GPTs are GPTs: An early look at the labor market impact potential of large language models

The pun in the title of this paper—"GPTs [Generative Pre-trained Transformers] are GPTs [General Purpose Technologies]"—is clever, but the data inside is deeply serious. The authors offer a methodical look at the exposure of various occupations and tasks in the US labor market to large language models. As a founder who builds software tools, this is an incredibly valuable catalog of market demand. It shows that high-wage, information-processing tasks are actually the most exposed to automation, a complete inversion of how previous automation waves impacted the economy.

For our engineering and product teams, this research is a guide to where the value is moving. If writing, summarizing, and basic programming are highly exposed, then building applications that simply do those things will quickly become commoditized. The real value is in building complex orchestration layers that can chain these exposed tasks together to solve actual business problems. It confirms my belief that the next generation of SaaS will not be about helping humans do their work faster, but about automating entire task pipelines end-to-end.

What stuck with me

• Inverted exposure curve: Unlike past technology waves, LLM exposure is concentrated heavily in high-income, highly educated white-collar roles.
• General purpose status: LLMs behave as true general-purpose technologies, meaning their utility spreads across almost every sector of the economy.
• Orchestration layer premium: As individual cognitive tasks are automated, the value of combining these tasks into larger systems rises.