Organizations are rushing to add AI to workflows they do not fully understand. 

The risk is not simply that the output will be wrong. The bigger risk is that AI will accelerate the gap between the way work is documented and the way it actually happens. When that gap is filled with slow decisions, unclear authority, hidden workarounds, and exhausted people compensating for a broken system, automation does not create transformation. It helps the organization do the wrong things faster. 

That is why PMO and transformation leaders need to look at AI differently. The most strategic opportunity may not be another productivity feature. It may be the visibility AI creates into the operating model beneath the work. 

AI enters two organizations at once 

David Dean, author of An Inbox Between Us and an enterprise AI strategist, describes organizations as having a written contract and an unwritten contract. 

The written contract is the formal version of work. It includes job descriptions, policies, governance models, process maps, workflows, values, project plans, and the documented steps everyone is expected to follow. That is the organization leaders believe they have designed. 

The unwritten contract is the version people actually experience. It is the collection of conversations, relationships, shortcuts, informal approvals, workarounds, judgment calls, and assumptions that allow work to continue when the formal system is incomplete or too slow. 

Every organization has both. The problem begins when leaders try to improve the written system without understanding the unwritten one. 

A process may show that a decision belongs to a particular role. In practice, the person in that role may not believe they have the authority to decide. A workflow may show a clean handoff. In practice, the handoff may depend on one trusted employee translating the request, chasing missing information, and calling three people who are not named anywhere in the process. 

That invisible coordination is not an exception to the work. It is part of the work. 

AI can help make that behavioral record easier to see. Emails, meeting transcripts, spoken updates, questions, delays, and repeated points of confusion can reveal where the organization is functioning differently than the process says it should. The value is not surveillance. The value is organizational self-awareness. 

AI can amplify behavior before it improves work 

Technology does not enter a neutral system. It enters the system people have already created through priorities, incentives, decision rights, trust, habits, and leadership behavior. 

David put the risk clearly: “AI is here to industrialize human behavior.” 

That means AI can scale what is working. It can also scale dysfunction. 

If people do not know which outcomes matter, AI can generate more activity without creating more value. If decision rights are unclear, AI can produce recommendations that no one feels authorized to use. If status reporting exists primarily to prove that work happened, AI may create more polished reports without improving a single leadership decision. 

This is not an argument against AI. It is an argument for better problem definition. 

The visible problem may be a slow approval process. The real issue may be that leaders have never clarified what can be decided at each level. The visible problem may be late status updates. The real issue may be that the reporting structure asks people to translate messy work into a format that does not reflect how they communicate or what leaders need to decide. 

AI can help with both examples, but only after the organization understands the human behavior underneath them. 

Use AI to tell the story of the work 

One example from the episode shows what a more human-centered use of AI could look like. 

Status reporting often requires a project leader to open a form, translate the work into predefined fields, and submit information that may already be scattered across meetings, messages, and project systems. The PMO then spends more time chasing the update and cleaning the data than helping leaders act on it. 

David suggested a different experience. Ask the person to speak the update naturally. Let AI compare that story with the available project information, clarify what is missing, and help shape an accurate update. 

The important shift is not dictation. It is designing the interaction around the way people communicate and the decision the update needs to support. 

Laura connected that idea to storytelling. Executives do not need more information simply because AI can generate it. They need a clear story about what is happening, what is changing, where they need to lean in, and what decision is required. 

PMO and transformation leaders are well positioned to create that clarity. They see the dependencies between teams, the gap between strategic intent and delivery reality, and the patterns that repeat across projects. AI can help connect those signals. Human leaders still have to interpret what the signals mean. 

Slow down before you accelerate 

When leaders feel behind on AI, slowing down can sound like the least responsible option. David’s advice was the opposite: slow down to speed up. 

This does not mean launching a six-month assessment or waiting until every policy is perfect. It means choosing one work pattern that people across the organization already agree is difficult and examining what is actually happening. 

Start with a process where frustration is shared. Approvals may stall. Priorities may keep changing. Decisions may be repeatedly escalated. Teams may spend hours producing status information that does not lead to action. 

Then look beyond the documented steps and ask: 

Where does the work really stop? 

What decision is not being made? 

What information do people not trust? 

What authority do people believe they lack? 

What workaround has become the real process? 

What conversation is everyone avoiding? 

The answer may not require a technology investment. It may require a clearer decision right, a better leadership conversation, or permission for someone closer to the work to act. 

Solve that human problem first. Then determine where AI can remove friction, improve sense-making, or create a better experience without weakening accountability. 

The PMO opportunity is bigger than automation 

PMO and transformation leaders should not limit their AI role to automating reports, summarizing meetings, or making dashboards more efficient. Those uses can help, but they are not the strategic opportunity. 

You sit close enough to the work to see the friction and close enough to leadership to understand the intended outcomes. You can see where the strategy assumes a level of clarity, capacity, authority, or coordination that the operating model does not actually provide. 

AI can help you make that gap visible. 

It can help identify recurring questions, repeated delays, conflicting expectations, invisible dependencies, and patterns of work that cost the organization time and capacity. That insight can strengthen prioritization, decision quality, resource focus, and value realization. 

The technology can surface the pattern. Your leadership determines what happens next. 

Start with a spoken digital journal 

David’s practical first step does not require a large initiative. Create a digital journal of your work and speak into it rather than trying to write polished notes. 

Capture what you are noticing. Where did a decision get stuck? What felt harder than it should have? Where did someone work around the process? What does the organization say it wants, and what do its decisions actually support? 

After you have collected enough entries, ask AI to help you identify the repeated themes and the questions worth exploring. Train it to challenge your thinking rather than simply agree with you. Use it as a thinking partner, not as a substitute for judgment. 

You remain accountable for what the system produces and for every decision you make with it. That is the human standard David returns to throughout the conversation: AI does not understand confidence, risk, or consequence. People do. 

The question is not whether AI will touch your work. It already has. 

The question is whether you will use it to automate the organization you imagine you have or to understand the organization that is actually doing the work. 

Press play above to hear the complete conversation with David Dean and learn how AI can help PMO and transformation leaders see the operating model more clearly before they try to accelerate it.