Working Paper 26-093

Delegating to AI:
Beliefs and the Organization of
Work at New Ventures
Innessa Colaiacovo
Rembrand Koning

Delegating to AI:
Beliefs and the Organization of
Work at New Ventures
Innessa Colaiacovo
University of Oregon

Rembrand Koning
Harvard Business School

Working Paper 26-093

Copyright © 2026 by Innessa Colaiacovo and Rembrand Koning.
Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may
not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.
Funding for this research was provided in part by Harvard Business School.

Delegating to AI:
Beliefs and the Organization of Work at New Ventures∗
Innessa Colaiacovo
University of Oregon
innessa@uoregon.edu

Rembrand Koning
Harvard Business School
rem@hbs.edu

June 18, 2026

Abstract
When a technology emerges that may enhance productivity, managers must decide whether and how
to apply it to their firms. We recruit founders of North American tech startups willing to share indepth information about their experiences with generative AI and their perceptions of its impacts on
their ability to scale. We first document substantial heterogeneity among the AI users in our sample:
while founders in our sample estimate that in the absence of AI they would need 55% more employees,
the median is 17% and nearly a third estimate no additional employees would be needed. Founders
who formally integrate AI into their workflows estimate their required headcount increases would be
nearly four times higher than those of informal users, and are more likely to report that AI has changed
hiring and management. We then show that founders can have different preferences for delegating to
AI, holding constant their beliefs about how well it performs a given task relative to humans. We
randomly expose founders to optimistic or pessimistic information about AI performance and show that
even when these task performance beliefs and delegation preferences update on average, substantial
managerial heterogeneity persists in willingness to delegate to AI. These results underscore that, as with
prior transformative technologies, the impact of AI on work and organizations is not only related to what
the technology can do but also what managers choose to do with it.

1

Introduction

A fundamental task for managers is allocating work between employees and technology (Simon, 1965;
Drucker, 1975). When a transformative technology emerges, managers have been critical to translating
technological innovation into economic gains (Bloom et al., 2012; Brynjolfsson et al., 2021a,b; Johnson and
Acemoglu, 2023), including by making necessary changes to work structures and other critical complementary
investments (Bresnahan et al., 1996; McElheran, 2015; Kim et al., 2026).
However, managers and the nuanced choices they make when adopting new technologies have been
absent from many studies of how AI will impact the labor market. Existing work on this emerging class of
technologies generally focuses on where in the economy various AI technologies have been adopted (Acemoglu
et al., 2022; McElheran et al., 2024; Bonney et al., 2024b; Yotzov et al., 2026), the worker-level productivity
impacts conditional on being given an AI tool (Felten et al., 2021; Brynjolfsson et al., 2025b; Cui et al.,
∗ We thank Kate Houston for exceptional research assistance.

Thank you to participants at the Smith Entrepreneurship
Research Conference (University of Maryland) and Hyunjin Kim for helpful comments.

1