AI is moving from experimentation to real deployment inside companies, and that shift is exposing a familiar set of challenges. As organizations try to put AI into production, they're discovering that success depends less on access to powerful models and more on trust, workflow fit, and execution inside real operating environments. We've been watching this play out up close for more than two years, and what we keep seeing is that the enterprises moving fastest aren't the ones with the best access to frontier models. The enterprise companies moving the fastest are the ones working with vertical AI platforms built by people who came from inside the industries they're now rebuilding.
Compressing Vertical Value-Add
This moment mirrors something we've navigated before, just at a pace none of us have seen. The SaaS and cloud computing era took 10 to 12 years to produce vertical SaaS. This didn’t take over a decade because the technology was slow to improve, but because enterprises needed time to trust it and the go-to-market playbooks needed to mature enough for domain-expert founders outside Silicon Valley to run them. AI is compressing that same arc into 2 to 3 years.
After Claude released its Cowork module, AI stopped being an IT initiative and became a board-level strategic imperative across every market sector we track. The startup companies we back aren't solely focused on the SMB or mid-market sectors of their markets anymore. They're now closing seven-figure enterprise deals while putting forward-deployed engineers to work automating the core workflows of some of the largest organizations in the world.
The question we get from LPs isn't whether AI is real. Instead, they’re asking us whether the market is overvalued, and whether a handful of frontier models will capture all the automation opportunities in every industrial vertical. We don't think either framing gets at what's actually happening.
The historical pattern across every major industrial revolution is the same: massive infrastructure investment concentrates value at the foundational layer first, and then that value migrates toward the application layer as domain expertise catches up. Railroad buildout in the 1840s cost roughly $3 trillion in today's dollars. The telecom and internet build-out of the late 1990s came in under a trillion before the markets burst. AI has already hit $2.6 trillion in cumulative investment since 2013, and we aren't done. The market analysts pricing frontier models as if they'll capture all software automation opportunities are making the same mistake they made during the recent SaaS-pocalypse, when they were slow to recognize that value was already shifting and decried that all SaaS investments should be marked down due to AI’s rapid progress. When this happens, keep in mind that value isn't disappearing from AI, it's moving to where domain expertise lives.
Every major release from the frontier models is a horizontal release, going after analyst roles, legal and financial workflows, and general productivity of the end-user. That's real value creation, and we respect what those foundational platforms have built. But when you go deeper into retail and supply chain, healthcare, or energy and industrials, the workflows are too complex, the data too non-normalized and proprietary, and the regulatory environment too specific for a general-purpose tool to handle end-to-end. The frontier models stop at a certain point by design. That's not a weakness. It's a strategic choice that creates the opening for what comes next, which is vertically-focused AI platforms built by founders who spent years operating inside the industries they're now transforming.
The Importance of Founder-Market Fit
We've spent three years refining what we look for in those founders, going back through every Mercury portfolio company across Funds III, IV, and V and being honest about what actually predicted success. We believe the Era of Vertical AI will be defined by entrepreneurs with strong Founder-Market Fit, and it comes down to four things:
- Subject-Matter Expertise: Deep subject-matter expertise built over 4 to 8 years inside a specific industry
- Adaptability: the hustle, grit, and agility to move faster and smarter than the market and your legacy and new-found competitors
- Vision: the ability to sell a compelling vision of the future, not just to customers and investors, but to the ML engineers who could work anywhere and are choosing to bet their careers on an early-stage vertical AI company
- Network: a robust network of customers, partners, and distribution channels creating near-zero customer acquisition cost from day one
We’ve seen these traits - especially a founder’s network - manifest in a few vertical AI founders that we bet on over the past few years. The founders of Mercury portfolio company Pie left Toast with hundreds of restaurant and retail relationships already in hand. One of Collide's co-founders started a millennial-driven community of energy professionals, while the other ran one of the largest PE firms focused on fracking and knew every buyer in the market before Collide wrote a line of code. That kind of distribution isn't something you replicate with outbound sales, and in vertical markets, it's the moat.
That conviction runs through much of Mercury’s vertical AI portfolio. Retail and brand- focused Nectar Social, founded by two former Meta executives, closed a $30M Series A less than 12 months after our Seed investment and have seen 5x ARR growth in the past 3 months. Another vertical AI portfolio company focused on brick-and-mortar SMBs is in the process of closing a $20M Series A led by Lightspeed on similar timing. Energy-focused Collide closed a Seed+ round, a 3x valuation step-up in 12 months, while expanding into the largest operators in the energy sector. Clinical trial-focused OmniScience hit 8 figures in bookings in Q1 2026 alone, and is already working with eight of the top 20 global pharma companies on clinical trial management. These aren't isolated data points. They're early proof of a pattern we expect to play out across every major industry vertical over the next decade.
Our thesis is straightforward: the Vertical Era of AI will be led by founders who are regionally located, industry-trained, and building platforms tailor-made for specific markets. This is the same way vertical SaaS was built - by founders who understood their industries from the inside rather than trying to serve every customer from a distance. Value always migrates from the foundational layer to where domain expertise lives, and that pattern has held across every major technology transition we've ever seen. AI will be no different.
We're backing the founders who earned the right to win in their industries long before the models caught up to them.
