AI's biggest winners aren't giants—they're scrappy middle-market tech firms

Brad Bernstein, managing partner at FTV Capital, frames the AI era as belonging not to the largest technology incumbents or the most-hyped startups but to mid-market software companies—the "middleweights" of enterprise technology. He opens with a boxing analogy: while fighters like Muhammad Ali and Mike Tyson are the most famous heavyweights, Sugar Ray Robinson, a middleweight, defeated the heavier Jake LaMotta in a famous 1951 match with a 13th-round TKO by virtue of superior speed, footwork, intelligence, and stamina. Robinson's advantage was completeness, not size.

Bernstein applies this logic to software companies. The market assumes that big companies will capture the most value from AI, but recent examples suggest otherwise. Klarna, valued at $6 billion（約9600億円） in 2024, announced an OpenAI-powered chatbot that it claimed could handle millions of conversations and replace 700 customer service employees. Customers rejected the rollout, and by 2025 Klarna was hiring back human workers. Jasper, an AI-native startup, watched its value evaporate after ChatGPT commoditized its offerings. These failures underscore that AI is difficult to implement correctly at any size.

The question Bernstein poses is direct: who wins when everyone can get AI wrong? His answer is that the biggest long-term gains will flow to "scrappy middle-market technology companies"—firms with proven growth, customer trust, domain expertise, capital structure, and speed, but without the organizational mass and technical debt of legacy enterprises. Horizontal platforms built by hyperscalers like Microsoft Copilot and Salesforce struggle with the messy, regulation-heavy, category-specific workflows of the real world. Middleweights win by making their software the system of record that AI calls into, rather than software that AI replaces.

Bernstein identifies five traits shared by winning middleweights. First is disciplined self-assessment: these companies test honestly where AI generates value versus where it merely consumes engineering capacity and budget. Seat-based pricing is one area for brutal reassessment; when autonomous agents do the work, revenue models should reflect outcomes, not users. Intercom exemplified this in 2023 when it priced its AI agent Fin at 99 cents per resolved conversation. Fin became the company's core offering, and Intercom recently sold to Salesforce for $3.6 billion（約5800億円）. Second is agility: enterprise companies are weighed down by technical debt and legacy infrastructure, while middleweights—growing 20% or more with strong unit economics and a tech-first mindset throughout the business—can experiment with minimal approval overhead. Toast, a restaurant software company, showed this by having product leads use AI to cut documentation work, then building a flywheel connecting new features to external communications with LLMs continuously editing instructions for AI agents.

Third is workflow ownership: the strongest moat in the AI era is owning a complex workflow. Middleweights have spent years integrating into customer systems, accumulating specialized data, and learning operational nuances that differentiate performance—moving a claims process from 95% accurate to 99.5%, where the last 4.5 percentage points are the true competitive moat. Agiloft, an FTV Capital portfolio company, illustrates this by absorbing the entire decision workflow around contract management, learning why internal teams made specific decisions through approvals, negotiations, and redlines. Fourth is technical capacity: most large companies are stuck in AI pilot purgatory, and the market underestimates the operational demands of AI deployment. Middleweights like ReliaQuest, founded in 2007 as a service-heavy cybersecurity business, have spent years operating in more than 1,000 customer environments—including large global enterprises—learning deep detection logic that it can now encode into AI and move SOC analysts into higher-value roles. Fifth is a well-capitalized balance sheet: companies with clean balance sheets can move faster, absorb experimentation costs, pursue selective acquisitions, and invest through disruption, whereas legacy software firms carrying heavy leverage and optimized for 5–10% growth cannot reallocate capital aggressively enough into AI R&D.

Bernstein concludes by emphasizing the time sensitivity of this window. Per PwC data, three-quarters of AI's economic gains are now being captured by just 20% of companies. Speed is a middleweight leader's advantage; waiting to perfect an AI strategy before executing is a losing move. Companies should identify key workflows, map them, and assess whether AI makes them more defensible or more exposed. That answer may determine whether they give up a round or win the match.

The article challenges a widespread assumption in business and venture capital: that size and resources guarantee dominance in AI adoption. The author uses boxing as a metaphor—citing Sugar Ray Robinson's middleweight triumph over a heavier Jake LaMotta—to argue that organizational completeness beats raw scale. This framing directly refutes the assumption that hyperscalers (large cloud providers) and well-funded startups will capture AI's upside simply because they can spend more money and hire more engineers.

The evidence the author presents centers on failure patterns among both camps. Large enterprises like Klarna, despite a $6 billion（約9600億円） valuation and access to OpenAI's technology, could not execute an AI customer-service rollout that customers accepted—forcing a retreat to human employees. Similarly, Jasper, an AI-native company, saw its value collapse after ChatGPT commoditized its core product. These failures suggest that AI is genuinely difficult to implement well, regardless of firm size or funding, and that advantage flows not from capital alone but from how a company deploys it.

The author identifies five traits that separate winning middleweights from the rest: disciplined self-assessment (ruthless honesty about where AI creates value versus where it merely consumes budget), agility (20%+ growth with strong unit economics and minimal approval layers), workflow ownership (the accumulation of specialized data and operational knowledge that creates a sustainable moat), technical capacity rooted in years of customer deployments, and financial flexibility (a clean balance sheet that allows aggressive reallocation to R&D). Intercom's shift to outcome-based pricing—charging 99 cents per resolved conversation rather than per seat—exemplifies this thesis: the company recognized the business model transition AI would force and moved before competitors did, positioning itself for acquisition by Salesforce for $3.6 billion（約5800億円）.

Critically, the author notes that with three-quarters of AI's economic gains concentrated in just 20% of companies per PwC data, the window for transformation is finite. This creates urgency for middle-market companies to act decisively rather than wait for a perfect AI strategy. The implied challenge to established large software firms is that their leverage, cost-cutting culture, and 5–10% growth orientation are structural disadvantages in an era that demands speed and experimental freedom.

No comments yet. Be the first to share your thoughts!