The transformative power of vertical AI agents

Following its release in November 2022, OpenAI’s ChatGPT reached 100 million users in just two months.¹ Since this remarkable success, advancements in both multimodal and specific-focus agentic AI applications have continued to accelerate. The emerging AI technology landscape presents significant opportunities for companies that innovate beyond general-purpose large language models (LLMs) to create industry-specific applications that are far more powerful and have much larger total addressable markets than traditional software-as-a-service (SaaS). Innovation in vertical-industry-specific agentic AI is a burgeoning investment opportunity that is only beginning to take shape across private markets. 

Agentic AI will not only help industries like law, consulting, and financial services replace low-level-knowledge workers, but it will also be a tremendous benefit in areas like health care and government where skilled labor shortages, combined with growing demand for services, are creating consumer pain points. Here, we explain what’s driving the vertical-agent investment opportunity, how agentic AI will usurp legacy SaaS and human labor, and which industries represent the greatest sources of demand. 

The next big shift for AI: Higher-order functions at far less cost

To understand the transformative potential of vertical AI agents, a bit of historical context helps. Among major technological advancements of the last decade, few had greater impact than cloud and mobile computing. Building on the first wave of general applications for those innovations, vertical software companies developed solutions tailored for specific industries. Companies like Toast (restaurants), Procore (construction), Veeva (life sciences), and nCino (banking) customized cloud and mobile solutions to become systems of record for their adopters. That success helped those innovators expand into higher-value fintech offerings such as payroll processing and other services, transforming them into foundational operating systems for their industries.

Today, a similar dynamic is playing out with agentic AI. Building upon early generations of LLMs, agentic AI progresses beyond simple task completion toward strategic decision making and problem solving by adapting autonomously in real time. Traditional SaaS automated simple functions like payments, workflows, and data aggregation. The next generation of AI-enabled tools goes further, executing the resulting tasks that SaaS currently allows humans to do: addressing payment-related errors, optimizing workflows, and interpreting aggregated data, for example. In addition, while agentic AI is still in early development, these agents will soon execute industry-specific tasks and procedures with the same — or better —skill and efficacy than humans.

The key advantage of the agentic layer is its ability to process unstructured data before it reaches traditional systems of record. (Unstructured data includes meeting minutes, clinicians’ notes, audio and video recordings, emails, social media commentary, and so on.) This shift in data flows could erode the power of existing systems of record, as agentic AI virtually eliminates those systems’ data advantage. For vertical industries, this shift presents tremendous potential for operational efficiency and cost savings. AI agents can identify, analyze, and solve existing problems — and anticipate unforeseen ones. They automate complex, sensitive, and labor-intensive workflows. They facilitate decision making and minimize errors. Above all, they lower capex. The race to reduce machine-intelligence spending is already on. China’s DeepSeek recently touted a 94% decline in costs,² and given the capital flowing into GPU infrastructure, businesses that aim to leverage AI for highly specialized, higher-order tasks may be able to do so for minimal capex. (History again hints at the potential: In the mid-2000s, Amazon Web Services’ pay-as-you-go model slashed cloud storage costs, catalyzing an explosion in SaaS startups.)

Implications of agentic AI for labor markets and industry growth

In their ability to perform tasks for end users, agents can tap a much larger total addressable market (TAM) than enterprise software alone. Enterprise software is currently a US$315 billion market.³ In contrast, agentic AI has the potential to narrow the gap in future shortages of human labor, enabling vertical AI companies to grow much larger than previously imagined.

In our view, the companies most likely to succeed in the vertical-agent market will: 

• Produce agents capable of processing copious amounts of unstructured, industry-specific data
• Integrate agents with customers’ existing systems of record, rather than require total replacement
• Start with — and continuously build — knowledge of target vertical industries’ evolving needs
• Keep up with rapid technical development and competing disruptors in agentic AI

Largest potential markets for vertical agents

While many industries are set to accelerate the adoption of agentic technology, we believe early adopters will be those that face significant labor challenges and/or complexities and have unique data sets that could benefit from specialized vertical solutions. For us, the most interesting categories today are health care, government, and the small-business and independent economy. 

Health care 

Agentic technology may be able to process massive amounts of unstructured health care data and deliver personalized recommendations across multiple dimensions with greater speed and accuracy. 

Health care is a complex, high-cost industry drowning in information. Many countries, including the US, face the dual challenges of worsening labor shortages and rising demand from aging populations. By 2030, one in six people will be aged 60 or older.⁴ Given rising demand for care, the World Health Organization projects a shortfall of approximately 18 million health care workers by the end of the decade.⁵ In the US, the National Center for Health Workforce Analysis projects a 10% shortage in registered nurses by 2027.⁶ Clinician and administrator shortages can lead to attrition across health care systems, exacerbating the problem. At the same time, approximately 80% of medical data is unstructured.⁷ Even small efficiency gains and error reductions can help yield cost savings and improve patient outcomes. 

We believe the quickest initial adoption of agentic solutions will be in non-clinical areas that make doctors and nurses more efficient. AI-enabled medical scribes have already driven a substantial return on investment for hospital systems by processing clinical conversations. The University of Chicago Medicine recently reported a 40% increase in “patient comprehension” from using Abridge’s AI scribing technology.⁸ To help physicians — who report spending nearly as much time on paperwork as they do on patient care⁹ — AI assistants can automate charting, coding, and routine diagnostics, saving time. On the patient-intake side, Paratus Health facilitates these tasks with voice-assisted AI. Other administrative processes, such as confirming and handling insurance information, can also be streamlined. 

Agentic solutions will eventually be implemented in clinical applications as well. For instance, a physician’s AI assistant can share a patient’s vitals, past medical history, and take notes during an appointment. Suki, a California startup, already offers something similar. Clinical decision-support systems will be able to provide physicians with multiple treatment options in real time; AI will be able to assist radiologists in interpreting scans and images with greater precision; and so forth. The possibilities extend beyond what we can cover here.

Government

Agentic AI can enhance efficiency and drive cost savings. By automating routine tasks and supporting human staff, AI can help local governments manage workloads, improve public safety, and better serve their constituents.

Governments are increasingly focused on efficiency and cost savings. In the US, there are over 90,000 local government units, each with unique systems and procedures. This fragmentation challenges traditional software solutions. Agentic AI, trained on specific municipal data, can scale across departments, offering a significant opportunity in a market where US state and local governments spend approximately US$3.9 trillion annually.¹⁰

Government data, from permits to voting forms, is largely unstructured. AI agents process this data faster than human staff, reducing administrative burdens and burnout, increasing efficiency, and providing residents with quicker access to services. These improvements may enhance trust and satisfaction among constituents. AI can also automate time-consuming tasks such as processing permits and licenses. In business licensing, AI guides applicants through forms, verifies documents, and makes approval decisions. It can also evaluate zoning rules, facilitate inter-department communication, and send renewal reminders, ensuring compliance.

Federal, state, and local governments often face staffing challenges. In the US, older workers are retiring, and young people are joining the ranks of public service in smaller numbers. Government agencies’ IT departments in particular face growing strains as young workers opt for higher salaries in the private sector. AI assistants can execute repetitive tasks like application reviews, payroll processing, and routine inquiries, around the clock. Virtual help desks powered by AI can answer frequent questions, supply voter information, and assist with appointments, allowing employees to focus on strategic activities and critical thinking. 

For public safety, paramount for local governments, agentic solutions ensure timely response and resource deployment. For example, New York-based Prepared uses AI voice agents for emergency dispatch centers, transcribing and summarizing 911 call audio, and integrating with legacy systems to help emergency centers respond faster and maintain better records. AI also plays a role in real-time alert monitoring for natural disasters, supporting first responders with timely information. It can optimize transcription and record-keeping for emergency services, ensuring accurate documentation.

Small, independent businesses

AI agents can provide access to resources and help small companies operate with the efficiency and speed typically associated with large enterprises.

Between 400 million and 500 million small- and medium-sized businesses (SMBs) exist globally.¹¹ We believe agentic AI can revolutionize SMB operations by offering access to resources that were traditionally affordable only by large enterprises. In other words, agentic AI can allow SMBs to compete at enterprise levels without enterprise budgets. In the skilled trades, carpenters, electricians, welders, and plumbers are contending with labor shortages amid growing demand from infrastructure build-out, real estate development, and the energy transition.¹² We believe that if it were easier to start and run a small business, more people might be enticed to enter the trades. 

SMBs often juggle multiple priorities with limited resources, including back-office tasks like invoice and receipt management. AI models can automate document processing, extract key figures, and apply accounting rules, helping SMBs function efficiently without extensive accounting teams. Agentic technology can help small company owners operationalize both the front and back ends of their businesses, thus allowing tradespeople to focus on developing their skill set. Agents can streamline workflows such as customer service and inventory management, allowing SMBs to operate well and leanly, with less staff and overhead costs. 

For customer service, agentic chatbots and voice-enabled agents offer immediate support, enhancing customer engagement and protecting brand reputation. These agents can manage inquiries, troubleshoot issues, and ensure consistent customer experience. In sales, agents can triage customers, generate leads, and engage prospects, freeing sales teams to focus on closing deals. Automated customer relationship management keeps the sales pipeline robust by updating data, suggesting leads, and crafting outreach templates. Agentic “salespeople” can engage prospects and follow up on leads, driving sales growth.

Back-office tasks like tax, finance, and inventory management benefit from agentic accuracy and efficiency. Agents tackle bookkeeping, invoicing, and expense tracking, while optimizing inventory by minimizing stockouts and overstock. They also coordinate logistics and analyze sales trends, ensuring smooth operations and driving supply chain optimization. For example, Cactus is building an AI copilot for solopreneurs that handles administrative work. Their AI technology answers the phone, qualifies customers, and books revenue —so entrepreneurs can run their businesses on autopilot.*

Conclusion 

Vertical AI agents are revolutionizing industries by enhancing efficiency, reducing costs, and addressing labor shortages. By processing unstructured data and integrating with existing systems, these agents offer significant advantages. In highly complex, fragmented, and data-heavy areas like health care, government, and the small business economy, agentic AI can unlock new levels of productivity and innovation. As we continue to explore and develop these technologies, the potential for growth and improvement across various sectors is immense. AI-driven solutions are becoming foundational to industry operations. We are committed to researching and engaging with innovative early-stage companies that deliver vertically integrated solutions for these and other sectors of the economy. 

¹David Curry, “ChatGPT revenue and usage statistics,” Business of Apps, 14 May 2025. | ²“Anthony Di Pizio, “DeepSeek isn't the only low-cost AI startup. Here's what it means for OpenAI and Nvidia,” The Motley Fool, 31 January 2025. | ³“Enterprise software – worldwide,” Statistica. | ⁴“Aging and health,” World Health Organization, 1 October 2024. | ⁵Charles H. Jones and Mikael Dolsten, “Health care on the brink: navigating the challenges of an aging society in the United States,” npj Aging, 6 April 2024. | ⁶“Nurse workforce projections, 2022 – 2027,” HRSA, November 2024. | ⁷Brian Eastwood, “How to navigate structured and unstructured data as a health care organization,” HealthTech, 8 May 2023. | ⁸Guru Sandar, “From pilot to enterprise-wide implementation: Three partners share why they expanded use of Abridge,” Abridge, 22 January 2025. | ⁹ABC News, “Doctors may spend nearly half their time on paperwork, study says,” ABC News, 6 September 2016. | ¹⁰“Government receipts and expenditures,” Bureau of Economic Analysis, data from 29 May 2025. | ¹¹“OECD SME and entrepreneurship outlook 2023,” Organisation for Economic Co-operation and Development, 27 June 2023. | ¹²Ezra Greenberg, et al, “Tradespeople wanted: The need for critical trade sills in the US,” McKinsey & Company, 9 April 2024. 

*Not all companies referenced are owned by the funds managed by Wellington Management. For the full list of our Early-Stage strategy investments, please refer to our investments page.