Vertical AI: how industry-specific intelligence is transforming the business landscape
By White Star Capital’s Early Growth Fund team
Notable Horizontal AI tools, such as ChatGPT and Github’s Copilot, have been widely adopted, with 96% of our White Star Capital’s portfolio already leveraging these technologies. Will we see the same market penetration as Vertical AI tools come to market?
As part of our research, we surveyed our 90+ portfolio to understand the opportunities, risks, and challenges for companies building Vertical AI solutions. The results of that survey are included throughout this deep dive.
What is Vertical AI?
While Horizontal AI excels in versatility, Vertical AI leverages deep, industry-specific knowledge to offer precise, tailored functionalities that address distinct challenges.
However, while Vertical AI’s precision holds great promise, its success hinges on solving clear, specific problems. Otherwise, implementing AI just for the sake of it, whether vertical or horizontal, can result in low engagement and underwhelming outcomes, a challenge many companies face today.
For example, customer service is a prime area for the adoption of vertical AI applications, lending itself to natural language processing and the pattern of automation of low-value work — 52% of portfolio companies surveyed have integrated AI tools for customer service.
The success of these models is dependent on the proprietary industry-specific data, as well as publicly available data, required to train them. An AI chatbot is only as effective as the quality of its training data. In customer service, high-quality data ensures accurate, relevant, and appropriate responses, enhancing customer satisfaction, increasing the product’s value, and strengthening brand loyalty.
The release of GPT-4 in March 2023 was a breakthrough with its enhanced accuracy and safeguards against hallucinations. But, it wasn’t ready for customer service “out of the box”.
Many companies are now building their own AI tools, and exploiting their datasets, to create tailored solutions that address specific industry challenges at scale (69% of our portfolio companies have built their own AI tools).
While building AI tools requires significant expertise and resources (successful AI applications rely on deep domain knowledge), the potential payoff is huge.
The Vertical AI advantage: why domain-specific models outperform horizontal approaches
The early days of AI were dominated by a ’horizontal’ strategy, with companies like Google, Amazon, IBM, and Microsoft creating broad AI solutions for various tasks. These Machine Learning as a Service (MLaaS) tools are like Swiss Army knives: versatile and capable of handling multiple tasks but not optimised for any specific one. They often act more as features than standalone products, easily integrated into existing platforms.
Vertical AI enables companies to access niche markets by significantly increasing the value delivered and automating tasks that were previously not feasible to scale. This automation leads to increased efficiency and reduced costs, as seen in examples like EvenUp in the legal sector or Harvey, the AI law assistant, which can perform tasks typically done by associates.
Vertical AI can address both core and supporting workflows within industries.
Core workflows involve primary job functions, such as contract drafting for lawyers or diagnosis in healthcare, while supporting workflows include ancillary tasks, such as marketing for dentists or administrative work in finance. By automating high-cost, repetitive tasks in sectors like legal, healthcare, and finance, Vertical AI has the potential to tap into a broader Total Addressable Market (TAM) that was largely out of reach for legacy Vertical SaaS due to complex, labour-intensive processes.
In contrast to traditional seat-based software, Vertical AI is therefore enabling a fundamental change in cost structures and delivering measurable value in novel ways, translating to a higher willingness to pay from customers, which ultimately translates to a larger baseline TAM.
Companies focusing on Vertical AI can build strong defensibility by honing in on industry-specific nuances and integrating with sector-specific systems, thus creating unique experiences that are difficult for horizontal solutions to replicate.
Rather than directly competing with existing SaaS products, Vertical AI often complements them, enhancing and extending their capabilities without the need to replace them. This synergy allows Vertical AI to not only compete in but also transform entire sectors, offering enhanced functionality and democratising access to services like legal assistance, healthcare, and financial planning.
However, it’s important not to position Vertical AI as inherently superior to Horizontal AI. While some argue that Horizontal AI is less attractive, it’s too early to draw definitive conclusions. Horizontal SaaS, for instance, took off much sooner than Vertical SaaS, with its first IPO a decade ahead. Instead of comparing the two approaches, it’s more constructive to recognise that both Horizontal and Vertical AI have roles to play in the evolving AI landscape.
AI-native vertical applications
AI is rapidly gaining momentum in vertical markets due to its ability to automate tasks and drive innovation, making it an attractive solution for industries seeking efficiency and precision. This trend has spurred the rise of AI-native vertical applications, particularly in sectors like healthcare and law, where automating routine processes can save significant time and resources.
For example, Causaly is an AI-enhanced search platform tailored for biomedical research, optimising data retrieval and analysis. In the legal industry, Harvey, a vertically focused chatbot, assists law firms with tasks such as contract analysis, due diligence, and regulatory compliance.
Both examples illustrate how new entrants are introducing specific solutions that meet needs unmet by existing general-purpose AI tools.
Understanding the push for domain-specific AI solutions
Many markets remain underserved by Vertical AI solutions, particularly foundational industries relying on unstructured data. This includes industries with constrained TAMs, slow sales cycles, or low annual contract values. Unstructured data, which includes raw text, multimedia files, and web pages, presents a challenge due to its complexity and lack of predefined formats.
However, recent advancements in AI, particularly in large language models (LLMs), have improved the ability to handle such unstructured data. Data lakes and data lake houses now provide centralised repositories capable of storing both structured and unstructured data, facilitating the development of AI solutions for these underserved sectors.
The increasing adoption of Generative AI (GenAI) across industries has highlighted the limitations of generic AI solutions and underscored the need for more specialised approaches (as highlighted by our portfolio survey below).
According to a report by McKinsey, GenAI has the potential to add between $2.6 trillion and $4.4 trillion annually to the global economy. Industries such as banking, high-tech, and life sciences stand to benefit the most, with the potential to generate substantial value through enhanced productivity and efficiency.
In sectors where accuracy is paramount, such as healthcare and finance, Vertical AI offers a solution by providing more accurate, industry-specific insights. As Horizontal AI solutions often yield standard outputs lacking in nuance and expertise, Vertical AI’s ability to leverage specialised knowledge makes it a valuable asset.
Practical approaches to implementing Vertical AI effectively
For businesses considering the deployment of Vertical AI, the first step is to clearly define the problem.
Unlike Horizontal AI, which can be applied broadly, Vertical AI must be tailored to fit the specific needs of the organisation and its industry. Organisations should start small, addressing simple, high-impact problems before scaling up, and ensure they understand the technology intimately, including who is developing it and how it works.
By doing so, businesses can harness Vertical AI to fill gaps left by more generalised AI solutions, ultimately adding significant value.
To better understand their use of AI and its impact on their operations, we surveyed our portfolio of 90+ companies about AI adoption, areas of application, challenges faced, in-house development, and future AI plans.
Key survey findings
- 63% of our portfolio companies surveyed strongly agree that using AI within their enterprise is crucial to long-term success.
- 69% of the portfolio companies surveyed have built their own AI tools, either building ChatGPT integrations or building from scratch.
- 96% of our portfolio companies have adopted ChatGPT and Github’s copilot: 50% have adopted Chat GPT, 43% have adopted Github, and 4% have adopted a mix of both.
- GPT’s adaptability across workflows suggests incumbents may be hard to surpass, but vertical AI tools show strong potential, especially in areas requiring proprietary data or domain expertise.
- Moreover, there are opportunities for companies making vertical solutions, as ChatGPT can be harder to integrate across multiple workflows.
- The survey primarily focused on horizontal AI applications. Thus as a result, we found limited adoption of AI applications overall. However, when looking at specific functions like customer service, there is a clearer use of specialised AI solutions tailored to improve efficiency and quality. This indicates the targeted use of AI in certain core functions, reflecting a more specialised approach to AI integration.
- Customer service was by far the largest functional area which adopted AI applications (52% of portfolio companies surveyed have integrated AI tools for customer service) — lending itself to natural language processing and the pattern of automation of low-value work.
A clear opportunity: the customer service vertical
Customer service is the most widely adopted business vertical for AI applications across our portfolio, with companies using vertical AI solutions to enhance and streamline their processes.
With advances in Natural Language Processing, Customer Service has become an obvious pain point that Artificial Intelligence can attack. Horizontal applications use a company knowledge base as well as an LLM such as GPT4, to produce answers to customers that can automate 30–50% of queries (the earlier figure from our survey, the latter from Fin by Intercom). In addition to this, when we look at the main challenges faced by companies in implementing AI, Customer Service applications add evidence to our list of Key Success Factors within an AI application.
Customer service applications do not take skilled personnel to implement, and with its advanced competitive landscape of incumbents and start-ups, integration with existing systems has become a key part of many of the products. Looking at it now, 3 years after Yuma was founded, it is clear that Customer Service was a good domain for AI applications to automate — not only has advances in natural language processing made it easier but it’s clear that customer service is a pain point in enterprises.
Although good customer service can improve the LTV of a customer, it isn’t an obvious source of revenue for a customer and a lot of tasks involved are low-value and time-consuming, making it an obvious target to automate in the same way it was quite often outsourced in the past.
Included below is a summary of some of the most popular tools within this vertical.
Critical factors when selecting and adopting AI tools
The survey highlighted key factors when choosing AI tools. Leading the list, the quality of output is a priority for 90% of respondents, followed by usability (65%) and price (38%). AI requires unpredictable investment for results that are harder to control, unlike traditional tech products where investment and outcomes are more easily anticipated.
The approach to AI budgeting varies widely, but most organisations do not set aside a dedicated budget for it. Instead, AI spending is often absorbed into broader categories such as the tech stack, engineering, or product development budgets. Some companies take a more ROI-driven approach, prioritising AI initiatives that show returns within a year. A few have dedicated R&D budgets covering roles, tools, and consultancy, ensuring compliance and safety in their AI efforts. For others, AI is seen as an integral tool, embedded in existing systems rather than warranting a separate budget category.
Tool reliability is also a key decision factor for 34%, along with user interface, considered by 31%.
The survey revealed that many companies are building their own AI applications rather than buying them. Our portfolio companies, for example, have developed various AI tools, from simple ChatGPT integrations to advanced data analysis from wearables.
Klarna’s rollout of its AI assistant, powered by OpenAI, is a significant example of building in-house tools to redefine the vertical SaaS landscape. In just one month, this AI assistant managed 2.3 million customer interactions, handling two-thirds of all customer service chats, and achieving the workload of 700 full-time agents while maintaining a customer satisfaction level on par with human agents.
While there’s a risk that businesses will build their own AI if purchasing becomes too costly, this is uncommon. Building AI is often too complex and expensive for most companies, so they usually buy solutions unless prices become unreasonable.
Successful AI applications often require deep expertise in specific fields, as exhibited by the founders of Eolas and Synthavo, who are both medical professionals and engineers. This domain knowledge is crucial for creating effective AI tools. Larger enterprises often build their own AI tools to better tailor them to their needs. This trend is supported by data, with Accenture alone securing over $1 billion in generative AI bookings from September 2023 to February 2024.
This brings us to some other key risks:
- Dependence on proprietary data: Exclusive access to non-public data provides defensibility in vertical applications but also makes startups reliant on maintaining that access. Any disruption in data, or loss of domain expertise, could significantly harm the product’s effectiveness and the startup’s viability.
- Regulatory risks: AI applications in sensitive sectors like healthcare, finance, and legal services raise regulatory and ethical concerns. Startups in healthcare must carefully navigate regulations to avoid being classified as medical devices. Governments are increasingly scrutinising AI for data privacy, fairness, and accountability. Additionally, AI-driven job displacement could trigger a public backlash and lead to stricter regulations.
Why AI tools (Vertical or Horizontal) struggle: key implementation challenges
The most prominent issue highlighted by our portfolio is integrating AI with existing systems, as reported by 56% of respondents.
Additionally, 34% cite a lack of skilled personnel, likely referring to the expertise needed to implement and operate these tools effectively. Data privacy and security concerns are raised by 31%, while high implementation costs affect 25%.
Resistance to change (19%), ethical and regulatory challenges (9%), and job displacement (3%) are also noted. The advantage often lies with incumbents who can integrate AI into their established systems more easily, while smaller companies struggle with a shortage of skilled professionals required to leverage these technologies.
Outlook
Over the next 1–3 years, our portfolio companies plan to expand their use of AI, especially in customer service, sales, internal processes, and product development. Key goals include improving efficiency, automating repetitive tasks, and reducing costs. AI is expected to enhance workflows, increase data insights, and improve customer experience.
Strategically, businesses see AI driving faster product iteration, automating complex tasks, and reducing reliance on manual labour. It’s expected to become central to both products and internal operations, with a transformative impact on industries like healthcare and insurance.
Conclusion
Vertical AI offers the strongest potential for sustainable value as companies pursue automation and improved output, while Horizontal AI faces challenges from incumbents and limited defensibility. Successful Vertical AI startups leverage proprietary data, domain expertise, and strong feedback loops, making them more defensible and capable of delivering high profitability.
White Star Capital sees Vertical AI as a major opportunity for industry-specific, high-impact solutions, driven by the growing demand for tailored AI.
As investors, we are particularly excited about the shift from traditional SaaS pricing to models that charge for actual work delivered, like legal AI startup EvenUp, which bills for completed legal summaries. This approach taps into headcount budgets, enabling faster growth than traditional SaaS, which depends on IT budgets. Projections suggest that by 2030, Vertical AI could automate a large portion of tasks at a fraction of the cost of human labour, unlocking substantial market opportunities.
Stay tuned for our upcoming piece, in which we break down the shifting AI infrastructure market, and examine how each segment of the AI value chain interacts within the broader ecosystem. From chip manufacturing and energy infrastructure to application layers and monitoring stages, we’ll guide you through the key players shaping the AI market today.
