What changes will occur in an organization when AI becomes a co-founder?
What if the first "employee" of your startup isn't a person, but an AI agent? This highly challenging question became the focus of a one-semester collaboration between Microsoft and the Technology MBA program at New York University's Stern School of Business. Thirty students were divided into six startup-like teams. They were given access to Microsoft 365 Copilot with the latest agent capabilities and were asked to reimagine how work would be carried out when AI is embedded from day one. Their task was to break traditional work processes, build a "frontier company," and explore the future of human-AI collaboration.
We envision "frontier companies" as organizations that not only adopt AI but are built around it. From the day of their establishment, these companies integrate AI agents into every function, redesign work processes, decision-making methods, and team structures to maximize human-AI collaboration. In the context of this project, the "frontier company" model encourages students to go beyond automation and explore what it truly means to lead and scale an organization where AI is a fundamental team member (rather than an add-on).
The results are both inspiring, disruptive, and thought-provoking. The most successful teams not only use AI to complete tasks but also strive to tap into its creative potential, such as building organizational charts, reshaping project presentations, and even proposing hypothetical business models to prompt human leaders to respond. In the process, they discovered a new work paradigm where AI is an active and evolving team member.
This article distills their research results into four transformative themes that redefine our understanding of work, teams, and leadership in the AI era. These observed changes are not isolated but closely connected: The way work is initiated affects recruitment needs, and the way teams are managed shapes the decision-making process. Together, they provide a roadmap for building and leading AI-native organizations.
Research Methodology
Students participating in the collaboration project between New York University's Stern School of Business and Microsoft were divided into teams. Each team was given a startup proposition and asked to create their business. They could access a virtual business environment where they could hold meetings, send emails, and create documents, and they had access to Microsoft's latest AI capabilities. Startup ideas included a home solar company, a consumer app offering personal finance advice, and a health and fitness wearable device company.
The project was carried out in two phases. First, students used Copilot to simulate launching a startup, assigning management roles, drafting plans, creating content, and testing work processes. Then, they thought about how an "AI-first" company (what Microsoft calls a "frontier company") might operate in the future and developed a blueprint for an organization that integrates AI agents into every function. The goal of the project was to explore the "AI agent-enabled future of work" and understand how to leverage AI in a business environment. The key is that starting from scratch gives students an advantage: they are not bound by any legacy structures.
AI Becomes the First "Employee"
Students said that AI quickly became the team's "first employee" and was used in basic roles such as strategist and analyst. For example, one team tried to have AI design their organizational chart based on resume analysis. Other key tasks included developing market entry strategies, writing job descriptions for recruiting third-party consultants, creating financial models, and designing brand manuals and logos. The ability to access a wide range of functions instantly allowed human founders to take on multiple roles without being overwhelmed and to handle tasks beyond their core expertise. As a result, students were able to quickly transition from ideas to execution. Whether it was a low-risk or high-risk decision, it was first discussed with AI, introducing a new decision-making paradigm for entrepreneurs.
This shift has far-reaching implications for the entire startup ecosystem. In startups, many "to-do tasks" are vague or temporary, and recruitment can be a major distraction for founders trying to move quickly. AI's ability to handle vague tasks and produce tangible results helps clarify a company's real needs, providing founders with a more solid foundation to conduct business. In the early stages, AI's work is often sufficient to support a company's operation with minimal human resources.
What can we learn from this? AI changes the way recruitment is considered. Founders may now ask not "Who do I need to hire?" but "To what extent can AI help me from the start, and what gaps still exist?" This repositiones AI as an economic lever rather than a cost center. It enables leaner teams, faster iteration and prototyping, and more strategic recruitment and expansion. Even if a company hires in the future, being able to discuss with an AI "co-founder" and develop (and respond to) drafts of various deliverables gives the company an opportunity to understand its real needs.
Work Starts with Dialogue
One of the most significant observed shifts is how work has moved from static documents to dynamic dialogues. Students scheduled real-time meetings to discuss background information as the "seeds" of concepts with AI so that AI could focus on drafting actual reports or presentations. This changed the work experience: instead of opening an application and starting from scratch, it starts with a dialogue with Copilot. As one student said, "I have an idea, and I can just tell Copilot, and it will generate a draft for me. This really helps me get started with work."
This conversational work model, which one team called "pair programming for every task," not only improves output efficiency but also reduces people's fear of new tasks. One group used AI to create an investor presentation and then discussed in a meeting what insights were missing and how to adjust the content for different audiences. AI was responsible for execution, and humans provided insights. This means that the role of humans has changed. Students are no longer just meticulous content creators but have become both idea generators (providing rough, unstructured starting points) and screeners and optimizers (guiding, editing, and perfecting AI-generated drafts). Students said that working with AI feels more like collaborating with colleagues than using software.
One team found that while AI is very good at generating various logo options, the team must use their understanding of the brand and target audience to select the appropriate logo. This shift from creator to inspirer and screener has profound implications. It shows that the most valuable skills in the future will be the ability to shape and guide AI output. Domain expertise, critical thinking, and editorial judgment become new superpowers. When natural language becomes the new user interface, what matters is not knowing which menu to click or what formula to write, but clearly expressing what you need and then evaluating the content provided by AI. Knowing how to communicate, that is, clearly expressing intentions and expected results, will become a highly valued skill set in the future. This has a huge ripple effect on the cultural construction within a company: in traditional concepts, over-communication is often regarded as an obstacle to an efficient work process and is therefore not favored.
Another aspect of this theme is that documents themselves may become secondary to dialogues. Continuous dialogue with the AI agent means that teams no longer think about planning with Word documents, budgeting with spreadsheets, or making proposals with PowerPoint. Instead, they focus on discussing ideas and problems with each other while letting AI create documents. And these documents can be viewed during meetings to see if the team likes the direction they are discussing. The content exists in the dialogue, and AI will generate different documents as needed. In the AI era, establishing context is crucial: every recorded meeting or dialogue contributes to the context that AI will dominate in the future. Since AI generates the first draft, the value of human contribution lies in what only humans can do: establishing context, perceiving nuances, applying context, and giving meaning. This elevates the role of humans from executors to coordinators, thus changing the social contract of work.
The Role of Human Knowledge
This view of humans as inspirers and screeners shows that the role of human knowledge has changed. Students don't need to be in-depth experts in every task they undertake because they can use AI's power to supplement their knowledge.
That being said, they also expressed concerns that AI's confidence might create a false sense of security, and they might not be able to evaluate it if they don't have the corresponding knowledge. Several times, AI's output ignored nuances or contained misleading data, just like an overconfident junior analyst.
They emphasized the idea that "Copilot needs to challenge, not just please." One student shared, "You have to be an expert. You have to be the one who says 'This is right, this is wrong,' and you have to be the one who makes the final decision."
The core positive outcome is that AI significantly reduces the cost of decision-making. A team considering marketing methods asked AI to predict the results if they shifted their budget from events to online advertising. AI quickly generated a rough comparison, which traditionally would have required a lot of effort. This near-instant intelligence makes exploring various alternatives cheaper and faster and allows humans to focus on using their knowledge to evaluate a large number of quickly generated ideas. Teams can ask "What if...?" more frequently because AI can immediately analyze various scenarios. In the AI era, there is no cost to being curious.
It also shifts the burden of low-risk transactional decision-making to AI: for topics of low value to the entire team, students can use AI to brainstorm. Some of the students participating in the project had limited actual management experience, but they found that they could make important strategic suggestions because AI provided data and even reasoning. This indicates that the hierarchical structure may become flatter. Decision-making can become more decentralized: people closer to the front line may make decisions that previously required "reporting upwards" because they have AI-generated insights as support. Therefore, conversations between people also become riskier and potentially more meaningful: transactional conversation topics have been entrusted to the human-machine interface.
The Dynamics of Human-AI Teams
Perhaps the most groundbreaking insight from this project is how AI reshapes team dynamics. Since AI is responsible for task execution, teams are smaller and act faster. Students described a "multi-agent network" where different AI agents handle different areas, such as customer relationship management, scheduling, finance, etc., while humans act as conductors, coordinating these agents and making final decisions. The idea is that each agent can have its own role, knowledge, and perspective, so the integration of each agent's skills becomes crucial.
This model subverts the traditional team structure. Instead of people using multiple tools, people manage many AI "workers" that use tools. This means that the work model has shifted from humans directly using machines for labor to humans leveraging digital labor, and digital labor then uses machines to complete work. A new work process is thus established.
Redefining AI as a form of labor, rather than just a function, has far-reaching implications for leadership. This indicates that managing AI will become a core competency, requiring new skills in coordination, supervision, and ethical governance.
...
The collaboration between New York University's Stern School of Business and Microsoft gives us a glimpse into a rapidly emerging reality: AI-enhanced organizations. It shows that AI can be not only a productivity tool but also a teammate, a strategist, and even a co-founder. This is not a distant future vision but a phenomenon that is already reshaping the nature of work, innovation methods, and competitive advantages. The experiences of the students provide a clear blueprint for how leaders can navigate this transformation.
To achieve this vision, more than just technology is required. It requires a shift in mindset, from seeing AI as a tool to seeing it as a collaborator. It requires new skills, new structures, and a new way of thinking about work. As we stand on the brink of this transformation, one thing is clear: the future of work will not be built on outdated assumptions about work. This prompts us to think about a key question: If today were the first day of the history of work, how would we plan our work methods?
J.P. Eggers, Sarah Ryan, Alexia Cambon, Jared Spataro | By
J.P. Eggers is a professor of entrepreneurship at New York University's Stern School of Business. Sarah Ryan is the director of the venture acceleration program at New York University's Stern School of Business. Alexia Cambon is the head of research on Microsoft Copilot and the future of work. Jared Spataro is the chief marketing officer for Microsoft's artificial intelligence business in the work scenario.
Qiang Zhou | Edited and Proofread
This article is from the WeChat official account "Harvard Business Review" (ID: hbrchinese). The author is HBR-China. Republished by 36Kr with permission.