How Might AI Help Founders Move Faster—and Still Build Their Judgment? A Call for Collaboration from Entrepreneurship Educators and Accelerators
A call for educators and accelerators looking to test how AI can both speed founders up and strengthen their judgment through structured training, which can be scaled with AI-enabled delivery.
We've known for decades that entrepreneurs taught to make decisions systematically perform better. The uncertainty created by AI makes developing this skill more vital than ever.
The challenge for those that help founders develop - is equipping founders to leverage GenAI's decision-making capabilities without undermining the founder’s development of these same skills?
This is the issue we're going to address through a rigorously designed field experiment. To do it well, we need to build a coalition of those that support founders to collaborate. If that sounds like you, please reach out.
The Current Reality
Leading founders, especially university-aged ones, are ahead of society in adopting AI tools. This lets them generate business plans, pitch decks, and prototypes in hours not years. But what if this speed costs those founders the development of critical business judgement?
A field experiment where entrepreneurs in Kenya were equipped with GenAI found that the high performers improved outcomes, while low performers saw worse outcomes compared to control groups. The key finding: entrepreneurs need to know how to engage AI to benefit from it1.
Findings like this make the work of entrepreneurship educators as dynamic as ever. Helping founders make decisions quickly has never been more important, the tools have never been more widely available, and the unconscious obstacles have never been so prevalent.
What We Know Works
Thousands of entrepreneurship courses have explored the best approaches to decision-making since Harvard launched the field nearly eight decades ago. The best evidence we have says that teaching entrepreneurs a scientific approach to decision-making—forming theories, testing them rigorously, learning from results—significantly improves their performance. This approach is not new, it builds atop methods like the business model canvas, and learn startup with its reliance on hypothesis studies. Importantly though, entrepreneurs trained in the scientific approach out-perform those trained on the business model canvas according to rigorous experimental studies involving 759 firms across four randomized controlled trials2.
Tired of reading the “AI-Slop” that started to appear in my university entrepreneurship capstone, I started teaching the scientific approach myself. It rapidly accelerated how quickly students began to think critically about their business. This led me to double down and have now taught it several semesters in a row. As one student reflected:
“Getting ghosted during our first hypothesis test was the first moment when I realized how little I knew about generating reliable data from tests. Learning effective ways to test theories was a trial by fire, and ultimately gave me a deeper appreciation for what de–risking entailed”
Integrating this rigor helps students think critically. But how do we do that while empowering them to move faster using AI tools?
The Challenge and Opportunity
AI is a super power for entrepreneurs, letting them generate business and financial models near instantly. This will be valuable for experienced entrepreneurs with the judgement skills necessary to evaluate outputs. Those new to entrepreneurship however, may not have built such critical thinking in these areas (or they may have atrophied with AI-dependence). We know that left to their own device, founders will use generative AI to confirm existing beliefs, not to challenge them.
In the broader AI in Education domain, there is reason for hope and for pause. Large scale aggregation of studies have found that AI in education leads to significant increases in learning 3. Systematic reviews are acknowledging however, the potential for student over-reliance on AI to undermine critical thinking4. In response to such concerns, Google is optimizing versions of their AI models for learning and showing positive results5.
The Research Opportunity
Perhaps more than any educator given the adoption of AI by founders, entrepreneurship educators need tools for using AI to improve learning. The research we’ll be conducting responds to that by asking: How can we support founders in moving faster with AI but also moving smarter?
With the help of entrepreneurship educators like yourself, we will run a field experiment to train hundreds of entrepreneurs in the scientific approach, and explore how AI-powered training tools designed specifically to teach can help.
The research is informed by my time on both sides of this equation as a founder and educator, wrapped up in my PhD research. Beyond exploring entrepreneurial decision-making myself and with hundreds of founders, I’ve completed rigorous reviews of academic research on this topic (summarized here), and evaluated the latest tools.
After hitting gaps with off-the-shelf training tools, I spent months building an AI tool specifically for entrepreneurship education. It guides entrepreneurs through hypothesis formation, rigorous testing methods, and structured reflection on results. The goal of the AI is to build judgment, not bypass it.
The Ask: Partner on Exploring AI in Entrepreneurship Education
I'm seeking partners from accelerators and university entrepreneurship programs to pilot this AI-powered training tool. The Research Questions We'll Address Together: How can we structure entrepreneurial engagement with AI tools around to improve decision-making abilities?
Here's what participation involves:
For Your Founders:
Asynchronous training on the scientific approach to decision-making (either as a single sitting, or over a longer period)
Evidence-based skill development proven to improve performance
Structured guidance through real decision-making scenarios
For Your Program:
Direct experience with pedagogically structured AI training
Detailed, anonymized debrief on findings
Influence on the implementation
What’s Next
The case has been made, the ethics approvals secured, and the training tool is being tested by pilot groups of academics and founders. Now we need to identify several hundred entrepreneurs or entrepreneurship students to test this with. That will only be possible with dedicated field partners.
The anonymized, aggregate findings from this research will be presented at the Conference on Field Experiments in San Francisco October 10th. We’ll then continue follow-on testing based on these results.
By participating, you're contributing to rigorous research on how entrepreneurship education can evolve with AI while maintaining its core purpose: developing founders' judgment and decision-making capabilities.
Ready to Explore This Together?
If you manage an accelerator or university entrepreneurship program and want to give your founders access to evidence-based, AI-powered training that builds rather than bypasses their judgment, please reach out for a call, send a message, or sign up to have me reach out. We’re ready to go when you are!
Have questions but don’t need a call?
Luke DeCoste is a PhD candidate in Organizational Psychology, former founder (CDL/Techstars alumni), and entrepreneurship educator researching how AI can empower founders to develop stronger judgment while building faster.
https://www.hbs.edu/faculty/Pages/item.aspx?num=65159
https://hbr.org/2024/07/why-entrepreneurs-should-think-like-scientists
Lu, W., & Lin, C. (2025). Meta-Analysis of Influencing Factors on the Use of Artificial Intelligence in Education. The Asia-Pacific Education Researcher, 34(2), 617–627. https://doi.org/10.1007/s40299-024-00883-w
Tlili, A., Saqer, K., Salha, S., & Huang, R. (2025). Investigating the effect of artificial intelligence in education (AIEd) on learning achievement: A meta-analysis and research synthesis. Information Development, 02666669241304407. https://doi.org/10.1177/02666669241304407
Zhai, C., Wibowo, S., & Li, L. D. (2024). The effects of over-reliance on AI dialogue systems on students’ cognitive abilities: A systematic review. Smart Learning Environments, 11(1), 28. https://doi.org/10.1186/s40561-024-00316-7
LearnLM outperformed other AI models in a recent technical study. (2024, December 19). Google. https://blog.google/feed/learnlm-technical-report/