This essay first appeared in shorter form on LinkedIn, in a publication called Pivot Point where I have written for several years in a more diagnostic register. What follows is the in-depth version — for a more settled reading and expanded interpretation. The two publications travel on parallel tracks. Same author. Same matrix. Different treatment, for readers who arrive at the question with different needs.

Now to the matter at hand.

The tragedy isn’t that AI is overhyped. It’s that most leaders are too in love with their first instinct about it to notice what kind of exchange it actually requires.

We underestimate the value of a good conversation.

Most of us equate it with being eloquent — with the precision of our words, the strength of our expression, the clarity of our argument. We rarely measure the two things that actually decide whether a conversation worked: how well we listened, and how our words actually landed on the other person.

Conversation, it turns out, is not a soliloquy performed well. It is a symphony conducted well.

Most failed conversations are not failures of vocabulary. They are failures of recognition — a refusal to notice that the person across from us requires a different register than the one we arrived with. The same is true of most failed strategies, most failed marriages, and most failed boards. We are reluctant to abandon the register we have already chosen.

Now consider how most leaders are talking to AI.

Some are giving orders to a machine they don’t trust. Some are asking a machine to tell them what to do because they don’t trust themselves. Some are playing with it like a toy. A few are negotiating with it like a colleague.

Only the last group is creating any real value.

The tragedy of AI in the enterprise today is not that it has been overvalued. It is that it has been miscast. And the cost of that miscasting is paid quietly, quarter after quarter, in initiatives that look like transformation and behave like theater.

The Ego States We Bring to AI

In the 1960s, the psychiatrist Eric Berne proposed a simple model of human interaction. In any conversation — and the model has survived six decades because this is true of every conversation that matters — we occupy one of three ego states: Parent (commanding, judging, authoritative), Adult (analytical, present, negotiating), or Child (curious, dependent, emotional). Healthy conversations match the ego states to the moment. Dysfunctional ones don’t.

The model travels well to AI. Every leader brings an ego state to their AI strategy. And every leader projects an ego state onto the AI itself — whether they realize it or not. The combination determines whether AI creates value, wastes budget, or quietly corrodes decision-making.

Plot the two together and nine archetypes emerge.

The 3×3 of AI conversations — nine archetypes formed by the leader’s ego state and the ego state projected onto AI.

Each of these mismatches has a name. What they share is older than any of them: the executive’s reluctance to let go of the first posture they chose. The hubris of every failed AI strategy I have watched is not arrogance about AI. It is loyalty to a stance the leader took before they understood what they were dealing with.

Four Cells That Matter Most

Four of these cells matter more than the others. Let me walk through them in the order most leaders actually pass through, if they’re lucky enough to graduate at all.

Sandbox Learning (Child + Child) — Necessary, Not Sufficient

A leader new to AI plays with it. They ask it to write a poem, summarize a memo, generate a chart. The AI plays back. There is no agenda, no risk, no commitment. This is where capability is built — and it should be. The best executives I have watched onboard themselves to AI did not commission a strategy. They opened a chat window on a Saturday.

The danger is staying here. Sandbox learning produces curiosity, not outcomes. It is the warm-up, not the workout.

Magical Thinking (Child + Parent) — The Most Dangerous Cell

A leader who has not done the sandbox work, but is under pressure to act, reaches for AI as an oracle. Tell me what to do. Tell the board what to do. Tell the market we are doing it. AI becomes the parent the leader needs because the underlying business question is too hard to face directly.

This is the cell that is quietly failing companies right now — usually in mature, monolithic organizations whose fundamentals are eroding and whose leaders have learned that an AI announcement buys a quarter of grace. The investment is real. The benefit is narrative. The decline continues.

This is hubris in its purest form — not the swagger of certainty, but the refusal to do the work of decomposition. The leader has not asked what AI actually is, what their organization actually needs, what the underlying configuration actually permits. They have skipped the precomputation and arrived directly at the conclusion that AI will save them. Every other failure in the 3×3 is a variation on this one move: arriving at the conclusion before doing the work.

The signature of magical thinking is a leader who can describe the AI vision in detail and cannot describe a single workflow it has changed.

Underutilization (Adult + Child) — The Silent Failure

A capable, analytical leader treats AI as a productivity tool. Better email drafts. Faster slide decks. Cleaner meeting notes. The leader is mature. The relationship is not. AI sits at the periphery of every important decision, never invited into the room where the decision is made.

Look across functions and the pattern repeats. The CFO who uses AI to clean up board materials but never asks it to stress-test the margin bridge. The CRO who lets AI draft sales emails but never asks it to find the deal pattern the team is missing. The VP of Engineering who runs AI on code review but never on design trade-off analysis. Different functions, same failure: AI is allowed near the work, never near the decision.

The signature of underutilization is a leader who uses AI personally and cannot point to a single executive decision where AI changed the outcome.

The Golden Quadrant (Adult + Adult) — The Negotiation

The leader brings judgment, context, accountability, and the strategic frame. The AI brings synthesis, pattern recognition, scale, and tireless iteration. Neither defers. Neither commands. They argue.

The leader challenges an AI recommendation; the AI surfaces a factor the leader missed; the leader integrates it into a decision the leader still owns. This is not human-in-the-loop. This is human-and-machine-in-conversation. It is rare. It is also where every meaningful AI advantage of the next decade will be built.

What this looks like in practice depends on the function — and, more importantly, on the corporate moment.

Take a sales and GTM leader. The conversation with AI looks one way during a spin-off, where the entire motion is being built from a clean sheet — pricing logic, territory design, channel mix all under construction. Here the Adult-to-Adult conversation is generative: the leader brings the strategic intent and the field’s tacit knowledge, AI proposes territory models and segmentation patterns the leader stress-tests against on-the-ground reality. AI is most valuable when it accelerates pattern formation in a vacuum.

Now take the same leader two years later, running the post-spin integration of an acquired adjacent business. The conversation is completely different. AI is no longer filling a vacuum — it is arbitrating between two competing GTM motions, two compensation philosophies, two definitions of “qualified pipeline.” The Adult-to-Adult conversation here is forensic: the leader asks AI to surface where the two motions actually conflict beneath the surface PowerPoint reconciliation, and AI’s job is to find the inconsistencies the integration team has political reasons not to name.

Same leader. Same function. Same AI. Two profoundly different conversations. Most leaders are having only one of them — usually the wrong one for the moment they are in.

The same texture applies elsewhere. The finance leader uses AI to challenge the margin assumption everyone else has stopped questioning. The product engineer lets AI propose three design optimizations and rejects two because the AI doesn’t understand the manufacturing constraint. None of these leaders are deferring to AI. None are commanding it. They are arguing with it — and the decision is better for the argument.

The signature of the golden quadrant is a leader who can describe, specifically, how an AI exchange changed their mind about something that mattered.

The Other Five Cells

The remaining cells are the texture of dysfunction.

Command-and-Control (Parent + Child) is safe and low-ceiling — common in regulated industries where it is sometimes the right answer. Friction (Parent + Adult) is the seasoned leader who hired AI as a colleague and keeps treating it as staff. Cold War (Parent + Parent) is two authorities, neither yielding, AI eventually shelved. Abdication (Adult + Parent) is the data-driven leader who hides behind the model when accountability arrives. Inverted Mentorship (Child + Adult) is useful in private and dangerous in public — a CEO learning from AI is healthy; a CEO led by AI is not.

What to Do on Monday

Three questions every leader should answer honestly before the next AI budget cycle.

Which cell are you actually operating in? Not the cell you describe in the board deck. The cell your behavior reveals.

What is the next cell you need to move to? Most leaders need to graduate from Sandbox Learning to the Golden Quadrant. Many are stuck in Underutilization. A dangerous few are anchored in Magical Thinking and don’t know it.

What conversation are you actually having with AI? If you can’t answer that — if the relationship has no shape, no posture, no register — you are not yet having one.

The leaders who win the next decade will not be the ones who spent the most on AI. They will be the ones who were willing, finally, to let go of the first posture they chose toward it — and to do the slow work of arriving at the right one before the moment demanded it.

Subscribe now