Agentic Transformation: Can an Existing Organization of People Learn Every Day with AI Agents?

Research note · Version 0.1 · October 28, 2026 · Status: Running (the knowledge core); Early (an agent for every person)

Mike Lee, drafted with his AI agent · Global Student Startup Foundation · GSSF Press

Cite as: Mike Lee, "Agentic Transformation: Can an Existing Organization of People Learn Every Day with AI Agents?," research note, version 0.1 (Agentic Organizations, GSSF Press, October 28, 2026), https://agenticorganizations.org/research/agentic-transformation.

1. Summary

Most organizations meet AI as one tool for everyone, and each person uses it differently; the tools are capable while the organization stays forgetful ¹². This note describes a live test of a different design at the Global Student Startup Foundation, a small, mostly volunteer nonprofit: one shared knowledge core (the one place where the organization keeps what it knows) at several levels of access, an agent for each person who wants one, built from a shared template and acting only with that person's access, and a way for what one person learns to reach everyone, by consent ¹²; the template design is in ³. The knowledge core is Running; an agent for every person is Early; measures of what it saves are Planned ¹. We describe the setting, the design, the record kept of each change, what has been observed in the first days, and what would show the design is not working, drawing on research on organizational learning and on evidence about AI assistance at work ⁴⁵⁶.

2. The research question

Can an existing organization of people be changed so that it learns every day, with AI agents doing the labor of keeping its knowledge current and people making every decision?

1. Does one shared knowledge core, at levels of access, keep everyone's answers consistent (dates, policies, who to ask)?
2. Do people use their own agents more fully when their private knowledge is protected, and reaches the organization only by consent?
3. Does a change made once, in the shared template or the core, reach everyone without retraining each person?
4. What does the transformation save, and for whom: newcomers, volunteers, or leaders?

3. Background and related work

Research on organizational learning describes what a learning organization does: it creates, acquires, and transfers knowledge and changes its behavior to reflect it ⁴, and Argyris and Schön distinguished fixing a mistake from changing the rules behind it ⁷. Nonaka and Takeuchi described how organizations turn what individuals know into knowledge the organization can use ⁸. Von Hippel showed that users, not only manufacturers, are often the ones who innovate ⁹. In our reading, the obstacle to putting these ideas into practice was the labor of capturing and reconciling what people know ¹.

Evidence on AI assistance at work suggests the gains are uneven. In one large study of human customer-support workers, access to a generative AI assistant raised productivity by 14 percent on average, with the largest gains, 34 percent, for novice and low-skilled workers ⁵. Analysts have asked whether organizations are ready for personalized AI agents in the workplace ⁶, and legal scholarship has begun to ask how the law of agency applies to them ¹⁰. Risk frameworks organize AI risk management into four functions, govern, map, measure, and manage, applied throughout an AI system's life ¹¹.

We have not found studies of the transformation itself, though our search so far has been informal: what happens when an existing organization, with its habits and documents, adopts a shared knowledge core and an agent for each person, and what that changes for the people in it.

4. Setting and method

The setting. The Global Student Startup Foundation is a U.S. 501(c)(3) nonprofit that runs the Global Student Startup Competition; it reached more than 450 students in 2026 ¹¹². Its team is small, mostly volunteer, spread across many time zones, and much of it changes every year ¹.

The design under test. One knowledge core at four levels of access, from public to the most restricted; each fact kept once, at the lowest level allowed to see it; an agent for each person, built from one template and working with that person's own access; private knowledge reaching the shared core only by consent; and websites and dashboards drawn from the core rather than copied ¹²³. An AI Agent Policy, adopted and frozen, says that an agent works only for its person, that its memory is private, and that decisions about people are made by people ¹³.

What is recorded. Changes to how the organization works are recorded, dated, with why it was made, what was done, and what was learned, in the appendix of The Knowledge Core, "Actions Taken" ¹⁴.

How evidence is judged. By that record of changes, and by measures planned for the 2027 competition cycle ¹. Because the author designed the system and also takes part in running it, findings are reported as a case, with their status marks.

5. What we have observed so far

The record of the first days, October 17 to 25, 2026, is in the online appendix of The Knowledge Core, for its signed-in readers ¹⁴. A shared template replaced building each person's agent by hand. A one-page agent policy was adopted because people would not use an agent fully without knowing what happens to what they tell it. When changing one person's title touched six places, each kind of data was given one master record. When replies to outsiders drafted from internal knowledge could quote things not meant for them, a public level was added below the others. And when a person switched AI accounts, keeping the agent's memory in the person's own repository meant the same core opened from every account ¹⁴.

Status, as of this note: the knowledge core is Running; an agent for every person is Early, with several members of the team working with their own agent; measures of what it saves are Planned ¹. What is not yet known is whether the design saves time, for whom, and whether it survives the yearly turnover of a volunteer team.

6. What would change our mind

What would count against the design. If people who have agents use them little; if facts still reach people in conflicting versions after the core is in place; if what one person learns rarely reaches others even with consent; or if the knowledge of people who leave is lost anyway, then the transformation is not doing what it claims.

The strongest objection. Organizations may adopt AI tools faster than they change how they work, so that agents add another place to look rather than removing work, and the gains are uneven. In the customer-support study, the largest gains, 34 percent, went to novice and low-skilled workers ⁵, so who benefits depends on who is helped. This test is designed to show who, if anyone, the design helps, and the measures are planned for that reason.

Open questions. How much structure must a volunteer organization accept for its knowledge to stay current? Which decisions should never be delegated to agents ¹³? What is the smallest version of this design a small organization can adopt first?

7. Ethics and privacy

Each person's agent works only for that person, and its memory is private; the organization never reads it ¹³¹⁴. What a person's agent learns reaches the shared core only with that person's consent ²¹³. The policy says, in its section 2: "Decisions about people, such as who is admitted, judged, chosen, or helped, are made by people" ¹³. Students and partners are not named in this research.

8. How it reaches students

What GSSF learns is written into The Knowledge Core and into Building Agentic Organizations, one of GSSF Press's textbooks for students ¹⁵.

9. What this means for your organization

Start with one shared place for what the organization knows, and an agent for each person that works only with that person's access; let what one person learns reach everyone only with their consent. Which facts in your organization are kept in more than one place today, and who would notice if two of them disagreed?

10. References

1. Mike Lee, The Knowledge Core: Leading a Population of AI Agents (GSSF Press, 2026), chapter 1, "The Agentic Organization". https://theknowledgecore.org/cite/the-agentic-organization
2. Mike Lee, The Knowledge Core: Leading a Population of AI Agents (GSSF Press, 2026), chapter 6, "An Agent for Everyone", section 2. https://theknowledgecore.org/cite/two-kinds-of-knowledge-kept-apart
3. Mike Lee, Building Agentic Organizations: How to Build an Organization of People and Agents (GSSF Press, 2026), chapter 9, "An Agent for Every Person". https://buildingagenticorganizations.org/book/ch09
4. David A. Garvin, "Building a Learning Organization," Harvard Business Review 71, no. 4 (July–August 1993): 78–91. https://hbr.org/1993/07/building-a-learning-organization
5. Erik Brynjolfsson, Danielle Li, and Lindsey R. Raymond, "Generative AI at Work," NBER Working Paper 31161 (April 2023, revised November 2023). https://www.nber.org/papers/w31161
6. Gregory Vert, Brad Kreit, and Abha Kulkarni, "Are Organizations Ready for Personalized AI Agents in the Workplace?," Deloitte Insights, October 12, 2026. https://www.deloitte.com/global/en/insights/topics/emerging-technologies/personal-ai-agents-at-work.html
7. Chris Argyris and Donald A. Schön, Organizational Learning: A Theory of Action Perspective (Reading, MA: Addison-Wesley, 1978).
8. Ikujiro Nonaka and Hirotaka Takeuchi, The Knowledge-Creating Company (New York: Oxford University Press, 1995).
9. Eric von Hippel, Democratizing Innovation (Cambridge, MA: MIT Press, 2005).
10. Deven R. Desai and Mark O. Riedl, "Using Agency Law to Tame AI Agents," Berkeley Technology Law Journal 41 (2026): 465–516. https://doi.org/10.15779/Z38KW57M6R
11. National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (Gaithersburg, MD, January 2023). https://doi.org/10.6028/NIST.AI.100-1
12. Global Student Startup Foundation, "GSSC 2026," Global Student Startup Competition, accessed October 25, 2026. https://globalstudentstartup.org/#/gssc-2026
13. Global Student Startup Foundation, AI Agent Policy, version 2 (October 24, 2026). https://foundation.globalstudentstartup.org/#/ai-agent-policy
14. Mike Lee, The Knowledge Core: Leading a Population of AI Agents (GSSF Press, 2026), appendix, "Actions Taken," entries of October 17–25, 2026. https://theknowledgecore.org/cite/actions-taken
15. Agentic Organizations, "About Agentic Organizations," accessed October 28, 2026. https://agenticorganizations.org/about

A research note is a working document: it states what we are testing and what the record shows so far, and it is revised as the record grows; each version is dated. It is not peer reviewed. If you are testing something similar, we would like to compare notes: [email protected]. See something wrong? Report an error. All research areas · Our standards