AI is changing the work. Now organizations have to change, too.

Most conversations about artificial intelligence begin with technology. Which model are you using? What can it automate? How much time can it save?

Willy Walker’s recent Walker Webcast conversation with Paul Cheek, a senior lecturer at the MIT Sloan School of Management, moved quickly past those questions.

Paul argued that AI tools plus AI training do not equal AI transformation. An organization can give its people sophisticated technology, teach them how to use it, and still operate much as it did before.

The larger opportunity begins when we stop asking how AI can make existing work faster and start examining whether the work should still be organized the same way.

Productivity is not transformation

Willy described a conversation with Walker & Dunlop professionals who were using AI to become more knowledgeable and insightful. When he asked how that increased knowledge had helped win business, the examples were limited.

That gets to one of the tensions surrounding AI investment today. An employee can save an hour without the company creating an hour of additional value. The economic benefit depends on what happens with the capacity AI creates.

Perhaps someone can serve more clients. Perhaps an analyst can spend less time gathering information and more time interpreting it. Perhaps a professional can identify an opportunity that would otherwise have been missed. Or perhaps a process that required four people can support substantially more activity without adding resources.

Those outcomes require more than productivity software. Paul compared AI investment to higher education. A student does not evaluate a four-year degree based on the wages from a freshman-year campus job. The investment builds capabilities that create value over a much longer period.

Companies face a similar challenge with AI. Short-term productivity metrics can capture part of the return, but they do not capture what happens as an organization develops new capabilities and then redesigns work around them.

Faster workflows can expose outdated structures

Willy described the traditional structure of a commercial real estate banking or investment sales group: a professional with strong client relationships supported by one or several people dedicated to that professional’s business.

Now imagine AI dramatically improves how quickly that group can ingest information, analyze a transaction, or prepare materials. The group has become more efficient. But if the volume entering that group remains unchanged, much of the newly created capacity sits inside the same organizational boundary. Technology improved the process. The structure limits the benefit.

Paul responded by arguing for organizational redesign. He described the future organization as a collection of “nodes,” some human and some AI agents, working toward a common purpose.

The terminology is new. The management question is familiar: How should resources be organized to produce the greatest value?

AI changes the answer because it changes the constraints.

Processes designed around limited human capacity, sequential handoffs, and information that took hours or days to assemble may make less sense when some of those constraints disappear. Simply adding AI to each step can speed up an old process without addressing whether every step, handoff, or organizational boundary still needs to exist.

The most consequential question may be what stays human

Paul offered four questions for evaluating where to apply AI:

• Is it feasible?
• What is the business impact?
• What is the risk?
• Should AI do it, or should a human do it?

Commercial real estate offers a clear example. AI can already help collect, organize, and analyze enormous amounts of information. As the technology improves, the range of decisions it can support will expand. But commercial real estate finance is not simply an information-processing business.

Willy described money as the ultimate commodity. A dollar from Walker & Dunlop is not inherently different from a dollar available elsewhere. Differentiation comes from the people involved: their relationships, judgment, ability to structure transactions, understanding of risk, and knowledge of their clients.

Automation that improves those capabilities can strengthen the service. Automation that strips them out indiscriminately can standardize the very experience that differentiates one firm from another.

Paul applies a similar principle to his own use of AI. His AI agents can analyze a meeting, identify follow-up items, and draft the next steps. They cannot send an email from him without his review. As he put it, if something is coming from him, it needs to come from him.

That boundary will look different across companies, roles, and workflows. Defining it requires understanding what the technology can do and, perhaps more importantly, where human involvement creates value.

AI gives established companies a different kind of opportunity

The conversation also challenged the familiar narrative that AI inherently favors startups over established companies.

AI has dramatically reduced the resources required to build many new products and capabilities. A small group can create something today that once required a much larger organization.

Building something is not the same as building a business around it.

Paul said one question he frequently hears from entrepreneurs is some version of: We built the product. Can you help us find clients? His response is to return to the market.

Established companies already have something startups often spend years trying to create: distribution. They also have client relationships, proprietary data, institutional knowledge, operating experience, and an understanding of how their markets actually work.

Walker & Dunlop, for example, has decades of commercial real estate experience and relationships, along with a servicing portfolio that provides a significant base of information and market knowledge. AI can build on those strengths rather than starting from zero.

The competitive question becomes less about incumbents versus startups and more about speed. A startup can begin with an AI-native operating model. An established company has to evolve one while continuing to serve clients, manage risk, and run the existing business.

The advantage shifts toward established companies when they can combine what they already possess with a faster ability to experiment, learn, and build.

AI literacy has to become part of the work

Paul repeatedly returned to the speed of technological change compared with the speed of human behavior. Technology can change overnight. Habits rarely do.

His recommendation was refreshingly practical: Give people “tinker time”—not another presentation about AI or a list of approved use cases, but time to actually use the technology.

Experimenting in low-risk situations gives people a chance to discover where AI works, where it fails, how prompts and context affect results, and which tasks are worth approaching differently.

That experimentation also has to extend to leadership. Paul described working with executive groups and boards that deliberately set aside time to build their own AI literacy. If leaders expect employees to change how they work while leadership continues operating exactly as before, adoption will have a natural ceiling. Organizational behavior follows what leaders make time for, not simply what they say is a priority.

The harder work starts after adoption

Companies have spent the past few years asking how quickly they can adopt AI. Adoption is the easier part. The harder questions sit underneath it:

• What happens to the capacity AI creates?
• Which workflows were designed around constraints that no longer exist?
• Where can AI extend human capability, and where would automation diminish the client experience?
• How should people and AI agents work together?
• What becomes possible when institutional knowledge, proprietary data, distribution, and relationships are paired with dramatically faster execution?

AI does not answer those questions for an organization; it makes them harder to avoid.

See how AI is reshaping the way organizations work. Watch Willy Walker’s full conversation with Paul Cheek.