Middle market AI companies could have an edge, FTV’s Bernstein says

Middle market AI companies may be better positioned than many giant software incumbents or brand-new AI startups, according to Brad Bernstein, managing partner at FTV Capital. For retail investors watching every company add “AI” to its pitch, the argument is a useful filter: the winners may be firms that already own specific customer workflows and can add automation without breaking trust.

Bernstein says company size alone does not decide who benefits from artificial intelligence. AI can automate tasks, analyze data and support software agents, which are programs that can complete parts of a workflow with limited human input. Getting those systems into real businesses, however, requires customer knowledge, clean execution and pricing models that match the value created.

Why could middle market AI companies win?

Bernstein argues that mid-sized technology companies can combine speed with experience. In his view, they often have enough scale, proprietary data and customer relationships to make AI useful, while carrying less organizational drag than large legacy software companies.

He points to risks on both ends of the market. Klarna, valued at $6 billion in 2024, drew attention after saying its OpenAI-powered chatbot could handle millions of conversations and perform work equal to 700 customer-service employees, Bernstein wrote. He also cited reports that customers disliked the rollout and that Klarna was hiring humans back by 2025. Jasper, an early generative AI company, saw its valuation fall after ChatGPT made its core offerings easier to replicate, according to Bernstein.

The broader payoff is also uneven. PwC has said three-quarters of AI’s economic gains are being captured by 20% of companies, a figure Bernstein uses to argue that standing still is already costly for operators.

What traits does Bernstein say matter?

Bernstein lists five qualities he believes strong middle-market technology companies need in the AI cycle:

• Disciplined self-assessment, meaning boards and leaders test where AI creates real value instead of funding experiments that only absorb engineering time.

• Agility, or the ability to run practical experiments without layers of approval common at larger companies.

• Workflow ownership, which means the software sits inside a customer’s daily process and holds data that an AI agent needs to use.

• Technical capacity, including teams that know how customers actually use the product in production.

• A stronger balance sheet, giving a company room to fund AI work, absorb mistakes and pursue selective acquisitions.

One pricing example is Intercom. Bernstein notes that in 2023 the customer-service software company priced its Fin AI agent at 99 cents per resolved conversation, an outcome-based model. Outcome-based pricing charges for a completed result, rather than for each software seat or user. Bernstein says Fin became Intercom’s core offering and that the company was later sold to Salesforce for $3.6 billion.

He also cites Toast, where product teams used AI to reduce documentation and process work, and then connected product updates with external communications using large language models, or LLMs, which are AI systems trained to generate and edit text.

FTV portfolio companies appear in Bernstein’s examples as well. He says Agiloft’s contract software gains value by learning the approval, negotiation and redline history around agreements. ReliaQuest, founded in 2007 as a service-heavy cybersecurity business, has operated in more than 1,000 customer environments and shifted some in-house security operations center analysts into higher-value product development roles as automation advanced, according to Bernstein.

For investors, Bernstein’s case frames AI adoption as an execution test rather than a label. The companies he describes need defensible workflows, customer trust and enough capital to keep building while the technology changes.

This story draws on original reporting from Crunchbase News.