The Agile Transformation Mistakes Organizations Cannot Afford to Repeat with AI with Jim Highsmith

We are already seeing a familiar transformation pattern 

Organizations are moving quickly to adopt AI because the opportunity is real. Workflows can move faster. Information can be synthesized in seconds. Leaders can receive options, recommendations, and analysis that once took teams days or weeks to produce. 

But speed is not the same as judgment. 

AI can accelerate a decision without improving the quality of that decision. It can generate more information without making the signal clearer. It can recommend action without understanding the political, human, financial, or strategic consequences the organization will carry after the decision is made. 

That distinction matters because PMO and transformation leaders sit close to the decisions that turn strategy into action. We help leaders decide what work matters, where capacity should go, what risks require attention, when the organization should continue, and when it needs to stop. If AI changes the speed and volume of those decisions, then our role cannot remain centered on administering process. We need to help design the decision environment itself. 

Agile gives us a warning from the recent past 

In my conversation with Jim Highsmith, one of the co-authors of the Agile Manifesto, we went back to what Agile was originally intended to solve. Organizations were using large, rigid methodologies to deliver work in an environment where technology and customer requirements were becoming increasingly uncertain. The old approach did not fit the conditions. 

The response was not supposed to be another rigid operating religion. It was a mindset built around adaptability, collaboration, learning, and responding to what the work revealed. 

Yet over time, many organizations replaced the old prescriptions with new ones. Teams became more focused on whether a sprint was the correct length, whether every practice was followed, or whether someone was “doing Agile right” than on whether the approach was helping the organization learn faster and deliver the right business outcome. 

Jim described the difference as doing Agile versus being Agile. Doing Agile is visible. It produces ceremonies, roles, rules, and artifacts. Being Agile is harder. It requires leaders and teams to adapt their decisions to the uncertainty, evidence, risk, and consequence in front of them. 

That is the history lesson AI leaders should be paying attention to now. 

AI transformation can repeat the same leadership mistake 

It is easy to make AI implementation about the tool. Pick a platform. Automate a workflow. Deploy an agent. Measure the volume of work completed or the number of tokens consumed. Celebrate the reduction in cycle time. 

Those measures may tell us the technology is active. They do not tell us the organization is making better decisions or creating more value. 

The deeper question is whether the operating model around the technology is ready. Are the goals clear? Is the data trustworthy? Are decision rights defined? Do people know when they are expected to accept an AI recommendation, challenge it, or stop the process? Is the organization learning, or is it simply moving faster? 

The warning is not that Agile and AI are the same. It is that organizations can make the same leadership error: confusing adoption with transformation, process activity with capability, and faster output with measurable business IMPACT. 

Move from process-oriented management to judgment-oriented management 

For decades, managers have spent enormous energy improving processes. We standardize steps, reduce variation, create controls, and optimize flow. That work still matters, especially when reliability, safety, and traceability are essential. 

But as AI takes over more of the process activity, the leadership advantage shifts. The differentiator becomes the organization’s ability to make sound judgments about what should happen next, how much evidence is enough, which risk is acceptable, and when a human must remain accountable for the call. 

Jim calls this judgment-oriented management. It does not eliminate process management. It changes the emphasis. The process supports the decision instead of becoming the point of the work. 

That shift creates an important opportunity for PMO and transformation leaders. We can stop being the team that produces more status information and become the capability that improves how decisions are made across the strategy lifecycle. 

Do not automate a workflow until you map the decisions inside it 

Jim offered a practical place to begin: identify the workflows you are responsible for and map the decision points inside them. 

Choose one workflow that matters. Do not start with every possible use case. Trace how the work moves and mark each place where someone has to decide what happens next. 

Then ask four questions: 

🔹 What decision is being made at this point? 
🔹 What evidence, experience, and pattern recognition does that decision require? 
🔹 Who should be accountable for the consequence of the decision? 
🔹 Should AI make the decision, recommend an option, analyze the inputs, or remain outside this step? 

This is a very different exercise from asking where AI can remove manual effort. It forces the team to look at decision quality, not only process speed. 

It also exposes where the organization may be creating capability risk. When AI takes over the easy decisions, people may lose the practice that teaches them how to recognize patterns and make harder decisions later. That is the autonomous-car problem Jim described. The system handles the routine driving until conditions become dangerous, then hands control back to a person whose experience has been eroding because the system has been driving for them. 

Automation is not the same as augmentation 

Automation removes activity from a human. Augmentation improves what the human can see, understand, or decide. 

Both can create value. The danger is treating automation as the default answer without considering what the human and the organization need to learn. 

An AI system may be excellent at sorting large volumes of information, identifying anomalies, summarizing options, or highlighting patterns. That can reduce decision fatigue and help executives focus on what matters. But the organization still needs leaders who understand the business context, challenge the assumptions, recognize when the pattern is misleading, and own the consequence. 

AI should reduce the noise around judgment, not replace judgment. 

The PMO can become the architect of the decision environment 

PMO and transformation leaders already work across strategy, portfolios, delivery, change, risk, capacity, and value realization. That enterprise view makes us well positioned to connect the parts of the decision system that often remain fragmented. 

We can help leaders define what decision needs to be made, create the right information flow, clarify decision rights, identify where AI can strengthen analysis, and protect the human accountability required for consequential choices. 

This is how the PMO moves beyond reporting what happened and starts improving what the organization decides to do next. 

The organizations that thrive with AI will not be the ones that automate the most steps. They will be the ones that know which decisions to accelerate, which judgments to protect, and how to keep building the human capability required when the easy answer is no longer enough. 

Press play above to hear my full conversation with Jim Highsmith and learn how the lessons of Agile can help your organization build a stronger, more adaptable decision system for the AI era.