A new survey of tech founders finds that adoption hides a much bigger story about who is willing to restructure work around AI
Listen to this article:
AI will not reorganize your company for you. With increased automation and agentic AI systems inching into the workplace, leaders have a vital role to play in producing AI transformation in their orgs. The new working paper “Delegating to AI: Beliefs and the Organization of Work at New Ventures,” co-written by HBS AI Institute PI Rembrand Koning, explores this issue with a population that should be AI’s likeliest winners: founders of young, growth-oriented North American tech startups. If AI re-organizes work anywhere, it should happen here. Yet the benefits Koning and his co-author found vary dramatically, suggesting that the value of AI is not determined by adoption alone. It depends on how leaders and managers redesign workflows, make delegation decisions, and decide where human-AI collaboration happens.
Key Insight: Inside the Research
“Our sample is well-suited to study early stage firms but caution should be exercised when generalizing to the broader population of startups including those with hundreds of employees, or that have raised Series B financing.” [1]
To study how managerial discretion shapes technology adoption, the researchers surveyed 199 founders of growth-oriented B2B product and IT startups via Pitchbook, with 176 completing the full survey. The firms averaged 16.44 employees, were only 3.68 years old on average, and over half were venture-backed, raising an average of 4.86 million dollars. Within the survey, the researchers ran an experiment focused on customer service: founders were asked to imagine their firms 18 months in the future and estimate what percentage of inbound customer service inquiries AI could handle as well as, or better than, a skilled human. They were then asked how likely they would be to delegate customer service to a human alone, a human working with AI, or AI alone. Founders were also exposed to optimistic or pessimistic research summaries to see how their strategies updated. The researchers are candid about the limits of their survey: the sample is modest, the study itself cannot determine whether integration produced benefits, and they measured stated intentions about hypotheticals rather than lived outcomes. Self-selection is also possible, as recruitment explicitly highlighted AI and hiring challenges.
Key Insight: Where the Biggest Reported Gains Show Up
“Founders who formally integrate AI into their firm are also more likely to perceive AI as changing the management and coordination of their team.” [2]
Among the founders surveyed, 96% reported that AI was used internally either formally (44%) or informally (52%). The authors accept founders’ self-reported designations of formal integration, based on whether founders say that AI is built into internal workflows and systems, rather than merely being available for employees to use. Formal integrators described AI completing work like auto-triaging service emails, generating agendas and action items, assisting engineering teams with debugging and documentation, and accelerating MVP development. When founders were asked how many more people they’d need to match their current output without AI, formal integrators said 91% more headcount, which was nearly four times the 24% estimated by informal users. The pattern repeats when asked about time spent, with formal integrators estimating they would need 82.8% more hours without AI to complete the same work at the same quality, versus 33% for informal users. And formal integrators were also far more likely to report that AI changed whom they hire, with 76% saying it helped them avoid or postpone a hire versus 47% of informal users. Formal integrators were also more likely to hire or build their own AI agents, retrain staff, and lean on AI for team coordination, onboarding, and management advice.
Key Insight: Believers and Integrators Aren’t the Same
“Founders in our sample who integrate and delegate to AI in their firms are not simply technology optimists.” [3]
If formal integrators reported such dramatically better outcomes, you might expect them to simply be more optimistic about AI’s capabilities, but it appears they aren’t. The researchers found a gap of only 1.6 percentage points in performance beliefs between formal and informal integrators. Yet the two groups’ delegation preferences differed sharply: formal integrators were 15.7 percentage points less likely to favor a human working alone, and 8.4 points more likely to favor a human-AI collaboration, with no meaningful difference in preference for AI working entirely alone.
Why This Matters
For business leaders and executives, the new transformation goal is to decide and actually integrate AI into your organization. This means deliberately mapping it into production, using it across functions, and treating human-AI collaboration as an organizational capability. Without doing so, the gap between yourself and lean, tech-savvy startups is only poised to grow larger.
Bonus
This research highlights that AI gains won’t come simply from access to new tools, but from leaders’ choices about how to deeply integrate AI into the organization of work. For a related look at how AI value might accrue to firms with strong complementary assets like data, organizational knowledge, and integration capacity, check out GenAI Advantage May Belong to Firms That Already Have It.
References
[1] Colaiacovo, Innessa, and Rembrand Koning, “Delegating to AI: Beliefs and the Organization of Work at New Ventures,” Harvard Business School Working Paper, No. 26-093 (June 2026): 8.
[2] Colaiacovo and Koning, “Delegating to AI,” 18.
[3] Colaiacovo and Koning, “Delegating to AI,” 4.
Meet the Authors
Innessa Colaiacovo is Assistant Professor of Management at the University of Oregon Lundquist College of Business.
Rembrand Koning is Mary V. and Mark A. Stevens Associate Professor of Business Administration at Harvard Business School, and the co-director and co-founder of the Tech for All lab at the HBS AI Institute.