Key Takeaways
- • Vertical AI represents a $100M+ ARR opportunity by accelerating labor automation: Unlike vertical SaaS, which competes for IT budgets, vertical AI directly targets labor-intensive professional services workflows—a market 10x larger than software, representing 13% of U.S. GDP across business and professional services sectors.
- • Tiffany Luck, NEA Partner since 2023, identifies two core moat-building mechanisms: workflow depth and proprietary data as the primary defensibility factors for vertical AI founders, moving beyond generic chatbot capabilities to deliver firm-specific outcomes and automation.
- • The software paradigm is shifting from "system of record" to "system of actions": NEA's framework highlights four layers—data centralization, payments/billing, marketplace ecosystems, and agentic AI systems—with the fourth layer unlocking automation of previously untouchable language-intensive tasks in law, medicine, accounting, and professional services.
- • Proven vertical AI companies demonstrate rapid scaling potential: Examples including Abridge (clinical documentation), DroneDeploy (construction), EvenUp (personal injury law), and Fieldguide (audit workflows) show early-stage playbooks are validating the market thesis for industry-specific AI applications.
- • Declining foundation model costs and GPU infrastructure investment lower barriers to entry: Intense competition among LLM providers and hundreds of billions in capex investment have reduced the marginal cost of machine intelligence, making it economically viable to build large businesses serving narrow, specialized verticals.
Summary
Tiffany Luck, a Partner at New Enterprise Associates (NEA) since 2023 focused on early-stage AI, APIs, and B2B SaaS, has articulated a compelling investment thesis around vertical AI as the next generation of software category winners. Unlike the previous wave of vertical SaaS companies (Procore at $11B valuation, Toast at $21B, ServiceTitan at $9B), vertical AI operates under a fundamentally different economic model and market opportunity.
The core insight NEA presents is that vertical AI competes for labor budgets rather than IT budgets. Business and professional services account for approximately 13% of U.S. GDP—roughly 10 times larger than the software market itself. This distinction is critical because vertical AI automates high-cost, language-intensive workflows that traditional software could never address: lawyers reviewing discovery documents, doctors writing clinical notes, and accountants preparing tax filings. Large language models have unlocked the ability to automate these complex, nuanced tasks at scale, creating an addressable market far larger than vertical SaaS ever accessed.
According to NEA's framework, the nature of enterprise software is undergoing a structural shift from "system of record" to "system of actions." The evolutionary layers now include: (1) automation and centralization of customer data, (2) payments and billing workflows, (3) ecosystem marketplaces, and (4) agentic AI systems that automate remaining labor-heavy processes. This fourth layer represents the new frontier, enabled by declining foundation model costs and massive GPU infrastructure investment across the industry.
Luck identifies two primary mechanisms for building defensible moats in vertical AI: workflow depth and proprietary data. Companies cannot succeed by simply deploying generic AI agents; they must embed deep, firm-specific intelligence and control domain-critical workflows to differentiate from commoditized foundation models. Early momentum is evident across companies like Abridge (transforming clinical documentation), EvenUp (automating personal injury law workflows), Fieldguide (reimagining audit processes), and DroneDeploy (construction operations). NEA's research predicts that leading vertical AI companies will achieve $100M+ annual recurring revenue within historic timeframes—substantially faster than traditional SaaS scaling patterns.
The timing for vertical AI company formation has arrived, according to NEA's thesis, as foundation model capabilities have matured into generative, agentic, and multi-modal functions suitable for specialized industry applications. Competition among LLM providers and hundreds of billions in capital expenditure supporting GPU infrastructure have reduced the marginal cost of machine intelligence, making it economically feasible to build large, profitable businesses serving narrow verticals—a shift that could produce the next generation of software titans.
The core insight NEA presents is that vertical AI competes for labor budgets rather than IT budgets. Business and professional services account for approximately 13% of U.S. GDP—roughly 10 times larger than the software market itself. This distinction is critical because vertical AI automates high-cost, language-intensive workflows that traditional software could never address: lawyers reviewing discovery documents, doctors writing clinical notes, and accountants preparing tax filings. Large language models have unlocked the ability to automate these complex, nuanced tasks at scale, creating an addressable market far larger than vertical SaaS ever accessed.
According to NEA's framework, the nature of enterprise software is undergoing a structural shift from "system of record" to "system of actions." The evolutionary layers now include: (1) automation and centralization of customer data, (2) payments and billing workflows, (3) ecosystem marketplaces, and (4) agentic AI systems that automate remaining labor-heavy processes. This fourth layer represents the new frontier, enabled by declining foundation model costs and massive GPU infrastructure investment across the industry.
Luck identifies two primary mechanisms for building defensible moats in vertical AI: workflow depth and proprietary data. Companies cannot succeed by simply deploying generic AI agents; they must embed deep, firm-specific intelligence and control domain-critical workflows to differentiate from commoditized foundation models. Early momentum is evident across companies like Abridge (transforming clinical documentation), EvenUp (automating personal injury law workflows), Fieldguide (reimagining audit processes), and DroneDeploy (construction operations). NEA's research predicts that leading vertical AI companies will achieve $100M+ annual recurring revenue within historic timeframes—substantially faster than traditional SaaS scaling patterns.
The timing for vertical AI company formation has arrived, according to NEA's thesis, as foundation model capabilities have matured into generative, agentic, and multi-modal functions suitable for specialized industry applications. Competition among LLM providers and hundreds of billions in capital expenditure supporting GPU infrastructure have reduced the marginal cost of machine intelligence, making it economically feasible to build large, profitable businesses serving narrow verticals—a shift that could produce the next generation of software titans.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.