Insights from Emerging AI 100

Morning VC Brief is an expert daily curation of tech and venture news. Feel free to forward; the sign-up is here

As 2025 unfolds, with the breakthrough DeepSeek technology that reduces training costs while delivering outstanding performance, we’re keen to explore the latest discussions and focal points among VCs.

Here are 4 investment topics that are gaining traction, offering a glimpse into where the market is headed next.

2025, the Year of AI Agents

In 2025, we’re likely to see more successful commercialization cases for AI agent startups. According to Sierra AI founder Bret Taylor, the most commercially viable AI agents will be company agents, which serve as a new interface for customers to interact with businesses. These agents can handle everything a website does, make them the most applicable and scalable solution at this stage.

Taylor categorizes AI agents into three types:

1. Personal Agents: These act as personal assistants, handling errands for individuals. Due to their broad applications, they are seen as the most challenging to develop.

2. Persona-based Agents: Specialized in specific sectors, these are more achievable due to their focus on niche areas.

3. Company Agents: The most viable, these agents can take over tasks on business websites, creating seamless, scalable solutions for customer interaction.

AI agents are now moving towards outcome-based pricing models, moving away from traditional seat-based subscriptions. Innovators believe customers care more about ROI, and charging based on outcomes makes this clearer. Companies like Sierra AI and Crescendo (AI customer support incubated by General Catalyst) are already adopting this model, only charging based on customer satisfaction metrics.

This shift signals confidence from AI startups, who believe their agents can deliver results comparable to human workers. Sam Altman predicts AI agents will be a key breakthrough in AI foundational models and that the first batch will join the workforce by 2025, revolutionizing productivity.

As more AI agents enter the workforce, managing them will require new strategies. Agentic workforce management - tracking interactions, progress, and system records - will become a critical area to address, according to South Park Commons.

AI-driven innovations are already leading in areas like marketing, sales, and customer support. The customer support market alone is expected to grow from $460 billion in 2022 to $741 billion by 2030, with a 6% CAGR. Despite slow digitization in customer support (only 12%~15% of solutions are cloud-based compared to over 50% in CRM), the sector remains a heavy cost center. By leveraging AI agents, companies could reduce customer support costs significantly. Sierra AI’s CEO calculates that while human support costs $13 per call, AI agent support can bring it down to just $1 per interaction.

Vertical AI: Evolution from Vertical SaaS

Bessemer Venture Partners estimates the market opportunity for vertical AI is 10 times larger than vertical SaaS, with a potential value exceeding $3 trillion. In contrast, the combined market capitalization of the top 20 vertical SaaS companies is about $300 billion. Bessemer is confident in vertical AI’s future, predicting that the first vertical AI IPO could happen within the next 3 years.

Vertical SaaS has already shown success, with Toast and ServiceTitan leading the way. Founded in 2012, both companies took over a decade to capture market share — Toast with 10% and ServiceTitan with 1%. Toast is valued at over $20 billion, while ServiceTitan stands at nearly $9 billion.

In traditionally slow-moving industries like restaurants and home services, the digital shift has been gradual, presenting a significant opportunity for AI to drive change. In fact, while software only accounted for 1% of U.S. GDP in 2023, traditional business and professional services made up 13% — showing that the potential for AI-powered solutions in these sectors is 10 times greater than the current software market.

One example is Fieldguide, a startup using AI to enhance auditing across various sectors like information security, accounting, and payment security. By automating workflows, Fieldguide is increasing efficiency in traditional auditing, and its Series B funding round, led by Bessemer, is helping unlock the value of AI in this space.

Nina Achadjian from Index Ventures emphasizes that building AI agents from the ground up in specific industries helps demonstrate clearer ROI to customers.

This perspective is backed by Menlo Ventures survey, which found that enterprise customers prioritize ROI and industry-specific solutions over price when making purchasing decisions.

2025: The Next Big Consumer AI App

Jess Lee, partner at Sequoia Capital, predicts that consumer-facing AI will take center stage in 2025: “AI is great at chatting, making pictures, and videos, which has product-market fit with consumer. We will see brand new consumer social apps, novel forms of interactive media, chat based otome games, fresh takes on searching & synthesizing information, ambient memory capture, and fun messaging apps with generative UI.”

In past newsletters, we’ve hinted at new vertical AI search innovations coming in 2025. Sequoia also weighs in on the topic, calling AI search a likely “killing app” for consumers. They believe search will evolve into specialized, domain-specific AI tools. Instead of one-size-fits-all approach, AI search will be tailored for specific industries and professions. For example, Perplexity could become the go-to for investors, Harvey for lawyers, and OpenEvidence for doctors. As Sequoia notes, the future of AI search is all about domain expertise — helping professionals work smarter.

Further, South Park Commons investors highlight two key areas of interest for AI applications:

1. Content Consumption: Generative AI is reshaping how content is created and consumed, fundamentally changing the relationship between creators and audiences.

2. Content Discovery: With the rise of generative AI, content is becoming more abundant, but consumers face a new challenge — filtering and finding valuable information. As Forerunner Ventures notes, the next evolution of search will not just be about accessing information, but editing it, enabling a higher level of content curation and discovery.

AI for Science — Achieving Breakthroughs Sooner than Expected

Sam Altman’s AGI roadmap maps out the fourth level as “Innovators” — AI capable of aiding in invention, accelerating scientific breakthroughs. While Altman previously noted a significant gap between Level 3 agents and Level 4 innovations, he now believes this gap is closing much sooner than expected.

The application of AI in science has garnered significant attention from industry leaders. Vinod Khosla emphasizes AI’s potential to optimize nature resource usage and aid in discovering new materials, arguing that the real challenge is not a lack of resources, but our limited ability in finding new ones. Reid Hoffman, partner at Greylock and founder of LinkedIn and Inflection AI, is particularly excited about AI’s role in advancing genetics, medical discovery, and urgent disease research. Lux Capital shares this optimism, expecting “magical” breakthroughs in AI-driven healthcare, with biotech, chemistry, and physics all poised for acceleration through AI.

One example is Cusp AI, an AI-driven materials discovery startup that raised a $30 million Series Seed round in 2024, led by Lightspeed Ventures. Cusp leverages AI to recommend chemical combinations based on the desired characteristics of materials. Their primary use case is in carbon capture, where they aim to innovate new Direct Air Capture (DAC) materials using AI.