What a Generative AI Program Taught Me About Organizational

When I enrolled in MIT Professional Education's Generative AI program, I expected to learn more about the tools, how they work, and how they are integrated into agentic AI systems. Having completed the Agentic AI certification months earlier, I was aware that generative AI tools are the backbone of any agentic system. So I was curious to learn how Large Language Models actually work. I already knew what terms like embeddings, vector databases, RAG, context windows, and LLMOps meant, but I wanted to feel I could talk to an IT Director or a computer engineer with real expertise and engage in technical conversations. I thought that was the only way I could show expertise to help organizations lead digital transformation initiatives.

I did learn all of that. I also discovered there is still so much knowledge and depth to acquire, and so many skills to develop.

But the biggest surprise was how naturally I learned a technical concept and how quickly I was able to connect it to my own experience as an Organizational Development consultant. This happened more than once.

Most of all, many of the technical concepts did not feel distant from my work. The more I learned, the more I found myself returning to questions I have been exploring for years: how knowledge moves inside organizations, how expertise is developed and shared, how work is coordinated, how decisions are made, and how leaders help people adapt when the environment changes.

In that sense, the program gave me more than technical vocabulary. It gave me a new lens to think about organizational design.

The Information Problem You Already Know

For example, Retrieval-Augmented Generation, or RAG. At a technical level, RAG allows a Large Language Model to retrieve relevant information from external sources before generating a response. Instead of relying solely on what is already in the model, the system can search for information that may improve the answer's quality and relevance. Isn't this what we, as OD consultants, have been helping organizations and leaders acknowledge: how to broaden their perspective and retrieve more and more information rather than just what they had in front of them?

I remember a conversation I once had with a client. He came to me, erupting, about the latest information on a team's performance. It did not take me long to realize that he had reproduced a reaction (or an outcome, in LLM terms) based on very limited information, the one at hand, so I helped him see how it would be beneficial to get more information before making his decision.

Most organizations today are not suffering from a lack of data. If anything, they are producing more data than ever before. The challenge is that information is often distributed across multiple systems, departments, documents, reports, conversations, and platforms. People may have access to more information, but that does not mean they can always find what they need, understand what matters, or use it at the right moment.

This is where I see a strong connection between AI and Organizational Development. RAG is a technical solution, but the problem it tries to solve feels very familiar. Organizations also need better ways to connect people with relevant knowledge, especially when they are making decisions, solving problems, or learning from past experiences.

What Democratization Feels Like From the Inside

Being able to experience firsthand the democratization of expertise has been mind-blowing, even when I started learning about AI a few years ago. I am an organizational psychologist who lives comfortably in abstract constructs that need no technical expertise. So the fact that I was easily able to create websites, agents, my own LLMs, and analytical reports that I used to hire people to do for me led me to better understand the fears about the impact of AI on jobs. Democratization of knowledge means that, through natural language interfaces, tasks that once required specialized knowledge are becoming more accessible to people without technical training. Employees can now use AI tools to explore solutions in ways that were previously impossible.

So, if I experienced this shift in my own work, with tasks I used to subcontract, it is not hard to picture what is happening at large scale in organizations. I have read and listened to interviews with more optimistic and more pessimistic perspectives on this. I am honestly not grounded in one perspective yet. But my opinion now — it might change in a few months or weeks — is that the democratization of knowledge is opening doors for people to become more exposed to new skills and capabilities, to create and partner with AI tools, and that production capability and quality will continue to be incremental. That raises many ethical questions for me as a psychologist. Some have already mentioned that all of this is actually creating cognitive overload and a fried brain, due to the overwhelm of new tools we are now using at work.

At scale, with all the new work titles emerging, the necessary question is whether organizations are designed to develop those capabilities internally as needed. My inference is that the need for organizational redesign must accelerate dramatically, or the use of resources outside organizations — like the call center dynamic — will exponentially put jobs in the US at risk. As an OD consultant, this raises important questions and concerns.

The Leadership Question Nobody Is Answering Yet

What do the leaders of tomorrow need? Again, we keep hearing about the need for leaders with a greater capacity for critical thinking, empathy, curiosity, judgment, and perspective. But how specifically are those skills relevant and applied in the future? What leadership development programs do organizations need now, in this specific transition, as we operate in a hybrid model that has not fully embraced AI yet, while still working as we have in the past and testing the waters with the work of the future?

I do not have complete answers yet. But what the program helped me see is that the organizational conditions for leadership matter as much as the leadership qualities themselves.

And how will leaders create value, develop teams, and operate within the new organizational design? The questions that come up in AI systems, such as how to coordinate work, how roles are defined, how tasks are divided, how dependencies are managed, how a multi-agent system avoids unreliable outputs when one component depends on another, sound very similar to the challenges we see in organizations every day. So does this mean the conditions are basically the same? Maybe. But what really raises for me is how transferable the leadership capabilities that served us in the past will be going forward.

Organizations often assume that if they have talented people, performance will improve. However, talent alone is not enough. People need clarity about roles, responsibilities, priorities, communication channels, decision rights, and expectations. The same idea seems to apply when intelligent systems become part of the work. Adding more intelligence to a system does not automatically make the system more effective. The way work is organized still matters.

This is one of the reasons I believe Generative AI should not be treated only as a technology implementation. As organizations begin to integrate AI into workflows, leaders will need to think carefully about how work itself is designed and its impact on people. They will need to ask how tasks will be divided between people and systems, how accountability will be maintained, how quality will be reviewed, and how teams will coordinate when AI becomes part of the process.

Why Implementation Is Never the Finish Line

The concept of LLMOps also gave me an unexpected connection to change management. LLMOps focuses on the practices required to develop, test, deploy, monitor, and improve Large Language Model applications over time. This includes evaluation, iteration, and continuous monitoring because AI systems are not static. They need to be observed, adjusted, and improved as conditions change.

This reminded me of something we often see in organizational transformation. Many leaders treat implementation as the end of the process, when in reality it is often the beginning. A new system, a new strategy, a new culture initiative, or a new leadership model does not become part of the organization simply because it was launched. It needs reinforcement, feedback, adjustment, and continuous learning.

Studying LLMOps helped me see another parallel between AI and Organizational Development. Both require discipline after the initial launch. Both require ongoing attention to whether the system is producing the desired results. Both require the ability to monitor what is working, identify where problems are emerging, and make adjustments before the system loses credibility.

This may be one of the most important leadership lessons from the program. Digital transformation cannot be approached as a one-time project. It requires a learning mindset. Leaders must be willing to test, observe, listen, adapt, and redesign as they go. That is not always easy in organizations that prefer certainty, fixed plans, and quick results.

The Guardrails We Still Need to Build

The program also made me more aware of the ethical and governance questions that surround AI. In one of my assignments, I developed an ethics policy for the use of Generative AI at Culture To Fit. That exercise helped me think about issues such as confidentiality, intellectual property, responsible use, quality assurance, and human review. Since our work deals with deeply human topics like culture, leadership behavior, employee trust, service experience, resistance to change, and organizational identity, AI cannot be used casually.

For consulting firms, this is especially important. AI can help us draft, organize, retrieve, and summarize information, but it should not replace professional judgment or the sensitivity required when working with people and organizations. It also should not expose proprietary methods, confidential client information, or employee data. The more powerful the tools become, the more intentional we need to be about the guardrails we create.

Looking back, I entered the program expecting to understand Generative AI as a technology. I finished the program thinking much more about organizational design. The concepts I studied gave me a new language, but many of the deeper questions felt familiar. I found myself thinking about how knowledge is distributed, how expertise changes, how work is coordinated, how systems are monitored, and how leaders create the conditions for responsible adoption.

These questions are not new for Organizational Development professionals. What is new is the context in which we are now asking them.

Generative AI is forcing organizations to revisit long-standing assumptions about work, expertise, learning, and decision-making. It is also creating an opportunity for OD professionals to contribute in a meaningful way. We can help leaders move beyond the excitement of the tools and think more carefully about the organizational conditions needed for these tools to create real value.

For me, that was one of the most important takeaways from the program. I did not walk away thinking that I needed to become a technologist. I walked away understanding that I need enough technological fluency to participate in the conversation, ask better questions, and help leaders connect digital transformation with the human and organizational realities that will ultimately determine whether the transformation succeeds.