The Hiring at Top Startups Study (2025)

The 2025 HATS Survey includes questions designed to explore the relationship between entrepreneurs' beliefs about GenAI's capabilities and their anticipated willingness to delegate work to GenAI. Recent scholarship about GenAI and the future of work typically takes a task-based view of work to extrapolate to GenAI’s possible labor market impacts. It is common to model occupations as bundles of tasks and to theorize about a technology's possible labor market impacts by identifying which tasks it performs well and labeling occupations with large shares of those tasks as highly vulnerable to replacement (Autor et al, 2003; Eloundou et al., 2024). However, the belief a technology performs a given task well is distinct from (though likely related to) the belief that it should be deployed in a real organization. HATS 2025 includes questions to explore whether entrepreneurs are optimistic or pessimistic about GenAI's performance relative to humans, and whether and how this connects to how they hope to use it at their firms.

2025-04-28

2025-05-30

The key outcomes are entrepreneurs’ beliefs in GenAI’s capabilities and their delegation preferences conditional on those beliefs (see survey questions under "experimental design"). We will explore three dependent variables: beliefs, delegation preferences, delegation preferences after controlling for changes in beliefs.

We will also explore heterogeneity based on: the firm's age, size (number of employees), and whether it already handles customer service inquiries and has existing customer service staff.

The Hiring at Top Startups (HATS) Study is an annual survey of entrepreneurs at growth-oriented firms that explores their hiring and compensation-setting processes.

The 2025 HATS Study is particularly focused on understanding whether and how the availability of Generative AI technologies is impacting labor demand at top startups.

Entrepreneurs will separately be asked about their perceptions of GenAI’s evolving ability to handle an inbound customer service inquiry, and their willingness to delegate handling customer service at their own firm to a human, a human working with GenAI, or to GenAI.

They will be asked:
•In what percentage of cases do you believe GenAI will be able to handle inbound customer service inquiries at your firm as well as or better than a skilled human? (Beliefs)
•Imagine your firm in eighteen months. On a scale of 0% (never) to 100% (certainly),
how likely are you to delegate handling a customer service inquiry at your organization to [a human alone? A human working with GenAI? GenAI alone?] (Delegation Preferences)

Participants are then randomly assigned to one of two true sets of information about GenAI’s current capabilities: one optimistic and one pessimistic about its abilities. They are then given the chance to revise their answers to the questions above. Finally, they are asked an open-text response question about how they plan to use GenAI at their firms (if at all) over the next 18 months.

Randomization will be done by Qualtrics.

The unit of randomization is the individual survey respondent.

No

N/A

We will attempt to contact approximately 10,000 individuals. A response rate of 5% would produce 500 respondents.

Respondents will be assigned to two treatment arms (approximately 250 per arm).