AI Investing: Why Speed Beats Strategy and Labor Budget Is the New Software Budget
The pace of AI advancement has fundamentally broken our traditional frameworks for thinking about technology adoption, market timing, and competitive moats. What we thought might happen in 5-10 years is now happening in 12 months, creating entirely new rules for both investors and entrepreneurs.
The Speed Revolution: Execution Over Everything
“The number one thing that we look for in teams is speed of execution,” explains Talia Goldberg, Partner at Bessemer Venture Partners. “At the earliest stages, you really don’t have moats. You have teams you’re building, and how fast you can move and adapt to this rapidly changing market is so critical.”
This represents a fundamental shift in venture investing criteria. Traditional competitive advantages like network effects, switching costs, and proprietary data matter less in the early stages than pure execution velocity. The market is moving so fast that yesterday’s innovative feature becomes tomorrow’s table stakes, and only teams that can iterate at AI speed will survive.
Francis deSouza, COO of Google Cloud, reinforces this from the enterprise side: “In my three decades plus in tech, this is the fastest I’ve seen adoption of any technology ever. A year ago we were talking about pilots. A year later, we have Verizon, Seattle Children’s Hospital, and Toyota rolling out GenAI applications on factory floors, in hospitals, and call centers.”
From Software Budget to Labor Budget: The $10 Trillion Opportunity
The most profound shift happening in AI isn’t technical, it’s economic. We’re witnessing the migration from software budget to labor budget, where code is becoming labor itself.
“Markets used to tap into just software budget. We’re now seeing that change into labor budget,” Goldberg notes. This isn’t just about improving productivity; it’s about expanding the entire addressable market for technology companies.
Consider the implications: software spending represents 1-2% of most companies’ budgets, while labor can be 60-70%. When AI applications can genuinely replace or augment human work, not just automate software tasks, the market size explodes from billions to trillions.
This shift explains why healthcare, traditionally a laggard in SaaS adoption, is now leading AI implementation. Companies like Abridge in Bessemer’s portfolio are selling “tens of millions of dollars of contract value to large hospital systems” by automating medical scribing and back-office operations, directly replacing labor costs rather than just improving software efficiency.
The Paradox of Large Enterprise Leadership
Conventional wisdom suggests that technology adoption flows from agile startups to conservative enterprises. AI is flipping this script entirely.
“We’re seeing large companies lead the revolution as much as digital natives,” deSouza observes. “Typically, technology finds its footing first in smaller companies that move more quickly, but this wave is being adopted by large companies too.”
Why? The benefits are too significant and too immediate to ignore. When AI can deliver measurable labor cost savings within months rather than years, even the most risk-averse enterprises become early adopters. This creates an unusual dynamic where startups need to be enterprise-ready from day one, rather than gradually scaling up market.
The Multi-Model Future: Why Openness Wins
One of the most strategic insights from the discussion challenges the “winner-take-all” narrative that has dominated tech for the past decade. Both investors and platform providers are betting on a multi-model future.
“We don’t believe there’s going to be one model to rule them all,” deSouza explains. “That’s why in our Vertex AI platform we support over 30 models across different vendors—over 200 models total. You’re going to see models that show up that are specific for use cases.”
This openness isn’t altruistic; it’s strategic. Platform providers who lock developers into single models risk being displaced when better specialized models emerge. Startups building on closed platforms face existential risk when their foundation shifts. The winners will be those who architect for model diversity from the beginning.
Where Google Won’t Go: The Startup Opportunity Map
For entrepreneurs wondering where to build, deSouza provides remarkably clear guidance on Google’s strategic focus areas:
Where Google Invests:
- Infrastructure layer (TPUs, networking, planet-scale infrastructure)
- Core models (Gemini 2.5, image, voice, video models)
- Select horizontal applications (call centers, agent building, security)
Where Google Explicitly Won’t Go:
- Vertical market applications
- Specialized workflow tools
- Industry-specific solutions
- Next-generation SaaS replacements
“We rely on the ecosystem, all of you, because we believe a huge part of the enterprise IT market is going to be transformed by companies like you. That is a lot of ground to cover,” deSouza states.
This creates a clear roadmap for startup opportunities: focus on vertical specialization, industry-specific workflows, or novel application categories that major platforms can’t justify addressing due to market size or innovation timelines.
The New Venture Math: 60 People, $300M ARR
The leverage that AI provides to small teams is rewriting the fundamental economics of startups. Cursor, with just 60 people, is reportedly at $300 million in annual recurring revenue, a ratio that would have been impossible in previous technology waves.
“It is now possible to imagine companies that are less than 10 people that could be a hundred billion dollar company,” deSouza notes. “There’s just so much leverage for startups in leveraging AI themselves.”
This doesn’t mean the bar is lower; it means the ceiling is higher. Teams that can effectively leverage AI can achieve unprecedented scale with minimal resources, but they’re competing against others with the same leverage. The winners will be those who can most effectively multiply their capabilities through AI tools and automation.
The Data Goldmine: Conversational and Contextual Information
As AI agents become more sophisticated, the bottleneck shifts from processing power to data quality and context. Goldberg highlights “conversational data” as a critical frontier: “A lot of data happens like us having this conversation. You need that context for agents to work. It’s not just what’s sitting in your email or CRM, it’s what’s happening all around.”
Companies like Recall AI in Bessemer’s portfolio are building the infrastructure to capture and structure this previously inaccessible data from video calls, meetings, and conversations. This represents a new category of essential infrastructure for AI applications.
Strategic Implications for SaaS Leaders
For CEOs and Founders:
- Prioritize speed over perfection: In a market moving this fast, the biggest risk is moving too slowly, not building the wrong thing
- Think labor replacement, not software enhancement: The biggest opportunities lie in replacing human work, not just improving software workflows
- Build for model diversity: Architect your AI applications to work with multiple models rather than betting on a single provider
- Focus ruthlessly: Large platforms have infinite resources but can’t address every vertical, specialization is your competitive advantage
For Investors:
- Speed of execution becomes the primary diligence criterion: Team velocity matters more than initial product sophistication
- Labor budget TAM calculations: Evaluate opportunities based on labor replacement potential, not just software market size
- Platform risk assessment: Ensure portfolio companies aren’t overly dependent on single AI providers
- Category creation over competition: The biggest wins will come from companies creating entirely new categories rather than competing in existing ones
The Bottom Line
We’re in the early innings of the most significant technology transformation since the internet. The companies that will define the next decade are being built right now, by teams that understand these new rules: speed over strategy, labor over software, focus over breadth, and openness over control.
The question isn’t whether AI will transform your industry, it’s whether you’ll be the one doing the transforming or getting transformed. In a market moving at AI speed, there’s no middle ground.
Top 4 Unexpected Learnings
1. Healthcare Is Leading AI Adoption (Not Lagging) The biggest surprise? Healthcare. Historically, the slowest industry to adopt new technology is now among the fastest AI adopters. While healthcare companies were always behind on SaaS, they’re leading on AI because it directly addresses their biggest pain point: labor costs. When AI can replace expensive human tasks like medical scribing, even conservative hospital systems move fast.
2. VCs Are Hunting Down Startups (Not the Other Way Around) Forget cold emails and pitch decks. The best AI investments are happening when VCs are “banging down doors” to meet founders who are building something people desperately want. In a market moving this fast, great teams with obvious traction don’t need to hustle for investor attention—investors are hustling for them.
3. Google’s COO Uses AI to Plan His Daughter’s College Schedule. The most relatable AI use case came from Google’s COO, whose college-aged daughter used AI to map out her entire three-year academic plan, checking for conflicts and prerequisites across her double major. Sometimes the most mundane applications reveal the technology’s true power.
4. Personal Health AI Is the Ultimate Personalization Goldberg’s most compelling personal use case: aggregating a decade of medical records into an AI system that can answer health questions with “unbelievable detail and personalization.” As she notes, even her doctor parents couldn’t ingest and analyze 15 years of blood labs to spot trends. This level of personalization represents a category-defining opportunity waiting to be built.
