The Next Wave of Enterprise AI: From Horizontal Models to Vertical Solutions

We’re at an inflection point in enterprise AI. While much of the focus has been on the race between frontier model providers, the next transformation will come from AI systems that bridge the gap between state-of-the-art reasoning and physical world processes.

Our conversations with enterprise leaders and AI researchers consistently highlight this emerging opportunity, particularly in industries that have historically been resistant to technological change. Recent advances in diffusion models for robotics, for instance, are particularly exciting—these models can help robots learn complex physical manipulation tasks from demonstrations and adapt to new scenarios in real-time. While hardware limitations remain significant—from inadequate sensory feedback for distinguishing material textures to the inherent variability of physical systems compared to virtual ones–the combination of LLMs’ common sense and reasoning with traditional automation opens up possibilities for AI systems that can not only analyze and optimize processes but also begin to physically execute certain complex tasks.

This evolution mirrors the shift we saw in enterprise software, where horizontal cloud platforms gave way to vertical SaaS solutions built for specific industries. Now, as foundation models establish a new horizontal layer of AI capability, we’re witnessing the emergence of vertical AI solutions that combine industry expertise with AI to transform operations.

Identifying the Next Vertical Frontiers

Several key criteria signal domains ripe for disruption by end-to-end AI solutions:

1. Data-rich but Integration-poor Environments

• Industries with valuable data trapped in legacy enterprise systems
• Sectors with critical operational data that isn’t even being recorded yet
• Opportunities for novel instrumentation to capture previously invisible insights
• Environments where integrating disparate data sources creates powerful network effects

2. High Complexity Operations

• Domains where workflows are too complex for traditional automation or simple rule-based systems
• Use cases that require sophisticated orchestration of multiple systems and stakeholders
• Processes where complexity exceeds most enterprises’ internal capabilities to build and maintain solutions

3. Mission-critical Processes

• Operations with low tolerance for errors or system failures
• Environments where reliability and robustness are paramount
• Use cases where enterprises need best-in-class solutions rather than internal tools

4. Industry-standard Workflows

• Sectors where workflows and best practices are defined at an industry level
• Processes that follow standardized protocols rather than company-specific methods
• Domains where solution providers can build once and deploy many times

5. Strong Regulatory and Data Moats

• Industries requiring specific regulatory approvals or certifications
• Sectors where access to privileged third-party data creates competitive advantages
• Domains where compliance requirements create barriers to entry

Manufacturing, construction, and supply chain management stand out as prime examples where these criteria converge—featuring complex physical processes, mission-critical operations, legacy data systems, opportunities for novel instrumentation, and stringent regulatory requirements.

Why Vertical AI Will Win

Some argue that as frontier models become more capable, they will eventually be able to generate entire vertical applications from natural language prompts alone, eliminating the need for specialized solutions. While these models will certainly accelerate software development, the reality of building production-ready industry applications is more nuanced. Even if AI can generate robust code, implementing secure data integrations, maintaining compliance standards, and optimizing for specific industry constraints will continue to require dedicated teams with deep domain expertise. Moreover, the probability of failure increases geometrically across asynchronous agentic trajectories, making it impractical to rely entirely on end-to-end code generation for complex industry workflows. There will always be a need to connect to industry-specific data sources and handle domain-specific edge cases.

Just as the democratization of software development tools didn’t eliminate the need for vertical SaaS—despite making custom development cheaper by several orders of magnitude—the advent of powerful AI coding assistants won’t eliminate the need for specialized industry solutions. Instead, it will enable these solutions to be built faster and with smaller teams while maintaining their crucial role in complex industry workflows.

The biggest opportunities lie in vertically integrated solutions that combine:

• Novel instrumentation to capture previously unrecorded data
• Integration layers that unlock data trapped in decades-old enterprise systems
• AI systems that can reason about and optimize complex physical processes

Industry SaaS has been a huge success with highly valuable solutions like Veeva, Toast, and ServiceTitan. Just as vertical SaaS providers emerged to build industry-specific applications on top of cloud infrastructure, we’re now seeing vertical AI companies build specialized solutions on top of foundation models. The market opportunity for vertical AI will be even larger than vertical SaaS. These companies can not only replace legacy software but create entirely new capabilities by combining general intelligence with previously untapped data and domain expertise.

Joining Lightspeed to Drive This Transformation

This unprecedented opportunity is why I’m excited to join Lightspeed. Growing up in Silicon Valley, I developed a passion for innovation, entrepreneurship, and building things, spending many hours at Bay Area makerspaces. As an Enterprise Partner, I bring an academic perspective of AI from machine learning coursework and research at Cornell and industry experience in developing enterprise AI applications from C3 AI. There, I learned how AI can be applied to drive massive operational value and address challenges across manufacturing, supply chains, sustainability, healthcare, and beyond.

I joined Lightspeed because of our emphasis on deep technical experience and our track record of backing what we believe to be transformational AI companies such as Anthropic, Skild, Dexterity, Abridge, EvenUp, Glean, Mistral, and Databricks, among many others. I’m seeking to partner with founders who are building innovative end-to-end solutions that combine domain expertise, novel data sources, and efficient AI pipelines to drive enterprise value and societal benefit. If you’re working on ambitious AI products and building out a world-class team, let’s talk. I’d love to hear your story.

The content here should not be viewed as investment advice, nor does it constitute an offer to sell, or a solicitation of an offer to buy, any securities. Certain statements herein are the opinions and beliefs of Lightspeed; other market participants could take different views.