The Potential of Vertical AI (Law, Accounting, Architecture, Manufacturing, Healthcare, etc.)

"The next trillion-dollar company will be a software company disguised as a services company" — this prophecy from Bessemer Venture Partners cuts to the essence of Vertical AI. Vertical AI refers to AI solutions built specifically for particular industries such as law, healthcare, accounting, construction, and manufacturing. While general-purpose AI (Horizontal AI) like ChatGPT and Claude can "do anything but lacks depth," Vertical AI is deeply optimized for industry-specific domain knowledge, regulatory requirements, and workflows. Harvey, the largest legal AI firm, reached a valuation of $11 billion (approximately ¥1.65 trillion) and over $1 billion in cumulative funding in March 2026, recording ARR of $190 million — a 3.9x year-over-year growth. Healthcare AI company Tempus posted fiscal 2025 revenue of $1.27 billion (up 83% year-over-year) with a market cap of $8.9 billion, while Abridge raised $300 million in a Series E led by a16z and Khosla Ventures at a valuation of $5.3 billion. Enterprise payments company Ramp reached $1 billion in annual revenue at a valuation of $32 billion. a16z has assembled the largest AI-focused fund in history at $20 billion, advancing the thesis that "for every $1 spent on software, $6 is spent on services — and AI will consume that $6." Bessemer predicts that "more than five Vertical AI companies will surpass $100 million ARR within two to three years," while Sequoia has set a goal of "more than 10 portfolio companies exceeding $100 million ARR by 2027." The Vertical AI market is projected to expand from approximately $11–13 billion (roughly ¥1.65–1.95 trillion) in 2025 to $115.4 billion (approximately ¥17.31 trillion) by 2034 (CAGR of 24.5%). This article provides a comprehensive examination of the overall landscape of Vertical AI, key vertical companies and products, technical approaches, VC investment theses, trends in the Japanese market, and the outlook going forward.

What is Vertical AI — The "Last Mile Problem" of General-Purpose AI

Vertical AI refers to AI solutions built specifically for particular industries. It is fundamentally different from general-purpose models (Horizontal AI) such as ChatGPT or Claude in that it is deeply optimized for industry-specific knowledge systems — legal case law, medical clinical protocols, accounting standards, building codes, and so on.

General-purpose LLMs have a "last-mile problem." They possess broad knowledge, but their accuracy in specialized domains falls short. The founding story of Harvey (legal AI) illustrates this well. Co-founder Gabriel Pereyra (formerly of Google DeepMind/Meta) showed GPT-3 to his roommate Winston Weinberg (formerly of O'Melveny & Myers) and tested it with 100 legal questions on California tenancy law. Three attorneys evaluated the results, and 86 answers were judged to be of sendable quality for clients. However, the remaining 14 contained inaccuracies that would be unacceptable in a regulated industry. "Mostly correct" is simply not good enough.

Technically, state-of-the-art vertical AI adopts a hybrid approach: "fine-tuning for behavior, RAG for knowledge." Domain-specific vector DB encoders are used to understand industry semantics, and Agentic RAG merges agentic planning and execution capabilities with fact-verification. Chain-of-Thought reasoning logs have been shown to reduce hallucinations by 30%.

Legal AI — Harvey's Extraordinary Growth

The most mature segment of vertical AI is legal AI.

Harvey was founded in the summer of 2022 and reached a valuation of $11 billion (approximately ¥1.65 trillion) in just four years. It raised $200 million in a March 2026 round co-led by GIC and Sequoia, bringing total fundraising to over $1 billion. ARR grew 3.9x year-over-year, from $50 million at end of 2024 to $190 million in January 2026. It is used by more than 100,000 lawyers across a majority of the AmLaw 100, over 500 in-house legal teams, and more than 1,300 organizations in 60 countries. Allen & Overy deployed Harvey to 3,500 lawyers and processed approximately 40,000 queries. PwC rolled it out to 4,000 professionals across 100 countries.

When investing in Harvey, Sequoia described it as "the most compelling example of putting LLMs to work in real business operations." The story of meeting with OpenAI leadership on July 4, 2022 to secure the seed round and gain early access to GPT-4 stands as one of the most efficient fundraises in startup history.

Casetext used its GPT-4-powered AI legal assistant "CoCounsel" to perform document review, research memos, and contract analysis in minutes, and was acquired by Thomson Reuters in 2023 for $650 million (approximately ¥97.5 billion). EvenUp (an AI platform specializing in personal injury law) surpassed a $2 billion valuation in its Series E in 2025. Legal tech investment in 2025 hit an all-time high, driven by AI.

Medical AI — Proven Stage with Over $1 Billion in Revenue

Healthcare AI is the largest investment target in vertical AI, accounting for approximately half of all VC investment.

Tempus (founded by Eric Lefkofsky, co-founder of Groupon) went public in June 2024, achieving $1.27 billion in FY2025 revenue (up 83.4% year-over-year). Its market capitalization reached $8.94 billion as of March 2026. As a precision medicine AI platform, it integrates genomic and clinical data to support personalized treatment.

Abridge (clinical documentation AI) automatically converts physician-patient conversations into structured medical records. Its Series E in June 2025 ($300 million, led by a16z) valued the company at $5.3 billion, representing a 93% increase in valuation in just four months. ARR reached $100 million in May 2025, and the platform has been deployed across more than 100 of the largest health systems in the United States.

Hippocratic AI (healthcare AI agents) is valued at $3.5 billion with cumulative funding of $404 million. It has partnered with more than 50 health systems, payers, and pharmaceutical companies, processing over 115 million clinical patient interactions with zero safety incidents.

Viz.ai features more than 50 FDA-cleared AI algorithms and is deployed in approximately 2,000 hospitals across the United States. It supports the lives of more than 230 million people and achieved profitability in its healthcare business in 2025.

Accounting & Finance AI — [Ramp](/services/ramp) Annual Revenue $1 Billion

Accounting and finance is the domain where the impact of AI-driven automation can be most directly quantified.

Ramp reached $1 billion in annual revenue in August 2025, achieving a valuation of $32 billion. As a corporate card and AI expense management platform, its total payment volume (TPV) stands at $57 billion (up sharply from $22.3 billion in 2023). Through AI-powered expense tracking automation, accounts payable automation, procurement management, and spend analytics, it has become possible to do "three times the work on Ramp compared to two years ago" (co-founder and CTO).

Truewind (AI bookkeeping, YC alumni) uses AI agents to convert financial documents into GL journal entries, automating close management, reconciliation, and transaction coding. Digits is equipped with four AI agents — a bookkeeping agent, a payments agent, a finance agent, and a reporting agent — enabling comprehensive automation of accounting operations.

a16z analyzes the structural transformation AI is bringing to the accounting industry in "The Rise of Vertical AI in Accounting."

Architecture & Construction AI — Automation from Design to Estimation

Architecture and construction AI covers a wide range of use cases, from automated generation of design documents to cost estimation.

Swapp automatically generates comprehensive construction drawing sets from BIM models and performs compliance checks against building codes. It analyzes design practices and annotation conventions from past projects to build custom rule sets. Togal.AI uses deep learning to analyze construction drawings and automates the most time-consuming parts of construction cost estimation. Alice Technologies is one of the leading companies in the AI construction planning space.

Autodesk is progressively integrating AI capabilities into its existing products such as Revit and AutoCAD, driving AI adoption across the entire industry.

Manufacturing AI — Andrew Ng's "$500 Project"

Manufacturing AI is an area where Japan has particular strengths.

Landing AI (founded by Andrew Ng) enables AI-based manufacturing inspection using visual inspection AI that can be deployed within days. Ng has stated that "previous AI was for billion-dollar applications, but now $500 projects are possible," and is driving adoption in small-scale, industry-specific use cases such as inspecting cheese uniformity for pizza makers and determining wheat harvest timing on farms. The company participates in NVIDIA Metropolis and is integrated with the Jetson platform.

Sight Machine (cumulative funding of $85.5 million) provides industrial AI solutions for manufacturing process optimization and automation. Fictiv enables automation of AI material recommendations, manufacturing center selection, real-time inspection, and DFM validation. The AI visual inspection market is projected to grow from $4.13 billion in 2024 to $12 billion in revenue by 2033.

Other major verticals — insurance, education, agriculture, logistics

The wave of vertical AI is reaching every industry.

Insurance: Lemonade has fully automated 55% of all claims (resolved in seconds) as of end of 2025, with 96% of first notices handled by AI chatbots. Total in-force premiums reached $1.24 billion (up 31%). Tractable uses computer vision to process 90% of auto insurance estimates without human intervention, with 98% completed within 15 minutes.

Education: Khanmigo (Khan Academy + OpenAI) is a GPT-4-powered AI tutor at $4/month, available in over 180 countries. The education AI market is projected to grow from $6.9 billion in 2025 to $41 billion by 2030 (CAGR 42.83%).

Logistics: project44 has unveiled an AI-powered intelligent TMS that negotiates freight without human intervention. FourKites is automating over 40% of tasks with AI.

a16z's Vertical AI Thesis — "Converting Labor into Software"

a16z (Andreessen Horowitz) has developed the most comprehensive investment thesis on vertical AI.

In April 2025, it raised the largest-ever $20 billion AI-dedicated fund and provides portfolio companies with more than 20,000 NVIDIA GPUs through the "Oxygen" initiative. AI investments account for more than 40% of a16z's total assets under management.

The core of the thesis is "Services as Software." Against $313 billion in U.S. software spending, labor spending stands at $10.5 trillion——software represents just 3% of labor. By converting Labor into Software through AI, this vast market becomes accessible. Niche industries previously deemed "too small to address"——such as dry cleaners, chiropractic clinics, and veterinary services——become viable markets through Vertical AI.

Marc Andreessen predicts "hyper-deflation of business services via AI——things that cost $100 becoming a penny."

VC Chorus — Sequoia, Bessemer, Y Combinator

VCs beyond a16z are also making big bets on vertical AI.

Sequoia Capital is focusing on the "second wave" of AI adoption, positioning Vertical Agents (legal, healthcare, finance, sales, engineering) as a key investment area. Partner Sonya Huang is prioritizing vertical applications over infrastructure and has invested in companies like Harvey. The goal is for more than 10 portfolio companies to exceed $100M ARR by 2027.

Bessemer Venture Partners published "Building Vertical AI: An Early Stage Playbook for Founders" in January 2026, presenting 10 principles. The core insight is that "Vertical AI accesses the 'labor cost' line of the P&L rather than traditional IT software budgets, capturing exponentially larger budgets." They predict more than 5 companies will reach $100M ARR within 2–3 years.

The batch trends at Y Combinator are clear. AI-focused startups in 2025 numbered 1,140 (53% of the total). In the Summer 2025 batch, 88% of the 141 companies were AI-native, with domain-specific agents for insurance claims processing, mortgage applications, warehouse logistics, and more surging in numbers.

Wellington Management analyzes that Vertical AI agents hold a significant advantage over SaaS solutions, and predicts that agentic AI will address labor shortages, enabling Vertical AI companies to grow "larger than previously imagined."

Japan's Trends — Development of Regulatory Frameworks and the Rise of Domain-Specific Companies

The vertical AI movement is accelerating in Japan as well.

On May 28, 2025, the "Act on Promotion of Research, Development, and Utilization of AI-Related Technologies" (AI Promotion Act) was enacted. It is a basic framework aimed at balancing the promotion of innovation with risk management.

MNTSQ (Legal AI) combines the legal expertise of Nagashima Ohno & Tsunematsu with NLP technology, providing comprehensive support for contract drafting, review, management, and knowledge building through "MNTSQ CLM." In September 2025, the company announced "MNTSQ Legal Agent." Total funding raised stands at $20.48 million (led by MUFG).

Ubie (Healthcare AI) improves patient intake at hospitals through an AI symptom checker, contributing to enhanced diagnostic accuracy and patient outcomes. The PMDA, which oversees AI regulation as a medical device, has significantly reduced costs and administrative burden through the IDATEN system (Improvement Plan for Adaptive AI Pre-approval Through Enhanced Notification), which eliminates the need for re-approval with each update. The Cabinet Office has invested 22 billion yen in AI-assisted medical diagnosis support using generative AI.

In manufacturing, Preferred Networks is driving industrial applications of deep learning, and NVIDIA announced a partnership with Fujitsu for joint Vertical AI development (in healthcare, manufacturing, and robotics) in October 2025.

Vertical AI by the Numbers — Rapid Expansion of Markets and Investment

The vertical AI market is projected to expand from approximately $11–13 billion (roughly ¥1.65–1.95 trillion) in 2025 to $74.5 billion (approximately ¥11.175 trillion) by 2033 (CAGR 28.3%), and to $115.4 billion (approximately ¥17.31 trillion) by 2034 (CAGR 24.5%).

Total AI investment in 2025 stands at $203 billion, up 75% year-over-year. AI now accounts for 52.7% of global VC investment totals, surpassing a majority share for the first time. Investment specifically targeting vertical AI reached $3.5 billion (roughly triple the $1.2 billion recorded in 2024). Some 58% of AI investment is concentrated in mega-rounds of $500 million or more.

In terms of valuation multiples, AI companies command a median of more than 10x revenue, compared to less than 5x for traditional SaaS. Vertical SaaS unit economics outperform Horizontal SaaS by 25–30%.

Future Outlook — "Software Is Eating Labor"

Industry leaders are showing unusual optimism about the future of Vertical AI.

Through 2026: 40% of enterprise applications will integrate task-specific AI agents (a sharp rise from under 5% in 2025). Harvey (legal), Tempus (healthcare), and Ramp (finance) will establish precedents of "over $100M ARR" in their respective verticals.

2027–2028: As predicted by Bessemer and Sequoia, 5–10 Vertical AI companies will reach over $100M ARR. Following legal, healthcare, and finance, full-scale AI adoption will advance in construction, manufacturing, education, and agriculture verticals as well.

2030s: According to WEF projections, AI will create 170 million new positions while displacing 92 million existing ones (a net gain of 78 million). The redesign of human-AI collaboration will generate $2.9 trillion in economic value in the United States alone.

To borrow Bessemer's words: "For every $1 spent on software, $6 is spent on services." Vertical AI is coming for that $6. This means a transformation of the same scale as SaaS reshaping enterprise IT spending is about to happen — but this time directed at enterprise labor costs.

Impact on the Industry

First, vertical AI is establishing itself not as a "complement to general-purpose AI" but as an "independent, massive market." The scale of Harvey (valuation: $11 billion), Tempus (market cap: $8.9 billion), and Ramp (valuation: $32 billion) demonstrates that vertical AI companies can achieve valuations on par with general-purpose AI companies.

Second, a16z's "Services as Software" thesis shows that the true TAM for vertical AI is not IT software budgets ($313 billion in the US) but labor expenditure ($10.5 trillion in the US). The structural advantage of vertical SaaS unit economics being 25–30% superior to horizontal SaaS enables access to this enormous market.

Third, the fact that 88% of Y Combinator's Summer 2025 batch are AI-native companies, with a rapid increase in domain-specialized agents, indicates that the next-generation startup ecosystem is forming around vertical AI.

Fourth, Japan is steadily building the foundation for vertical AI adoption through the enactment of its AI Promotion Act, the PMDA's IDATEN system, the rise of domain-specialized companies like MNTSQ and Ubie, and the NVIDIA–Fujitsu vertical AI partnership. AI inspection in manufacturing and the regulatory framework for healthcare AI are areas where Japan could lead the world in practical implementation.

References: Grand View Research Vertical AI Market Report, Market.us Vertical AI Market, Crunchbase AI Funding Trends (2025), Harvey $11B Valuation (CNBC, 2026/3), Harvey Founding Story (Contrary Research), Sequoia "Partnering with Harvey", Casetext Acquisition by Thomson Reuters (TechCrunch, 2023), EvenUp $2B+ Valuation (Crunchbase), Tempus AI IPO (Fierce Healthcare, 2024), Tempus Stock (Motley Fool), Abridge Series E (Abridge Blog, 2025), Hippocratic AI Series C (BusinessWire, 2025), Viz.ai 2025 Record, Ramp $1B Revenue (Fortune, 2025), Ramp $32B Valuation (PM Insights), Truewind (Y Combinator), a16z "The Rise of Vertical AI in Accounting", a16z "Vertical SaaS: Now with AI Inside", a16z $20B AI Fund (Benzinga), Bessemer "Building Vertical AI Playbook" (2026/1), Bessemer "Ten Principles for Building Strong Vertical AI", Sequoia "AI in 2025", YC S25 Batch Analysis (Extruct), YC AI Agents (PitchBook), Wellington Management Vertical AI Agents, Satya Nadella AI Vision (Bank Info Security), Microsoft 2025 Annual Report, Jensen Huang CES 2025 (NVIDIA Blog), NVIDIA NIM Healthcare, Andrew Ng Landing AI (NVIDIA Blog, VentureBeat), Marc Andreessen Predictions (Fortune), Swapp AI, Alice Technologies, Togal.AI, Lemonade Claims Automation, Khanmigo (Khan Academy), Japan AI Act (Chambers, 2025), PMDA SaMD Report, MNTSQ (Tracxn), Ubie Healthcare AI, NVIDIA–Fujitsu Vertical AI Partnership, WEF Future of Jobs (2026/1), AI Workforce Trends (Gloat)