Entrepreneurs increasingly interact with AI systems such as large language-model–based chatbots for idea development, feedback, and scenario exploration. These systems are often prompted to “take the role” of an entrepreneur.

In our study (published in the Journal of Business Venturing Insights), we examined whether the patterned structure of the entrepreneurial mindset, well documented in human entrepreneurship research, also appears in synthetic responses generated by entrepreneurial AI personas.

Rather than focusing on isolated traits, the entrepreneurial mindset is best understood as a configuration of cognitive, emotional, and behavioral tendencies that coherently co-occur.

Our aim was not to explore how AI systems think (they do not). Instead, we asked whether such human-like structural patterns emerge on validated psychological scales when a model is prompted to behave “as an entrepreneur.”

How we studied it

Using synthetic persona-conditioned responses, we instructed a large language model to adopt an entrepreneurial persona and administered a broad set of validated psychological scales commonly used in human entrepreneurship research. Each scale was delivered in fresh, memory-free sessions and repeated many times to assess stability. Beyond this baseline comparison, we conducted several additional analyses to evaluate the distinctiveness of these patterns.

First, we compared the entrepreneurial persona with a manager persona, which is a standard comparison group in entrepreneurship research. Second, to test whether findings generalize, we compared the entrepreneurial persona with a range of occupational personas such as farmer, scientist, artist, teacher, and accountant. Third, we evaluated a widely circulated entrepreneurial advisory agent presented by Sam Altman from OpenAI in a public demonstration (“Startup Mentor”). This allowed us to examine whether entrepreneurial structure emerges only through persona prompting or also in purpose-built advisory agents. Finally, we compared the entrepreneurial persona with the model’s unprompted, default persona to identify baseline patterns.

Throughout, our analysis focused solely on output structure, not on internal cognition or psychological states.

What we found

Across repeated tests, the synthetic responses produced under an entrepreneurial persona prompt aligned with both individual trait levels and the higher-order structural constellation typical of human entrepreneurs.

For example, the model reproduced elevated opportunity alertness, proactive tendencies, creative engagement, balanced risk-taking, and domain-specific fear of failure, patterns well established in human studies. Traits and subcomponents that covary in humans also tended to covary in the synthetic profiles, suggesting that the model draws on structured entrepreneurial knowledge present in its training data.

The comparison with other personas confirmed the distinctiveness of these patterns. The entrepreneurial persona differed markedly from other personas such as manager, scientist, teacher, and accountant on core mindset dimensions.

The Altman “Startup Mentor” agent showed weaker entrepreneurial mindset structure than the entrepreneurial persona, indicating that simple persona prompting more effectively induced a distinct entrepreneurial profile.

Across all comparisons, the entrepreneurial persona displayed a coherent, human-like mindset structure, even though this does not imply genuine (human-like) cognition.

Why it matters

For practitioners, the key insight is that AI systems can reflect not just single tendencies but structural aspects of the entrepreneurial mindset when used in role-play mode. This makes AI a potentially useful tool for structured reflection, scenario exploration, and assumption testing from different entrepreneurial vantage points.

For educators, the structural consistency of these outputs creates opportunities for scalable simulations that expose students to recognizable configurations of entrepreneurial judgment.

It is essential to clarify what these findings do not imply. A structural match in synthetic self-reports does not mean that an entrepreneurial AI persona will act like an entrepreneur. Our study measured only patterned self-reports, not behavior. Whether AI personas can translate these simulated structures into entrepreneurial reasoning, decisions, or actions remains unknown. At this stage, practitioners should therefore view AI-enabled role-play as a tool for perspective-taking, not as an entrepreneurial actor or decision-maker. Testing mindset–behavior consistency in AI is an important next step for research.

Moreover, the structural patterns reproduced by AI originate from entrepreneurial discourse present in the training data. They may reflect cultural narratives and potential biases embedded in that discourse. As a result, these synthetic patterns can amplify familiar representations of entrepreneurs that circulate within particular cultural contexts.

What practitioners can do

Practitioners can use AI systems to explore different viewpoints and frames. When preparing pitches, evaluating opportunities, or weighing strategic options, founders can prompt the system to adopt various entrepreneurial personas, such as conservative, growth-oriented, or mission-driven founders, and compare resulting perspectives.

Educators can use similar exercises to help students identify structural aspects of entrepreneurial judgment and reflect on their own assumptions.

To ensure responsible use, practitioners and educators should treat these structured outputs as informative but tentative and remain aware of the cultural biases embedded in entrepreneurial discourse.

Read the full paper here: https://www.sciencedirect.com/science/article/pii/S235267342500068X

---

Author bio

Martin Obschonka is Professor of Entrepreneurship and Section Head of the Entrepreneurship and Innovation section in the Amsterdam Business School, University of Amsterdam. With his focus on person-environment interactions and transaction, his research aims at advancing the knowledge of psychological, economic, and technological factors and mechanisms relevant for human agency in context, with a special focus on entrepreneurship and innovation. 

Christian Fisch is an Associate Professor of Business Economics and Entrepreneurship at the Interdisciplinary Centre for Security, Reliability and Trust (SnT), University of Luxembourg. He founded and heads SnT’s Entrepreneurship, Innovation and New Technology (EINT) research group. Christian’s research examines entrepreneurial finance and innovation strategy, with a strong focus on new technologies (e.g. AI and blockchain), sustainability- and climate-related entrepreneurship, and space entrepreneurship.