Learn to manage abundant resources.
AI has drastically lowered the barriers to entrepreneurship, giving rise to abundance-driven entrepreneurship that is fundamentally different from lean entrepreneurship. Founders can leverage AI agents to incubate projects in batches, replacing resource allocation with attention allocation. However, abundance also brings risks such as homogenized creativity, false growth, and compliance issues. When resources are no longer scarce, the judgment to identify valuable opportunities will become the most important core competence for entrepreneurs.
Nearly two decades ago, entrepreneur and author Eric Ries proposed the “Lean Startup” methodology. Under this model, startups only need to prepare sufficient capital, a minimum viable product, and a core team, communicate fully with potential customers, and continuously iterate and pivot during the enterprise expansion process to launch projects. The Lean Startup concept integrates the lean production philosophy of Toyota and the customer development theory of Professor Steve Blank from Stanford University. This method quickly gained popularity since founders usually had limited resources at their disposal.
But today, with the support of AI, solo entrepreneurs can complete a large number of entrepreneurial tasks in parallel at extremely low cost: conceiving product ideas, simulating customer interviews, building product prototypes, developing websites, designing logos and other digital products. Emerging platforms allow founders to define their businesses at a macro level, assign roles such as CEO, engineer, and marketer to AI agents, who collaborate autonomously to complete tasks. In this brand-new model, entrepreneurs no longer perform tasks personally or even manage workflows. Instead, they set goals and constraints for the AI-operated organization. This capability not only accelerates the speed of entrepreneurship, but also represents a structural transformation, spawning a brand-new organizational design paradigm and reshaping the development methods of products and services.
In contrast to the Lean Startup, we call this AI-empowered new model Abundance-Driven Entrepreneurship. The entrepreneurial process is not completely free of all constraints, and many concepts of Lean Startup still apply. But there is no doubt that we have entered a new era of entrepreneurship, whose core feature is the excess supply of capabilities. The following is a comparison between the two models:
The following sections will describe these differences in more detail, list real cases of abundance-driven entrepreneurship, and discuss what AI-enabled entrepreneurship can and cannot achieve.
Background of the Era: Reconstruction of Constraints
The core driving force of this transformation is artificial intelligence. AI has drastically lowered the access threshold for information-based resources, including professional knowledge, data, and human-related resources such as creativity generation, analysis and judgment, and task execution. Founders can now produce code, design drafts and marketing content at low cost, and instantly obtain professional knowledge in finance, human resources, supply chain, and various entrepreneurial suggestions.
For example, a study on the incubation batches of Y Combinator over four years shows that startups that build products and business processes with AI have 25% smaller headcount than similar enterprises. Native AI startups have a 13% higher proportion of engineers, while frontline staff and management personnel are reduced by 15%. Such enterprises generally pursue two values at the same time: improving operational efficiency with the help of AI, and embedding AI into the products themselves.
There is a typical case of abundance-driven entrepreneurship: French entrepreneur Thibaud Louis-Lucas, as a solo founder, with only 10 employees, built and operated five software enterprises relying on AI. AI is not only used for product development, but also responsible for marketing, sales, and continuous product iteration. His idea is to launch multiple products in batches and only focus on the projects that have been verified by the market. He mentioned in an interview that the monthly revenue of his products exceeds 1 million US dollars.
Founders not only use AI to develop products, but also treat AI as an all-round co-founder. AI is replacing or assisting multiple functional positions, covering strategy, marketing, sales, and even human resources. To realize this AI productivity, it is usually necessary to schedule and orchestrate multiple AI agents, some of which are granted high autonomy. For example, solo entrepreneur Nat Eliason created an OpenClaw agent named “Felix”, gave it 1000 US dollars of startup capital and online payment interface, and instructed it to start an enterprise autonomously. According to the public dashboard data, Felix has founded three enterprises in total, with total revenue exceeding 200,000 US dollars.
However, entrepreneurs still face real constraints. Since everyone can easily access information, technology and professional knowledge, there is time competition to explore and implement feasible entrepreneurial ideas, and the feasibility judgment given by AI is not necessarily reliable. Another constraint is the founder’s attention. The question has changed from “Can we build it?” to “Which one should we choose to bring to the market?” As Steve Blank said in his entrepreneurship course: AI simplifies the process of customer research, but it will also “accelerate the implementation of bad ideas”. Overall, a series of transformations have taken place in the entrepreneurial process:
· From pursuing learning efficiency to generating and filtering creative options
· From focusing on the rigorous polishing of a single Minimum Viable Product (MVP) to the overall management of a multi-product portfolio
· From physical resource allocation to the allocation of the founder’s attention
· More need to distinguish between real and effective market signals (which are still scarce) and ubiquitous noise
These changes in the entrepreneurial process have many impacts. First, the number of startups will increase significantly. From November 2025 to January 2026, American entrepreneurs submitted a total of 1.56 million new business registration applications, hitting the highest record for the same period at least since 2004. We cannot determine how many of these enterprises build their businesses relying on AI, but a 2025 survey shows that more than 80% of American entrepreneurs believe that AI “improves operational and production efficiency” and “drives revenue and business growth” at the same time. AI has also lowered the industry access threshold, and the financing scale required for entrepreneurship (much of which comes from venture capital) is often lower.
Many founders use AI to build multiple projects at the same time, focusing on developing businesses that have achieved product-market fit. A global analysis covering more than 1,000 AI startups shows that 17% of founders operate multiple companies at the same time. In short, entrepreneurship is increasingly like managing a whole portfolio of low-cost experiments, which is in sharp contrast to the model of focusing on a single minimum viable product in the Lean Startup era.
Challenges of the Abundance Model
Although the AI-empowered abundance-driven entrepreneurship has huge potential and develops rapidly, practitioners of this model must deal with a number of challenges.
First, there is the risk of “progress illusion”: AI-generated content will create false signals of business growth. People-pleasing AI models find it difficult to tell founders directly that their ideas are flawed and that the enterprise is unlikely to succeed.
Second, AI may lead to the homogenization of entrepreneurial ideas. The lowered barriers to entrepreneurship have spawned a large number of similar projects. Although it is difficult to quantify, AI-driven entrepreneurship can easily give birth to many startups with identical ideas. Entrepreneurs may over-rely on the insights given by AI and ignore face-to-face communication with real and potential customers. Mature founders will know when to leave the computer and step into the real business world.
The convenience of code generation also brings potential hidden dangers. One founder we interviewed said that most of the code of her enterprise is generated by AI. Her customer satisfaction is acceptable, but she worries that customers can directly use AI to generate code by themselves to replace her application. The industry has been discussing that AI will replace traditional software vendors for some time, and now this risk also falls on startup software enterprises. All founders who use AI to develop digital products need to see this trend in the software industry — the window of competitive advantage for enterprises may be greatly shortened.
The last challenge: current AI may not be able to support the creation of scalable and successful enterprises. For example, there are several cases of vending machines operated by AI agents, where AI has made questionable or even radical decisions in marketing, pricing, and product display. Some observers have also questioned the use of AI-generated code by startups, arguing that the robustness of the output system is not sufficient to support large-scale expansion.
Professional Competence Remains Indispensable
Generative AI can provide professional knowledge in almost any field and write code on demand, but this does not mean that entrepreneurs no longer need industry expertise in their entrepreneurial tracks. In highly regulated industries such as finance and healthcare, the professional threshold is particularly high.
For example, Seth Dobrin, whom we interviewed, used to be in charge of AI business at enterprises such as IBM and Monsanto, and co-founded Arya Labs to develop new AI models based on mathematics and physics to solve complex problems in multiple fields such as drug and medical device R&D. The company has only 5 employees, and almost all the code is generated by AI. But Dobrin mentioned in the interview that the enterprise still needs vertical industry professionals to verify the models. He himself holds a doctorate in molecular and cell biology and has been engaged in genetics research for many years. He said that this professional background is very critical in sales communication with life science customers.
The case of MEDVi shows the problems that will arise when professional competence is lacking. This fast-growing telemedicine enterprise mainly operates GLP-1 drugs for weight loss and other health products. On the surface, it is a model of AI-empowered “solo unicorn”: with only two employees, the founder Matthew Gallagher and his brother, it achieved an annual revenue of 1.8 billion US dollars.
The New York Times reported: 41-year-old Gallagher, at his home in Los Angeles, used AI to write enterprise software code, write website copy, generate advertising images, texts and videos, and handle customer service at the same time; he built an AI system to analyze business performance, and outsourced all other work that he could not complete himself.
But problems emerged after in-depth investigation. In early 2026, the U.S. Food and Drug Administration (FDA) sent a warning letter to MEDVi, pointing out that its marketing website had misleading publicity, implying that self-made compound drugs were approved by the FDA. Subsequent reports also found other suspicious advertising behaviors of the enterprise.
Of course, enterprises founded under the traditional model sometimes have such compliance problems. But the entrepreneurial model that starts quickly relying on AI, lacks industry experience, and has a lean workforce will amplify such risks.
Enlightenment for Entrepreneurs and Mature Large Enterprises
The rise of abundance-driven entrepreneurship not only affects founders, but also is of great significance to mature enterprises that compete with them. Entrepreneurs should regard themselves as designers of systems and processes, not just product designers. Like entrepreneurs in the past, founders in the AI era still need to build competitive moats, which can be proprietary data, industry experience, or carefully accumulated professional cognition. They can conduct experiments, but need to grasp the scale to avoid excessive distraction of attention; they must establish a set of criteria for screening opportunities, relying on their own insights rather than simply trusting AI. Aesthetic taste, sense of responsibility, judgment, clear thinking ability and resilience are capabilities that will not be automatically obtained as resources become abundant.
Managers of mature enterprises must also use the same tools that startups are using to accelerate the speed of internal innovation and keep up with the pace. Many large organizations have adopted the Lean Startup concept, and now they also need to shift to AI-empowered abundance-driven entrepreneurship. Like startups, large enterprises need to learn to manage a larger portfolio of alternative projects.
In an era where digital products are readily available, competitive advantage comes from judging what is worth doing. For both entrepreneurs and mature enterprises, the challenge is no longer just to acquire scarce resources, but to learn to manage abundant resources. This brings rare opportunities for experiments and differentiation: forming new entrepreneurial teams, establishing new decision-making rules, designing new governance models, and exploring new methods of customer research. The rules behind abundance-driven entrepreneurship are still taking shape, which will reshape both startups and mature large enterprises; organizations that take the initiative to experiment in the new pattern and keep learning will occupy a favorable position in the future.
Victor P. Seidel, Bret Greenstein, Thomas H. Davenport | Article
Victor P. Seidel is Professor of Innovation Management at Babson College and Associate Research Fellow at Saïd Business School, University of Oxford. Bret Greenstein is the Head of AI Strategy at West Monroe, a global business technology consulting firm, who promotes the implementation of AI within the company and helps customers achieve quantifiable results through AI application. Thomas H. Davenport is Professor of Information Technology and Management at Babson College.
This article is from the WeChat Official Account “Harvard Business Review” (ID: hbrchinese), written by HBR-China, and authorized for release by 36Kr.
