💰 quantitative finance

Generative AI Fuels Solo Entrepreneurship, but Teams Still Lead at the Top

While generative AI has significantly lowered barriers to entry for solo entrepreneurs, leading to a surge in experimental product launches, team-based ventures continue to dominate the highest-quality outcomes.

Original authors: Hyunso Kim, Hyo Kang, Jaeyong Song

Published 2026-05-12

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Hyunso Kim, Hyo Kang, Jaeyong Song

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: A New Tool, A New Crowd, Same Old Winners

Imagine the world of starting a business is like a massive, crowded amusement park. For years, building a ride required a big crew of engineers, architects, and designers working together. It was hard, expensive, and only teams could do it.

Then, in late 2022, a magical new tool called Generative AI (like ChatGPT) was handed out to everyone. This tool is like a "super-powerful 3D printer" that can instantly draft blueprints, write code, and design interfaces.

This paper asks: Did this magic tool change who gets to build rides, and did it change who builds the best rides?

The answer is a mix of "Yes" and "No."

1. The Solo Adventurers Flocked In (The "Quantity" Boom)

Before the magic tool arrived, building a new product usually required a team. But once the tool was released, solo entrepreneurs (people working alone) rushed in.

The Analogy: Imagine a "Lego Castle" contest. Before the magic tool, you needed a whole family to build a castle because it took too long for one person. Suddenly, a robot arm appears that can snap bricks together instantly. Now, a single person can build a castle in an afternoon.
What the paper found: The number of solo entrepreneurs skyrocketed. In fact, in categories that used to be dominated by teams (like software and finance), the number of solo founders grew the most. The tool lowered the barrier to entry so much that anyone could try.

2. The "One-and-Done" Phenomenon (The "Experimentation" Boom)

However, the paper noticed something interesting about these new solo builders. Many of them were experimenting, not necessarily building a permanent business.

The Analogy: Think of these solo founders as people trying out different flavors of ice cream. With the magic tool, they can quickly scoop up a weird flavor (like "pizza-flavored ice cream"), taste it, and if they don't like it, they throw it away and try the next one.
What the paper found: A huge chunk of these new solo ventures were "one-shot" entries. They launched a product, got no updates, and disappeared within a year. They were testing wild, unique combinations of ideas. While teams also experimented, solo founders did it much more aggressively. They were exploring a wider, stranger territory of ideas, but many of these ideas were short-lived.

3. The Teams Still Rule the "Hall of Fame" (The "Quality" Reality)

Here is the twist: Even though solo founders flooded the park and tried thousands of new ideas, teams still won the top prizes.

The Analogy: Imagine the amusement park has a "Hall of Fame" for the best, most thrilling rides. Even though the solo adventurers built more rides, the rides that made it to the Hall of Fame were almost exclusively built by teams.
What the paper found: When looking at the highest-ranked products on the platform (the "Top 10"), teams actually increased their share of the winners after the AI tool arrived. Solo founders dominated the middle and bottom of the rankings (lots of experiments), but they didn't dominate the top.
Why? The paper suggests that while AI can do the "heavy lifting" of coding and design, the best products still need the "human spark" that comes from a group of people with different skills, experiences, and judgments. AI can't fully replace the complex, messy collaboration of a founding team when it comes to creating a truly world-class product.

The Takeaway: The Bottleneck Moved

The paper concludes that Generative AI didn't remove all the obstacles in entrepreneurship; it just moved the bottleneck.

Before: The hardest part was just starting (getting the tools and skills to build something). AI fixed this, so anyone can start.
Now: The hardest part is finishing and refining something great. The barrier to entry is gone, but the barrier to excellence remains high and still favors teams.

In short: AI made it easy for one person to try a thousand things, but it still takes a team to build the one thing that stands out as the best.

Technical Summary: Generative AI Fuels Solo Entrepreneurship, but Teams Still Lead at the Top

Problem and Motivation
Recent advances in generative artificial intelligence (AI), particularly the public release of large language models like ChatGPT-3.5 in November 2022, have lowered barriers to entry for creating digital products. While existing literature suggests that generative AI enables leaner organizational forms and allows individuals to perform tasks previously requiring teams (Cai et al., 2025), it remains unclear how this shift affects the quality of new ventures. Economic growth is driven not merely by the volume of venture creation, but by a small subset of high-growth, innovation-driven firms (Guzman and Stern, 2020; Botelho et al., 2026). The central research question is whether generative AI democratizes entrepreneurship by increasing solo entry, and if so, whether this expansion translates into higher-quality outcomes or merely a proliferation of low-commitment experiments.

Data and Empirical Setting
The authors utilize a novel dataset comprising over 160,000 product launches on Product Hunt, an online platform where entrepreneurs introduce new digital products. This setting is critical because it allows for the joint observation of entrepreneurial entry and immediate market-based evaluations.

Unit of Analysis: New product launches, treated as instances of entrepreneurial entry (transition from idea to public realization).
Key Variables:
- Founding Structure: Distinguished between solo ventures (one maker) and team-based ventures (two or more makers), with further breakdowns by team size.
- Entry Persistence: A "one-shot" indicator for ventures with no updates within 12 months of launch.
- Product Distinctiveness: Measured by the rarity of a product's tag combination relative to other products launched on the same day.
- Quality Outcomes: Assessed via Product Hunt's daily rankings (Top 10, Top 30, Top 50, and All), which aggregate user votes and engagement.
Timeframe: Monthly data from April 2020 to July 2025, providing balanced pre- and post-periods around the November 2022 release of ChatGPT-3.5.

Methodology
The study employs a quasi-experimental design leveraging the release of ChatGPT-3.5 as an exogenous shock.

Difference-in-Differences (DiD) and Event Study: The authors estimate a two-way fixed effects model to compare the differential changes in entrepreneurial entry between solo and team-based ventures before and after November 30, 2022. The specification controls for persistent differences between groups and month-specific shocks.
- Equation: $\log(N_{gt}) = \alpha + \beta(Post_t \times Treat_g) + \gamma_g + \delta_t + \epsilon_{gt}$
Quality Distribution Analysis: To assess how entry changes translate to quality, the authors analyze the share of team-based ventures within specific ranking tiers (Top 10, Top 30, etc.) using a pre-post specification that interacts the post-shock period with ranking tier indicators.
Heterogeneity Analysis: The study examines variations across product categories (classifying them as historically "team-heavy" or "solo-heavy") and analyzes behavioral shifts regarding persistence (one-shot entries) and product distinctiveness.

Key Results
The findings reveal a significant divergence between the quantity and quality of entrepreneurship following the emergence of generative AI:

Surge in Solo Entry: Entrepreneurial entry increased sharply post-ChatGPT (rising from ~29k in 2022 to ~41k in 2023). This growth was disproportionately driven by solo founders. By 2025, solo entry grew nearly 90% more than team entry.
Category-Specific Shifts: The increase in solo entry was most pronounced in categories that historically relied on teams (e.g., Software-as-a-Service, Fintech), suggesting AI is partially substituting for the need for co-founders in these domains.
Low-Commitment Experimentation: The expansion in solo entry is characterized by "one-shot" behavior. The share of solo ventures with no updates within 12 months rose from 95.1% to 97%. Furthermore, solo ventures exhibited a 4% larger increase in product distinctiveness compared to teams, indicating a broader, more exploratory search of the idea space with lower average commitment.
Team Dominance in Top Tiers: Despite the surge in solo entries, team-based ventures continue to dominate the highest-quality outcomes.
- The share of teams in the Top 10 increased from 50% to 53% post-ChatGPT.
- The share of teams in the Top 30 also increased.
- Conversely, the share of teams in the "All" category (including lower-ranked products) decreased.
- Gains in the Top 10 were concentrated among larger teams (5+ members), while solo and duo ventures saw relative declines.

Key Contributions

Entry vs. Quality Divergence: The paper provides one of the first large-scale empirical analyses showing that while generative AI lowers entry barriers and expands the volume of solo entrepreneurship, it does not proportionally increase the incidence of high-quality, growth-oriented ventures.
Mechanism of Experimentation: It demonstrates that generative AI reshapes who engages in experimentation and how. Solo entrepreneurs use AI to conduct low-commitment, trial-and-error searches across a wider range of product attributes, whereas teams retain advantages in refining and coordinating efforts to reach top-tier outcomes.
Constraint Reallocation: The study argues that generative AI does not eliminate constraints but shifts them. As barriers to initial creation fall, bottlenecks move from venture creation to the development, coordination, and refinement required for superior outcomes.

Significance and Claims
The authors conclude that generative AI induces a structural shift in entrepreneurship where the "bottleneck" moves from the ability to launch a product to the ability to sustain and refine it.

Policy Implication: Policies that reduce entry frictions may broaden participation and encourage experimentation but do not automatically generate high-quality outcomes.
Organizational Implication: Producing top-tier outcomes continues to depend on the recombination of diverse human knowledge and tacit experience, which AI models alone cannot fully substitute.
Macro Implication: The findings suggest a mechanism where micro-level productivity improvements (enabled by AI) may fail to scale into aggregate innovation if they remain concentrated in low-commitment, experimental activities rather than sustained development.

The paper maintains a modest scope, acknowledging that its findings are based on a single digital platform focused on software launches and may not fully capture venture creation in capital-intensive industries or later-stage business outcomes (e.g., survival, revenue growth).