There Was Never An Easy AI Era, And Investors Are Raising The Bar

By Maor Farid

Back in 2021, simply mentioning artificial intelligence on a pitch deck could secure funding. Small and mid-sized businesses rushed to adopt the technology, purchasing solutions rapidly. I observed companies acquire clients and raise capital, only to watch user engagement plummet afterward. Customer retention dropped significantly.

A large portion of those offerings were superficial: they mirrored traditional software functionality while merely adding artificial intelligence as a gimmick. Others saved minor amounts of time on tasks that held no critical importance to anyone’s job. A twenty percent boost on a trivial task creates an impressive product demonstration, but it guarantees a canceled subscription.

True artificial intelligence successes were never simple to achieve, even during periods when capital flowed freely. The market is currently correcting that imbalance. Businesses providing genuine value are expanding at a rate unmatched by anything I have previously witnessed in enterprise software. Investors are now focused on determining which of these firms can maintain their customer bases — and whether upcoming initial public offerings will highlight the divergence.

What makes an AI business defensible today

Before my co-founder and I wrote a single line of code, we spoke with over 900 mechanical engineers, ranging from entry-level staff to vice presidents. We inquired about how they spent their working hours and which of those hours they would gladly pay to reclaim. Those 900 discussions shaped my perspective on developing an artificial intelligence enterprise.

First: the performance gains must justify the financial layout. Achieving a ten percent time savings on an occasional assignment rarely alters operational methods. Conversely, reducing a mission-critical workflow from weeks to mere minutes, or producing substantial financial savings, certainly does.

Second: industry-specific knowledge. Frontier models are rapidly becoming standard infrastructure. Practically anyone with an API key can access a capable large language model. Commercial success depends on understanding a sector deeply enough to solve challenges that generic applications cannot address.

Third: proprietary context. To use engineering as an example, every firm that manufactures physical products has gathered decades of technical expertise. Much of that information remains trapped inside legacy blueprints or locked within the memories of veteran staff. Foundation models lack this specific data in their training sets. An artificial intelligence solution unable to access this background will struggle to become indispensable to an enterprise. This advantage compounds as the software integrates deeply into client workflows, making substitution expensive.

What this means for funding and IPOs

Upcoming artificial intelligence initial public offerings will subject these operational models to intense examination. Public market investors will scrutinize net revenue retention and gross profit margins to evaluate whether rapid top-line expansion converts into a viable business model. They will also analyze whether individual deployments decrease in support costs as the enterprise scales. The benchmarks they set will shape the criteria private investors apply to earlier-stage ventures.

I already observe this shift within my own fundraising dialogues. Several years ago, reaching $1 million in annual recurring revenue marked a major milestone. In the current landscape, a fresh product can hit $1 million in annual recurring revenue within roughly a year, only to shut down twelve months later. Consequently, annual recurring revenue figures alone no longer provide investors with adequate insight.

Instead, investors continually ask about account expansion. They want to verify whether clients increase their spending following the initial implementation. Account expansion serves as our most reliable indicator that a product has permanently altered an organization’s operations: clients have experienced its practical worth and allocated more internal budget toward it.

Throughout the next twelve to 24 months, I anticipate capital will continue favoring enterprises that pair profound sector expertise with major efficiencies in core business operations, alongside access to customer-specific proprietary knowledge. Such firms might expand via hands-on deployment rather than viral self-service adoption, yet they retain the ability to grow accounts where trust has been established.

Conversely, enterprises whose offerings represent thin interfaces over third-party models—backed by impressive client rosters and retention metrics nobody cares to display—will discover that securing their next funding round is substantially harder than the previous one. I would prefer steady, robust growth over five years rather than spectacular expansion for just five quarters.

---

Maor Farid is the founder and CEO of Leo AI, the first AI for mechanical engineering — a large mechanical model for physical product design. He conducted AI and mechanical engineering research at MIT as a Fulbright postdoctoral fellow and became the youngest Ph.D. graduate in the history of the Technion – Israel Institute of Technology. Farid has built a community of more than 60,000 engineers and supports underserved youth through his nonprofit initiative.

?Frequently Asked Questions

01What criteria do investors now use to evaluate AI startups?

Investors focus heavily on net revenue retention, gross margins, account expansion, and whether a product solves critical, high-value business problems rather than offering superficial features.

02Why is proprietary context important for AI companies?

Proprietary context includes internal company data and specialized domain knowledge that standard foundational models lack. Integrating this data makes the AI tool indispensable and difficult for competitors to replace.

03What is the danger of high ARR (Annual Recurring Revenue) alone in AI startups?

While an AI product can reach $1 million in ARR quickly, high initial revenue can mask poor customer retention if the product fails to deliver long-term value, leading to rapid churn.

Related Crunchbase query:

• Global Venture Funding To AI Startups In 2026

Illustration: Dom Guzman