Artificial intelligence is helping to level the playing field between startups and incumbents, and this is being fueled by a new class of organizations: AI-driven companies that leverage AI across all functional business areas from day one.
“AI-driven companies achieve incredible results with far fewer resources and in far less time,” said Paul Cheek, a senior lecturer at the MIT Sloan School of Management, in a presentation at the recent MIT Research and Development Conference. “Companies are becoming more efficient, entering the market, and taking market share from incumbent organizations faster than ever before.”
Organizations that implement AI across all functional areas from the beginning can develop products with fewer employees and less capital expenditure, and win customers at lightning speed. While it took Netflix years to reach 1 million users, Cheek said it took ChatGPT just five days.
“We’ve seen the product development lifecycle shrink significantly,” said Cheek, founder of the AI-Driven Enterprise Institute and senior advisor to MIT Entrepreneurship’s Martin Trust Center. “If you look back 50 years ago, product development that used to take years has been reduced to days, hours, minutes, and even milliseconds.”
It’s no longer just a product development process, he said, but the process of creating entirely new ventures is accelerating.
The 2026 AIDE Index, based on a framework developed by Cheek and AIDE Institute co-founders Felipe Scherer and Kate Reid, ranks Nvidia, Amazon, Meta, and energy producer SLB highest in terms of the most AI-mature companies in the S&P 500. Track objective signals of AI adoption using publicly observable data such as LinkedIn profiles and posts, earnings reports, job postings, patent applications, regulatory filings, and corporate communications.
Cheek explores the AI-first phenomenon in his forthcoming book, No One Works Here: How AI-Driven Enterprises Are Dramatically Redefining Business, Leadership, and Competition.
As AIDE begins to reshape its entrepreneurship and business more broadly, Cheek believes its competitors need to understand.
ARR to FTE is the new business metric to pursue
Historically, startup success has often been measured by the number of employees hired and the amount of investor funding secured. That has changed, Cheek said. Many entrepreneurs and investors measure the success of AI-driven companies and AI-native startups by looking at annual recurring revenue per full-time equivalent employee. This provides insight into the productivity of a company’s workforce.
A high ARR per FTE suggests that a company is efficient and can generate more revenue per person. Established companies or companies in industries with business models that don’t necessarily rely on recurring revenue can replace that metric with revenue per employee.
The ARR per employee numbers show a clear difference between declining unicorns and rapidly growing AI-native companies, Cheek said. He pointed out:
- Bench is a bookkeeping and tax advisory fintech company that had an ARR of just $23,000 per FTE before filing for bankruptcy in January 2025.
- Bolt.new, a browser-based AI app builder, had $1.3 million in ARR per FTE as of March 2025, five months after its founding.
“This is an efficient business,” Cheek said of Bolt. “This is a business that is growing rapidly and experiencing significant revenue increases per FTE.”
In this scenario, Cheek said, founders whose businesses initially looked strong based on traditional metrics are increasingly caught off guard. Even those with recent seed and Series A rounds with “tremendous metrics” are finding that the hurdles are already in motion when they want to raise additional capital.
“The goalposts have moved since we brought the last round of funding into the company. Investors are seeing much more efficient companies in the market and would rather put their money there,” Cheek said.
Lead an AI-driven organization
Face-to-face at MIT Sloan
Register now
Your teammate, direct report, or manager could be an AI agent
While most people still think of a “company” as a group of people working together, AIDE is best described as a network of nodes, some humans and some AI agents.
“I’m not saying we should eliminate people, but we need to redefine the word ‘organization’ to reflect what organizations really are today,” Cheek said, noting that in the real world there are already AI agents running companies that sell real products to real people.
“We’ve now sold a product that is profitable on its first AI agent-to-human sales unit. This is really important as we think about what happens next as companies continue to scale and how AI companies come into the market,” he said.
People are often the most inefficient part of an organization, he said, and technology can make up for that shortcoming. For example, an organization that includes AI agents in addition to its human population can easily operate around the clock, Cheek said, painting two scenarios for the audience.
- Human CEOs wake up every morning, take in the previous day’s data reports, reprioritize what they believe will move the organization forward, and delegate to their teams.
- An AI CEO who continuously ingests, reprioritizes, and delegates data to other AI agents “every hour, every minute, every 30 milliseconds, all night long.”
“Who would you bet on?” Cheek asked.
AIDE shows economic promise but is also dangerous for existing competitors
AIDE offers significant benefits to founders. Smaller teams have lower overhead costs and lower labor costs. Such companies also require much less investment capital to start up and scale, giving entrepreneurs more control over what they build.
“We don’t see these entrepreneurs giving up as much equity in their companies. They’re retaining more ownership, which is really exciting and good for entrepreneurs,” Cheek said.
As an example, Cheek cited an update for investors on a startup that has reinvented itself around AI-driven efficiency. Within three years, the company was able to increase revenue while reducing burn rate by more than 85%.
Collectively, such improvements could have broader implications for competitors and the economy.
A decentralized economy with thousands of AIDEs poses less systemic risk than relying on a few large employers or sectors, he said. When large, innovation-driven companies lose market share, jobs disappear. However, thousands of small and efficient AIDEs provide stability.
“We always think that innovation-driven companies are good for the economy because they create more jobs, but in reality [situation] “It also creates risk in the economy. If we had thousands of these AI-driven companies, we would have a more decentralized economy with less risk. That’s good for all of us.”
A new skill set for managers
Given that individual managers have the ability to manage far more agents than humans, Cheek believes that over the next 20 years the workforce will transition to one that consists of far more agents than humans.
While this change won’t necessarily result in long-term, widespread job losses, leaders will need a different skill set to manage agent-dominated workforces, Cheek said.
See: AI-driven enterprises — the new arithmetic of exponential growth
paul cheek Senior Lecturer and Senior Advisor in Entrepreneurship and Artificial Intelligence at MIT Sloan School of Management. MIT Martin Trust Center for EntrepreneurshipCo-founder and Executive Director of AI Driven Enterprise Instituteis the founder of entomology. He is the author of “.no one works here” and “Disciplined Entrepreneurship: Startup Tactics” and received the MIT Monothon Award for its impact on entrepreneurship education.
