The Emerging Vertical AI Landscape, And Our Vertical AI Market Map

By Aileen Lee and Matt Lu, Cowboy Ventures

At Cowboy Ventures, we’re excited about a coming wave of “vertical AI” software companies.

Historically, ‘vertical software’ — aka software purpose-built for a specific industry like construction, home healthcare or supply chain — has attracted fewer venture dollars and delivered relatively few $Bn+ sized outcomes relative to ‘horizontal’ software. Challenges have included smaller addressable markets (TAM), difficult customer acquisition (CAC), long sales cycles and lower price points (ACVs).

But a new window has been opened for vertical software — by AI. Horizontal products like ChatGPT and Midjourney have given millions of users a taste of the power of AI, but enterprises have barely gotten started (gen AI is < 1% of cloud spend, as noted in Menlo Ventures’ recent report). The time is ripe for founders to build AI-native software that delivers new capability to customers, leveraging models built on industry-specific data. We’re excited to see many more Toast, Procore, and Veeva-like outcomes with teams that build with AI and understand the nuances of a specific industry.

At Cowboy, we’ve been investing at pre-seed and seed in vertical software for many years (we enjoy bucking conventional wisdom about ‘limited’ market opportunities). Companies like Arcol (AECD), Homebase (brick + mortar SMB), Portex (logistics), SVT Robotics (warehouse logistics), and Uno (healthcare payers) are examples of high ROI software based on industry-specific pain points, and savvy navigation of historical vertical software challenges.

Why will customers choose vertical AI over general purpose AI?

For an enterprise to adopt new software (especially in the current economic environment), there’s a high bar for accuracy, time to value, ROI, security, support and ease of adoption, among other factors. Vertical software has the advantage of being purpose-built to meet those needs, designed with context and specific ROI in mind.

Building software on smaller, focused models with industry-specific data, documentation, terminology, processes and edge cases should deliver performance with lower cost than tuning a large foundation model. And focused scope makes it easier to achieve the accuracy required for customer adoption, plus can build a moat relative to more generalist competitors.

For example, Cowboy portfolio co Vic.ai has been building proprietary models trained on hard-to-obtain accounting data for 6 years. Their UX is tailored specifically for accountant and bookkeeper workflows; and it’s been imperative to prove high accuracy (over 99%) and ROI to gain trust to run mission critical business operations.

It will get noisy in vertical AI — startups should be prepared

With growing interest in all things AI and the lower barrier to entry in vertical AI (vs. needing $Bns for foundational models), there will likely be lots of funding and competition from both incumbents & startups. This is where access to data, strategic GTM, speed and quality of execution will matter. Choose your target business use case and ICP (ideal customer profile) carefully, aim to leverage proprietary data, and train your model with a focused dataset with high accuracy, ease of adoption and time to value in mind. Strategically onboard and nurture early customers, proving clear and lasting ROI, and build your company’s GTM muscle all along the way.

And, be thoughtful about your financing and burn rate strategy. Vertical software has historically had fewer outsized outcomes and has naturally smaller TAMs than horizontal software, so companies should be sensitive to capital efficiency. Companies are learning hard lessons right now from sins of the past — we look forward to helping teams avoid those pitfalls in the years to come.

We’re just at the starting line for a new generation of AI-native vertical software. Below, we share an early map of verticals and companies in each. More will emerge in the coming years — we’re tracking manufacturing, logistics, and hospitality as additional verticals, to name a few. We’ll update our map regularly to reflect the rapidly evolving landscape, and we welcome your suggestions of verticals and companies we missed.

Cowboy’s Vertical AI Market Map

Our market map includes companies using AI to create new user experiences and efficiencies in different ways. A few examples:

Automating work. Assort Health is in the healthcare vertical — their voice AI automates call center operations for healthcare providers, replacing costly, labor-intensive services, alleviating staff burnout and unlocking fast time-to-value for customers.

Similarly, Vic.ai automates 80% of invoice processing work for accounting firms. They can replace a formerly human-powered, error-prone 10 minute process per invoice with a higher accuracy, 45 second process using their proprietary model. This saved one customer over 60,000 hours in their first year, lowering costs by over 90%.

AI-powered search & retrieval. Harvey and Hebbia are two standouts. The former builds software for law firms, and the latter for financial institutions. Both verticals have largely untapped datasets that can be vectorized and parsed with LLMs — replacing low-level knowledge work and unleashing productivity.

Content generation. Media & entertainment, as well as gaming, have been quick to adopt AI tools to create beautiful content. In e-commerce, companies spend millions on expensive photoshoots for new merchandise. Startups like Creative Force leverage AI to power content creation so that brands can quickly iterate on campaigns and test at a fraction of the cost and effort.

At Cowboy, we’re excited to back vertical AI companies from the beginning of their journeys. Successful vertical AI teams will blend AI chops with specific domain expertise — including relationships in the industry or firsthand familiarity with pain points. While there have been many AI breakthroughs this year, the space is still early and evolving — we’re excited to work with agile founders who will iterate quickly to unlock value for customers.

Just as cloud ushered in vertical software standouts, we’re excited for the era of AI to usher in vertical AI standouts. If you’re a founder building a vertical AI company, we’d love to meet you — find us at hello@cowboy.vc, or on twitter at @cowboyVC, @aileenlee and @matt_lu_!

(P.S. Building in AI also requires a new infrastructure stack. Check out our thoughts and market map on the New Gen AI Infrastructure Stack here).

(Like this post and want more like it? Sign up to follow on Medium here)