Studying the Effect of AI Tools (Orbit/JetPack) on the Entrepreneurial Education Process

Each academic year, I am asked to supervise theses for students at MIT. Still, I am quite reluctant to do this because I am not a researcher, and I am much more interested in the practical application of entrepreneurship. I deeply respect good translational research that uses rigorous analysis to evaluate and guide practice.

This academic year, I had a very strong student, Billy Dale, who convinced me that his thesis topic exploring the role AI could play in the entrepreneurial process, including the entrepreneurship education context, was something that not only would be interesting but also could be concrete (considering our significant investment in the Orbit/JetPack tool). In fact, with my interests already percolating in this area, he convinced me to be his thesis advisor and got me truly excited about the work.

Last week, he turned in his thesis, and it was even better than I had hoped for, and I am delighted (with Billy’s permission, of course) to share it with you all now.

My printed-out copy of his thesis has highlights of key points and notes all over it, and I wondered how I would summarize it most effectively for you… And then I realized there was a great way to do that. Use AI as an assistant. So, attached is a summary I did in collaboration with NotebookLM by Google (if you don’t use this, you should). It is another awesome AI tool. I have also attached the full thesis for those who want more details and to see the data.

This thesis captures systematically, comprehensively, and rigorously the developments we had been seeing anecdotally over the past 18 months.

I want to give enormous credit to Billy Dale for pushing for this thesis and then following through. It is a great contribution to all important dialogue now about how to use AI in the field of entrepreneurship. Also, a huge shout-out to Doug Williams, who has been the leader in making the Orbit/JetPack platform what it is today, and who also spent countless hours with Billy getting the data. Props also to Paul Cheek and the other developers of JetPack, including but not exclusively Alfredo Garcia, Navroop Singh Sehmi, and Neelesh Bagga, who have brought Orbit/JetPack into this world and continue to help it mature to what it is today—and will be in the future.

This is also a wonderful example of the power of translational research and why we need a lot more of it in the future.

Now, without further ado, here is the summary of William (Billy) Dale’s MIT Master’s thesis:

The Effect an AI Tool (Orbit/JetPack) Has Had and Can Have on Entrepreneurship Education

Alright, folks, let’s cut to the chase. You’ve got a boatload of ideas, you’re maybe knee-deep in the trenches of building a venture, or you’re out there teaching the next generation how to do the same. You’ve heard about AI, generative AI specifically, and maybe you’re wondering, “Okay, so what’s the real story? How does this stuff help when you’re grinding through building a business?” This thesis, coming right out of MIT’s System Design and Management program, looks squarely at that question, specifically how a custom AI tool called Orbit/JetPack played out in our 15.390 entrepreneurship course, which is built around the Disciplined Entrepreneurship framework.

Now, this isn’t just some ivory tower theorizing. This is about understanding the messy, real-world interaction between people, a proven process (DE), a new technology (AI via Orbit/JetPack), and the learning environment we create. To figure this out, we looked at it like a System-of-Systems (SoS). Think of it like this: you’ve got independent parts – the students and their teams, the 24 steps of the DE framework, the Orbit/JetPack AI tool itself, and the course environment – all doing their own thing, but interacting to create something bigger than the sum of their parts. The big question is, what emerges from that interaction, especially when you throw a powerful new tool like AI into the mix?

Here’s the deal: This research dug into how students used Orbit/JetPack across three semesters, from Spring 2024 to Spring 2025, comparing how things worked with the initial version (v1) and the significantly upgraded version (v2). We looked at hard data – how many users, how many ideas, how many steps completed, how many iterations, how teams worked together – and layered that with feedback from course evaluations.

Orbit/JetPack as Your External Enabler

First big takeaway: Orbit/JetPack isn’t just a fancy chatbot. It acted as what we call an External Enabler. Academic researcher Professor Per Davidsson from Queensland University of Technology talks about how external enablers change the game for entrepreneurs by making things less “opaque” and reducing “agency-intensity”. The DE framework, while awesome, can feel daunting with its 24 steps. It’s appropriately and necessarily a bit complex. Orbit/JetPack helped reduce that opacity, making the steps clearer and more actionable. It also reduced agency-intensity – basically, the sheer effort needed to figure out how to do each step. By generating initial content or providing structure, it gave students a running start.

We saw this play out clearly when Orbit/JetPack evolved to v2. The underlying AI model got smarter (GPT-4o vs. GPT-4-turbo), and the user interface got a major facelift – more intuitive, better layout, easier tracking of changes. What happened? User adoption and activity went up. More people signed up, more people were actively using it month-to-month, and the volume of ideas generated per week increased. This wasn’t just random tinkering; the usage patterns tracked directly with the academic calendar, spiking when assignments were due. It tells you that when the tool got better (Perceived Ease of Use) and was explicitly woven into the course (increased Perceived Usefulness and Trust), people actually used it more effectively for their core coursework.

Transforming How We Learn (and Do)

This isn’t just about efficiency, folks. The research points to something bigger: a Transformation of Learning. Students weren’t just getting through the DE steps faster; they were engaging differently.

How did we see this?

• Deeper Iteration: This is huge. Especially in the v2 semesters, we saw a dramatic increase in iteration, measured by the number of versions students created for a specific DE step. The early, analytical steps like “Market Segmentation” were the champs here, with hundreds of versions created across users. This isn’t just hitting “generate” a few times; it suggests students were using Orbit/JetPack as a workspace to really grapple with and refine foundational concepts. The intensity of iterative work on these early steps was significantly higher than later steps.
• Increased Engagement with the Framework: Across the board, in the Orbit/JetPack v2 semesters, teams completed a significantly higher average number of DE steps compared to the v1 period. This suggests students were more willing and able to tackle more of the entrepreneurial journey within the structured framework, facilitated by the tool.
• Enhanced Course Experience: This is where the rubber meets the road. Course evaluations showed improved overall satisfaction and, importantly, higher ratings for whether learning objectives were met in semesters where Orbit/JetPack was used, especially with v2. This happened alongside an increase in the average number of hours students reported spending on the course outside of class. This isn’t a bug; it’s a feature. It means they weren’t spending more time because they were confused or struggling; they were spending more time because they were engaged more deeply with the material, facilitated by the tool. They were leveraging Orbit/JetPack to explore and validate their ideas more thoroughly.

This transformation is an emergent property of the SoS. It’s the student, the framework, the tool, and the course environment all working together. Orbit/JetPack, integrated into the curriculum, managing cognitive load for tough steps, and boosting students’ belief in their ability to navigate the process (self-efficacy), created an environment where this deeper engagement could flourish.

Navigating the “Jagged Frontier”

Now, let’s be real. AI isn’t magic, and this thesis doesn’t pretend it is. The transformation wasn’t uniform, and that brings us to the concept of the “Jagged Frontier”. This term describes the uneven capabilities of AI – it’s great at some tasks, not so great at others.

We saw this frontier clearly in how students engaged with different DE steps. Orbit/JetPack seemed particularly effective at supporting work on the more analytical steps, like market sizing or defining a beachhead market. These are tasks where AI can help crunch data, provide structure, or generate examples based on patterns. The deep iteration on “Market Segmentation” is a prime example.

However, steps requiring deep creativity, nuanced qualitative judgment, or synthesis of complex, messy real-world information showed less intense tool-supported iteration. It doesn’t mean students weren’t doing that work; it might mean they were doing it elsewhere (offline, team discussions, etc.) or that Orbit/JetPack’s AI wasn’t as effective at supporting that specific type of task at its current stage.

The jagged frontier also impacts different users differently. The analysis showed that students with certain self-reported “personas” engaged with Orbit/JetPack and the DE framework in distinct ways. “Founders with an Idea” averaged significantly more steps per user than “Founders without an Idea”. “Amplifiers” (focused on scaling existing concepts) showed the highest average steps per user, indicating intense individual engagement. “Investors” had the lowest, which makes sense; their goal isn’t necessarily to build out every DE step in the tool. This variability underscores that AI tools aren’t one-size-fits-all. How effectively you can leverage them depends on your starting point, your goals, and likely your own digital literacy and “AI self-efficacy”. You can’t just blindly accept AI output; you need to critically engage with it.

What This Means for You

Okay, so what’s the takeaway for entrepreneurs and educators?

For practicing entrepreneurs:

• Embrace AI as an Enabler: Tools like Orbit/JetPack, built around frameworks like DE, can genuinely reduce the friction of getting started and working through complex analysis. They can make the daunting feel doable.
• Use it for Iteration: Don’t treat these tools as magic answer machines. The real power, as the data shows, is in using them as a workspace for iteration and refinement, especially on those crucial early steps where ambiguity is high. Use it to challenge your assumptions, refine your customer segments, & clarify your value proposition.
• Know the Frontier: Be aware that AI has blind spots35…. It might give you solid analytical outputs, but it won’t replace your judgment, your creativity, or your need to talk to actual customers. Use AI to augment, not automate, the hard parts of entrepreneurship.
• Team Up Smart: The data suggests Orbit/JetPack facilitated more equitable workload distribution in teams in its later version. Leverage these tools collaboratively to make sure everyone is contributing and engaging with the framework.

For entrepreneurship educators:

• Think System-of-Systems: You’re not just adding a tool to your course; you’re integrating a new component into a complex ecosystem. Understand how the tool interacts with your students, your framework, & your overall course design. The success of Orbit/JetPack v2 was tied directly to deeper integration into the curriculum.
• AI is a Transformer, Not Just an Accelerator: The goal isn’t just getting students through the steps faster; it’s about fostering deeper learning and different engagement patterns. Design assignments and activities that leverage AI for iterative work and critical thinking, not just content generation.
• Address the Jagged Frontier Head-On: Teach students how to use AI effectively. Acknowledge its limitations. Guide them on when and how to rely on it and when to rely on their own judgment, creativity, and external validation (like customer interviews). Different students (and personas) will interact differently; consider how to support diverse needs.
• Tool Design Matters: The evolution from Orbit/JetPack v1 to v2 shows that good design (UI, model choice, features like version tracking) has a tangible impact on adoption and the quality of engagement. Push for tools that support iteration, team collaboration, and seamless integration with the entrepreneurial workflow.
• Measure Emergence: Look beyond simple metrics. How is the tool changing how students approach problems? How is it impacting team dynamics? How is it shifting overall learning outcomes beyond what you’d expect from just adding a resource?

Looking Ahead

This is just the beginning. The AI landscape is changing by the minute. Tools like Orbit/JetPack have shown they can be powerful catalysts for entrepreneurial learning, making the DE framework more accessible and fostering deeper engagement. But the “Jagged Frontier” is real, and we need to keep pushing tool design and pedagogical approaches to ensure these tools truly augment human capability, support creativity where AI is weak, and provide equitable benefits for all students.

The path forward involves continuous refinement of the tools, smarter integration into the curriculum, and ongoing research to understand these complex systems. The goal is to harness AI’s power to help more people successfully navigate the challenging, but incredibly rewarding, journey of building something from nothing. Let’s get to work.