Yesterday, JPMorgan’s Global CIO Lori Beer told Bloomberg that early AI tools are boosting productivity at the bank by as much as 30%. Jamie Dimon has framed AI as a core strategic priority, not a speculative bet, and is backing it with $19.8 billion in technology spending for 2026.
Beer also warned that AI is creating a new leadership challenge. The productivity gains are real, but so are the questions about picking leaders who can navigate this era, maintaining trust, and managing a set of risks that didn’t exist two years ago.
So the question for the rest of us becomes: what does an individual leader actually do to start using AI in a way that changes how they think and decide?
I found a practical answer in Geoff Woods’ book The AI-Driven Leader. And it connects directly to the leadership levels I teach in my Certified Agile Leadership CAL-1 workshops about how leaders operate at different levels of agility and the Transformation case study from Rabobank
The Leadership Agility Problem AI Exposes
In my CAL-1 class, we work through Bill Joiner’s Leadership Agility Model. The research shows that 45% of leaders operate at the “Expert” level, where they rely on their technical knowledge to solve problems themselves. Another 35% reach “Achiever” level, where they think more strategically and drive outcomes through teams. Only about 10% make it to “Catalyst” level, where they challenge assumptions, invite diverse perspectives, and create conditions for others to solve problems collaboratively.
An Expert-level leader treats AI like a search engine. Type a question, get an answer, move on. That’s the least valuable way to use it.
Woods interviewed over 200 leaders while writing the book. 100% believed AI was the future. 100% believed their company would adopt it. Fewer than 5% had done anything about it. They were too busy, they didn’t understand the technology, and nobody had shown them the simple path from zero to one.
The answer is putting AI directly in the hands of leaders so they could experience it as a thinking partner, not a content generator.
From Heroic Leader to Thought Partner
This is where the Joiner model and Woods’ framework reinforce each other. Joiner’s research draws a line between “heroic” leaders (Experts and Achievers who see themselves as ultimately responsible for having the answers) and “post-heroic” leaders (Catalysts who create conditions for better solutions to emerge collaboratively).
Woods draws a similar line. You are the Thought Leader. AI is the Thought Partner. The quality of what you get out of AI depends entirely on the quality of what you put in. Your context. Your judgment. Your questions.
The shift he proposes is one question: “How might AI help me solve this?”
Woods structures the expansion through three AI personas that leaders can assign during a conversation.
Persona 1: The Interviewer
What it does: Asks you targeted questions, one at a time, to pull structured thinking out of your head.
Why it matters: You often have the answer buried in your experience. The Interviewer acts like a skilled journalist who draws out the right information and organizes it for you. This is the AI equivalent of the coaching triads we practice in my CAL-1 workshops, where a coach asks powerful questions to surface insights the client didn’t know they had.
How Rabobank could have used it: When Rabobank shifted to their agile structure, selecting tribe leads became one of the highest-stakes decisions of the transformation. Some candidates who were previously earmarked for senior positions didn’t fit the new leadership vision. Others rapidly advanced. The Managing Board needed each tribe lead to have a specific blend: IT experience, content expertise, stakeholder management skills, execution capability, and the ability to guide large groups of people through ambiguity.
Imagine if a senior leader at Rabobank had used AI as the Interviewer before those selection conversations:
I’m selecting tribe leads for our agile transformation at a large bank. These leaders will manage tribes of 70-100 people organized into BizDevOps squads. I need people who can turn a traditional hierarchy upside down and make the new system work. I would like you to act as an expert interviewer with deep expertise in organizational transformation. Ask me one question at a time to help me clarify:
Here’s how that conversation might unfold. AI starts by asking: “What did your most effective leaders do in the old structure that you believe will still matter in the new one?”
The leader thinks about it and responds: “They built strong relationships across departments and they were good at getting results without needing to escalate.”
AI follows up: “You mentioned results without escalation. In the old structure, what gave those leaders the leverage to do that? Was it their formal authority, their expertise, or something else?”
Now the leader pauses. “Honestly, a lot of it was their tenure and the respect they’d earned over years. People listened to them because of who they were, not their title.” This surfaces something important: tenure-based influence won’t transfer to new tribe leads who are stepping into a structure nobody has operated in before. The new leaders will need to earn influence from scratch, in weeks not years.
AI pushes further: “If tenure-based respect won’t be available to new tribe leads, what alternative signals will their squads use to decide whether to trust this person’s leadership?”
The leader starts listing things: “Probably whether they actually listen in sprint reviews. Whether they remove blockers fast. Whether they admit when they don’t know something.” These are concrete, observable behaviors that never would have appeared on a traditional leadership competency checklist. But they’re exactly what predicts success in a squad-based structure.
After ten minutes of this kind of exchange, AI generates a structured evaluation framework with weighted criteria, suggested interview questions, and red flags to watch for. Because the Interviewer using this conversation pulled out what the board knew and organized into a usable system.
Persona 2: The Communicator
What it does: Takes complex ideas and turns them into clear messages tailored to specific audiences.
Why it matters: Think about how much of your week you spend translating the same idea into different languages for different stakeholders. The board wants financial impact. Your team wants to know what changes. Your customers want to know what’s in it for them. The Communicator handles that translation. In Leadership Agility terms, this is stakeholder agility, the ability to engage effectively across different groups with competing interests.
How Rabobank could have used it: Rabobank’s leaders faced a problem that every transformation leader knows: the workforce was anxious. For years, staff had operated in a top-down bureaucratic structure where decisions came from above. Now leaders were telling them the hierarchy was flattening, there would be fewer formal leadership positions, and teams would need to self-organize into squads of 7-10 people.
That message lands very differently depending on who hears it. A senior manager hears “your role might not exist.” A technical specialist hears “I might have to work with people from marketing now.” A frontline employee hears “nobody is going to tell me what to do anymore, and that’s terrifying.”
Rabobank’s leaders needed to communicate the same transformation in ways that addressed each group’s specific fears and motivations. This is where the SCARF model (Status, Certainty, Autonomy, Relatedness, Fairness) from the Rabobank case study matters. Each stakeholder group felt threatened on different dimensions.
A tribe lead could have used AI as the Communicator:
We’re implementing an agile transformation at a 125-year-old bank. We’re moving from a hierarchical structure to BizDevOps squads of 7-10 people, organized into tribes. Many employees are anxious about losing their roles or their reporting structures. I need to communicate this change to three different audiences: senior managers whose positions are being eliminated, technical specialists who will now work in cross-functional teams, and frontline staff who are used to clear top-down direction. Interview me one question at a time about each audience’s concerns. Once you have enough information, draft three versions of the same core message, each tailored to what that group cares about most.
Watch how this conversation goes deeper than a generic communication template. AI starts: “Let’s begin with the senior managers. How many of them will lose their current title or reporting line, and what are you offering them instead?”
The tribe lead responds: “About 30% will move into product owner or area lead roles. Another 40% will become chapter leads focused on a specific skill area. The remaining 30% honestly don’t have a clear landing spot yet.”
AI asks: “For that 30% without a clear role, what is your timeline for resolving their placement, and what have you told them so far?”
“We haven’t said anything specific yet. We’re waiting until we have answers.” This is the moment AI earns its value. It’s surfaced the fact that silence is the current communication strategy for the most anxious group. Silence from leadership, in a transformation, is never interpreted as neutral. It’s interpreted as bad news the leaders aren’t brave enough to share yet.
AI responds: “Understood. Now, the technical specialists moving into cross-functional squads. What’s their biggest concern: losing depth in their specialty, having to report to someone less technical, or something else?”
The tribe lead thinks about it. “It’s credibility. Our best Java developers have spent years being recognized as the go-to expert in their domain. In a squad, they’ll sit next to a UX designer and a marketing analyst. They’re worried their expertise won’t matter as much.”
This is a Status threat in the SCARF model, and it requires a completely different message than the Certainty threat the senior managers are feeling or the Autonomy concerns of frontline staff.
After walking through all three audiences, AI generates three distinct drafts. The senior manager message leads with what’s being preserved (their expertise, their institutional knowledge, their relationships) before addressing what’s changing. The technical specialist message frames squad membership as expanding their influence, not diluting it. The frontline message focuses on what “self-organizing” actually looks like day to day, with concrete examples of how decisions will get made, because abstract language about empowerment creates more anxiety than it relieves.
The tribe lead now walks into three different conversations with three messages built for how each group actually processes change, instead of one deck that tries to be everything to everyone and ends up resonating with nobody.
Persona 3: The Challenger
What it does: Stress-tests your thinking by questioning assumptions, exposing blind spots, and surfacing second-order consequences.
Why it matters: This is the persona that moves you from Achiever to Catalyst. In my workshops, we talk about how the Ladder of Inference causes leaders to jump from observation to action without questioning the assumptions in between. The Challenger forces you back down the ladder. It asks the questions you aren’t asking yourself.
How Rabobank could have used it: The hardest part of Rabobank’s transformation was the personal shift leaders had to make from controlling “how” work got done to focusing on “what” outcomes the work should produce. The article describes this balance of maintaining oversight while relinquishing direct control as “a major change and often uncomfortable.”
Think about what that actually means in practice. A leader who spent 15 years building credibility by having the right answers now has to walk into a room and say “I don’t know, what do you think?” That’s not just a process change. It’s an identity shift. And most leaders, even ones who intellectually agree with it, will unconsciously recreate control through new mechanisms: overly detailed OKRs, mandatory check-ins disguised as “alignment,” or approval gates dressed up as “quality standards.”
This is where the Challenger would have been valuable. Before rolling out the new leadership operating model, a Rabobank tribe lead could have prompted:
I’m a tribe lead at a bank going through an agile transformation. I’ve been asked to shift from directing my teams on how to do their work to focusing only on what outcomes we need. I believe I’m ready to make this shift. I want you to act as the Challenger. Your job is to stress test my thinking to make sure I’m not only seeing the upside of this leadership shift but that I also see the downside and non-obvious second-order consequences. Ask me one question at a time to challenge my thinking.
AI opens with: “You say you’re ready to focus on outcomes rather than directing how work gets done. Describe the last decision one of your squads made that you disagreed with. What did you do?”
The tribe lead responds: “Two weeks ago a squad decided to skip writing automated tests for a feature because they wanted to hit their sprint commitment. I told them they needed to add the tests.”
AI comes back: “You overrode their decision on the ‘how.’ Was the sprint commitment an outcome you set, or did the squad set it themselves?”
“I set the sprint goals during planning.” The tribe lead starts to see it. He set the outcome (sprint commitment), the squad made a rational choice to meet that outcome (skip tests), and then he overrode the choice because he didn’t like the method. He was controlling both the “what” and the “how” while believing he was only controlling the “what.”
AI goes deeper: “If the squad had shipped without tests and the feature broke in production, whose reputation would take the hit in the next QBR: the squad’s or yours?”
“Mine. I’m accountable for the tribe’s delivery.” And there it is. The structural problem underneath the behavioral one. As long as the tribe lead personally absorbs the reputational consequences of squad decisions, they will keep intervening. The shift from “how” to “what” isn’t just a mindset change. It requires changing accountability structures so that squads own both the decision and its consequences.
AI asks one more: “You mentioned you’re accountable for tribe delivery at the QBR. If a squad fails and you didn’t intervene, how will the Managing Board interpret that? As empowerment or as negligence?”
This question hits a wall the tribe lead hadn’t thought about. The board says they want empowered teams. But if something goes wrong and the tribe lead’s defense is “I let the squad decide,” will the board actually accept that? Or will they quietly mark this leader as someone who lost control?
The Challenger surfaced three layers of problems he didn’t know he had: his own interventionist behavior, the accountability structure that drives it, and the board-level expectations that make the whole shift fragile. That’s twenty minutes of conversation that could have saved months of trust erosion between leaders and squads.
The Rabobank case study notes that leaders had to “plant the seeds for a new organizational culture.” The Challenger helps you figure out whether you’re actually planting new seeds or just replanting the old ones in different pots.
Combining Personas: The Rabobank Quarterly Business Review
The personas become more powerful when you chain them together. Consider what happened when Rabobank replaced traditional steering committees and progress reports with Sprint Reviews every two weeks and Quarterly Business Reviews (QBRs) where tribes resolved conflicting priorities. That’s a massive shift in how decisions get made, and the leaders facilitating those QBRs had to navigate competing demands from squads, areas, and the Managing Board simultaneously.
A tribe lead preparing for their first QBR could chain all three personas in a single session.
Start with the Interviewer to get clarity on the real conflicts: “I have six squads requesting priority for their initiatives next quarter, but I can only fully resource three. Interview me one question at a time about each initiative’s customer impact, technical dependencies, and strategic alignment so I can see which conflicts are real trade-offs versus which ones can be sequenced.”
The Interviewer pulls out that two of the six initiatives actually depend on the same backend service, so they’re not competing at all. They need to be sequenced. That insight alone reduces the decision from six competing priorities to four.
Switch to the Communicator: “Based on what we just discussed, I need to present these trade-offs at the QBR. My audience includes squad product owners who are emotionally invested in their initiatives, area leads who care about cross-squad dependencies, and a Managing Board member who wants to see delivery velocity. Help me frame the same prioritization decision in a way that each group can support.”
The Communicator drafts a narrative where the board member sees strategic focus (fewer initiatives, higher completion rate), the area leads see dependency management (the sequencing decision), and the product owners see that deprioritized work isn’t killed but explicitly scheduled for Q3 with committed capacity.
Then bring in the Challenger: “Before I present this, stress test it. Where could my proposed prioritization backfire? What objections haven’t I prepared for?”
AI responds: “You’ve deprioritized the customer onboarding improvements. If the squad working on that initiative has been promised this would ship by Q2 and you’re now pushing it to Q3, how will the product owner explain that to the commercial team who already committed the feature to a key client?”
The tribe lead didn’t know about that downstream commitment. Now they do, and they can address it before the QBR instead of getting blindsided during it.
Finally, practice the hard conversation by asking AI to role-play as a resistant stakeholder:
Role-play with me as a senior manager who just lost their direct reports to the new squad structure. I’m about to present our QBR priorities, and I expect this person to push back on resource allocation. Simulate their likely objections so I can practice my responses. Afterward, provide feedback on:
That’s four personas in one preparation session. The whole thing takes 30 to 40 minutes. Without AI, this same level of preparation would require separate conversations with a coach, a communications advisor, a strategy partner, and a trusted colleague willing to role-play a hostile stakeholder.
The Counter-Argument Worth Addressing
Some leaders push back on this. The worry is that leaning on AI for strategic thinking means outsourcing your judgment.
Woods addresses this directly: you are the Thought Leader, AI is the Thought Partner. AI doesn’t make decisions. It processes information, asks questions, and generates options. You provide context, judgment, and the final call.
Here is the real challenge: If you ask AI to build on your existing ideas without challenging them, it amplifies your biases instead of countering them. Joe Riesberg from EMC Insurance learned this when his agents used AI to “improve” their customer service ideas. AI stayed within the boundaries of their initial thinking and reinforced their blind spots. The fix was to ask AI to propose alternatives rather than polish what you already have.
This maps to what we cover in the CAL-1 workshops on mental models. If your Ladder of Inference is pulling you toward a conclusion based on selective data, and you ask AI to improve the plan built on that conclusion, AI will make a better version of the wrong plan. You need the Challenger persona to break the loop.
Your Path From Zero to One
JPMorgan has 2,000 AI use cases in production. You need one.
Here’s the simplest prompt from the book to get started:
I’m curious to explore using you as a strategic Thought Partner. I’d like you to interview me by asking one question at a time to identify a simple and valuable use case where you can help me clarify my thinking this week. Then continue interviewing me to help me clarify my thinking.
That’s it. You need to shift one question: from “How can I do this?” to “How might AI help me do this?”
Once you experience that first moment where AI surfaces an insight you didn’t see, or helps you collapse a week of thinking into thirty minutes, you won’t go back. And you’ll have taken your first step from Expert to Catalyst, by practicing it with a thinking partner that’s available any time you are.
Interested in going deep into leadership models and how AI can become your thought partner. Join My Certified Agile Leadership (CAL-1) workshop. I teach one workshop every month.
Inspired by Geoff Woods’ book The AI-Driven Leader, the Leadership Agility Model from Bill Joiner’s Leadership Agility, and the Rabobank agile transformation case study.
