Weekly new business applications in the U.S. are climbing at a pace that has caught economists' attention. Torsten Slok, chief economist at Apollo, published an analysis on March 7 arguing that AI and large language models are the most likely driver behind the spike. His core claim: AI has gotten so good at handling the grunt work of launching a company that the barrier to starting one has dropped significantly.
The argument is straightforward. Tasks that used to require hiring a contractor or spending weeks learning a new skill - building a website, drafting legal documents, writing marketing copy, setting up basic accounting - can now be handled by a solo founder with a ChatGPT or Claude subscription. When the cost of starting drops, more people start.
U.S. Census Bureau data backs up the trend line. Business formation has been elevated since the pandemic-era boom of 2020-2021, but recent weekly numbers show another leg up rather than the normalization economists expected. Slok's chart, sourced from Census Bureau and Macrobond data, shows the current spike pushing well above the post-pandemic baseline.
The Job Creation Angle
Slok makes a second, more contested claim: that this wave of AI-enabled startups will be a net job creator. The reasoning is that more businesses mean more hiring, even if each individual company runs leaner than it would have five years ago. It is an optimistic read. A founder who uses AI to do the work of three employees is, by definition, not hiring those three people. Whether the sheer volume of new businesses offsets that per-company efficiency gain is an open question.
What the Data Does and Does Not Show
The honest caveat here is that "likely driven by AI" is doing a lot of work in that headline. Business formation rates correlate with many things - interest rate expectations, stock market performance, immigration trends, regulatory changes. Slok does not isolate AI as a variable so much as point to it as the most plausible explanation for the current deviation from trend.
That said, the anecdotal evidence lines up. Talk to anyone running a small business launched in the last year and the conversation almost inevitably turns to which AI tools they use. The friction of starting something has genuinely dropped. Whether that produces durable businesses or a wave of short-lived LLM-wrapper side projects is the question nobody can answer yet.