Vertical AI Part 3: Business Models

The recent evolution of generative AI and Large Language Models (LLMs) has become a major driving force in expanding AI, which was previously utilized only in select industries, into a wide variety of sectors all at once. In particular, the "human-like" nature of LLMs, capable of processing a broad range of information such as text, images, audio, and code, has enabled the automation and streamlining of large-scale, language-intensive tasks that were difficult to achieve with existing software applications. When viewed as "Vertical AI" (industry-specific AI), innovative solutions that were previously unimaginable are emerging one after another in many specialized fields such as law, medicine, insurance, and construction.

Article:

In this article, we will cover three new business models that support the development of such Vertical AI: "Copilots," "Agents," and "AI-enabled Services," and explain their respective characteristics, concrete use cases, and trends in pricing strategies in an easy-to-understand manner. Furthermore, we will organize the hybrid and performance-based fee structures that leading companies in these new business models are beginning to adopt, and explore the key points of business models in the Vertical AI era.

1. The Expansion and Background of Vertical AI

1-1. The Potential for Industrial Transformation through LLMs

Just as cloud computing has transformed many industries over the past 20 years, the emergence of Large Language Models (LLMs) is permeating industry-specific AI solutions known as "Vertical AI" into various sectors. Especially in industries with high volumes of text, such as documents and contracts (law, healthcare, insurance, etc.), AI's language processing capabilities have the potential to significantly exceed the effectiveness of traditional software applications.

In fact, "Vertical AI" is beginning to not only streamline workflows but also automate complex and high-cost tasks that were previously considered impossible without human intervention. Fields with many human-centric professional tasks, such as contract review at law firms or medical record creation in clinical settings, are estimated to have a market size approximately 10 times larger than the software market, and expectations for Vertical AI are rising higher than ever.

1-2. The Importance of Business Models

However, no matter how innovative the technology or how high the performance of the AI model developed, a robust business model is essential to introduce it into the industry as a concrete service and provide continuous value. Without designing appropriate pricing and delivery formats for the output generated by AI, long-term business continuity becomes difficult.
In the latter half of this article, we will introduce how new business models such as Copilots, Agents, and AI-enabled Services have emerged and what pricing strategies they adopt, with concrete examples.

2. The Emergence of Copilots

Copilots refer to AI applications where AI works right alongside human workflows, supporting human judgment and input at key points. While users still perform the primary decisions and operations, AI improves work efficiency and accuracy dramatically by providing input assistance, suggestions, translation, and summarization.

As for the pricing model for Copilots, "subscription billing based on the number of users (per seat)" is the mainstream. This is similar to traditional cloud software pricing structures, where AI usage fees increase as the number of users in an organization increases. Major companies like Microsoft and Salesforce have also succeeded in expanding revenue by adding Copilot features to existing products and charging an additional $30 per user per month.

2-1. Copilots in the Code Field

"GitHub Copilot" is one of the first code-completion Copilots widely adopted in the market, offering features where developers can receive suggestions for subsequent code as they type, and ask the AI directly if they have any questions.
In actual use cases, a productivity improvement of approximately 55% has been reported when writing code, and developer satisfaction is said to be high. Furthermore, like Supermaven (which later merged with Cursor), solutions are emerging that utilize massive context windows (such as 1 million tokens) to understand the entire project codebase and provide more accurate auto-completion.

2-2. Copilots in the Text Field

Copilots are also active in tasks with high volumes of text, such as document creation and contract review. For example, Harvey specializes in the legal field, quickly summarizing contracts and providing immediate answers to questions about specific parts of the text. Sixfold AI for the insurance industry supports underwriters in quickly grasping customer risk profiles by consolidating different documents and data sources, contributing to productivity improvements in text-centric operational tasks.

2-3. Copilots in the Audio Field

Due to advancements in speech recognition (Speech-to-Text), automatic transcription and summarization of conversation content can now be achieved with high precision. In the medical field, Abridge provides a feature that transcribes conversations during medical examinations in real-time, detects the doctor's specialty and the patient's native language, and immediately translates them into an English medical record. Since doctors only need to edit the result, it can significantly reduce the cost and time required for creating medical records.

2-4. Copilots in the Image Field

Copilots specialized in image generation and image analysis are active in fields such as design and construction. For example, Workpack AI and Togal AI in the Architecture, Engineering, and Construction (AEC) industry automatically analyze blueprints, measure dimensions, and add labels, shortening the processes of estimation and design modification.

3. Evolution of Agents

While Copilots are designed to "support humans by their side," Agents are AI models that "automate tasks on behalf of humans." The goal here is to produce a certain output with almost no human intervention, which has the potential to directly improve corporate productivity by reducing headcount and labor costs.

3-1. Exploring New Pricing Models

Since agents often replace or supplement tasks performed by humans, pricing based on "labor cost savings" or the "value of the generated output" is beginning to be considered. It is expected that pricing based on metrics comparable to human work, such as "equivalent to the cost of hiring one SDR (Sales Development Representative)" or "equivalent to one seat in a call center," will become mainstream.

3-2. Representative Examples

• Bosh (Relevance AI): Automates the work of an SDR (Sales Development Representative) in the software sales domain. It handles everything from lead research and contact to meeting arrangements.

• LinkedIn Hiring Assistant: Automates candidate sourcing and drafting job descriptions in recruitment activities, reducing the workload for recruiters.

• Slang: AI handles initial responses to phone inquiries for restaurants, taking care of reservation requests and answering frequently asked questions. Significantly reduces customer service costs for understaffed restaurants.

• Assort Health: Turns medical institution call center operations into agents, automating patient appointment management and simple medical interviews.

• Tennr: Automates medical back-office tasks (document organization, transfer procedures, referral processing, etc.).

4. Potential of AI-enabled Services

AI-enabled Services place AI automation at the core of service delivery, a model that fundamentally replaces (or significantly streamlines) traditional forms of outsourcing and consulting. Unlike Copilots and Agents, they directly provide the final output required by the customer (such as a complete set of contracts or medical billing documents), so the customer does not need to operate the AI themselves.

4-1. Pricing Points

The price levels previously set by existing service providers (legal services, medical billing agencies, third-party insurance administrators, etc.) often serve as the benchmark for pricing AI-enabled Services. Compared to labor-intensive service costs, companies have the advantage of being able to offer services cheaper than existing players while maintaining profit margins, even while reducing costs through AI automation.

4-2. Representative Examples

• EvenUp: Automatically generates "demand packages" (complete sets of claim documents) related to traffic accidents and injuries for law firms. Improves both speed and accuracy for tasks that previously took paralegals a lot of time.

• SmarterDx: Automatically detects missed insurance claims and medical billing from hospital inpatient charts, supporting or replacing the work performed by Clinical Documentation Integrity (CDI) specialists.

• Reserv: Automates the insurance claims process and performs advanced analysis of data handled by carriers and MGAs (Managing General Agents). It is showing signs of replacing large TPAs (Third Party Administrators).

5. Business Model Examples of Emergent Vertical AI Leaders

As more companies adopt AI, models that link pricing to "outcomes" or "usage volume," rather than just subscription models, are gaining attention. In the case of usage-based pricing, it is common to directly tie service prices to output metrics, such as the number of documents generated or the number of inquiries resolved by automated responses. Additionally, hybrid models that set a certain base monthly fee or minimum usage guarantee while charging for additional usage are also increasing.

The following are examples of some companies.

• DeepL

  • Service Details: Multilingual translation service

  • Pricing Model: Billing based on the number of users + limits on translated files (number of editable files)

• EvenUp

  • Service Details: Automated generation of injury-related demand packages

  • Pricing Model: Pay-as-you-go based on the number of packages generated

• Intercom

  • Service Details: AI-powered customer communication

  • Pricing Structure: The “Fin AI Agent” charges $0.99 for each issue automatically resolved by AI

    • Copilot features include a free tier, such as “up to 10 free items per month”

• Zendesk

  • Service Details: Customer support and inquiry management

  • Pricing Structure: Usage-based billing based on the number of tickets automatically resolved by AI

In this way, when the output provided by AI delivers clear value to the customer (such as labor reduction, increased speed, or cost savings), performance-based pricing is effective. On the other hand, for customers who do not exceed a certain scale of usage, combining subscription or freemium models can lower the barrier to entry while also holding the potential for future expansion in usage.

Due to the rapid advancement of AI technology centered on large language models, a wave of major transformation is hitting industries that previously saw limited benefits from software adoption. “Vertical AI” holds great potential to partially automate even tasks that require human creativity and judgment, enabling the reallocation of resources to higher-value work.

The business models being newly defined within this space are broadly divided into three categories: “Copilots,” “Agents,” and “AI-enabled Services,” all of which are exploring added value and pricing that were not present in traditional SaaS (Software as a Service) models. While Copilots powerfully support human tasks, Agents aim for near-total automation, and AI-enabled Services seek differentiation through the provision of high-value services.

Regarding pricing strategy, there is a growing trend toward complex models such as “per-user,” “performance or output-based,” and “hybrid” models. This allows companies to keep initial AI adoption costs low while enabling flexible billing based on the scope of use and the expansion of results.

LLMs and peripheral technologies (such as voice recognition and image generation) will continue to evolve day by day, and even more diverse Vertical AI businesses will emerge. Whether startups or major tech companies can enter the market and quickly build appropriate pricing strategies will be a critical point that determines competitive advantage across the entire industry. Amidst these dynamic changes, “how to define the value your company provides and design a billing model that matches it” will be the key to business success in the Vertical AI era.

Related Articles