Critical Thinking in the age of AI

PART 1: AI Can Predict, But Only Humans Can Decide

 

In January 2009, US Airways Flight 1549 lifted off from New York’s LaGuardia Airport on what should have been an ordinary, uneventful climb. Minutes later a flock of geese struck the aircraft, disabling both engines. In the cockpit alarms sounded, screens flashed, and the automated systems began calculating the “optimal” return routes to LaGuardia or diversion paths to Teterboro. Air Traffic Control reinforced those options, urging the pilots to follow the system’s recommendations.

But Captain Chesley “Sully” Sullenberger saw something the computers couldn’t: the real‑time loss of altitude, the diminishing airspeed, the narrowing window for action. The system was offering possibilities while he was assessing probabilities. And in that gap between calculation and context, he made the call to override every automated instruction and land the plane on the Hudson River. His ability to judge the situation according to the context meant that all 155 people survived.

That moment wasn’t a triumph of technology, it was a triumph of critical thinking. Sully didn’t reject the system – he interrogated it. He understood its limits. He recognized that in complex, high‑stakes environments, blind trust in automated guidance can be as dangerous as ignoring it altogether.

watch video https://www.youtube.com/watch?v=jQANS8TjsHQ

The Overloaded Engine

Fast forward to today. We’re now facing a similar crossroads with AI. We’re increasingly letting algorithms propose strategies, shape decisions, and define what’s “true”. And we often do that without pausing to ask whether the system sees the full picture.

In global business and particularly in highly regulated sectors like banking, that’s a perilous assumption. Because when the stakes are high, we can’t simply follow the suggested flight path. We have to be like Sully in the cockpit: scanning the data, questioning the model, and knowing when to override the system to steer toward a safer, wiser outcome.

Think of AI as an incredibly powerful engine. It provides immense speed, but when a problem’s complexity overwhelms our “mental payload,” our reasoning doesn’t just slow down it breaks. Like an overloaded plane, our logic collapses under the weight.

You Can Override Autopilot

Critical thinking is your ability to override autopilot when the logic doesn’t hold. In the rush for AI-driven speed, individuals face a choice between being: a passive passenger, who doesn’t need to apply any critical thinking and is just along for the ride; a flight path checker who simply rubber-stamps automated output; or an active pilot who maintains intellectual command.

If your primary value is merely letting AI lead for you or verifying a robot’s opinion, you are quite literally documenting your own obsolescence. To remain relevant, you must shift from being a consumer of AI output to being the “pilot”– the one who ensures that ideas don’t just sound good, but they actually work in practice.

Recognizing the Cognitive Cargo Limit

Critical thinking isn’t just about navigating the journey, it’s also about assessing the capabilities and resources you have before you start the journey. In our research, we have identified this phenomenon as assessing the Cognitive Cargo Limit.

Just as every vehicle is built to meet specifications and has a safe weight limit, every project will have specific limitations and requirements. If you overload “the system” or your own thinking with unfiltered AI data, your project might not even be able to “take off.” When the complexity of the task exceeds the ability to process it critically, a collapse of logic can occur.

This can mean we fall into predictable cognitive traps when dealing with AI, such as:

  • Confirmation Bias: Accepting AI output simply because it agrees with our existing assumptions.

  • The Frictionless Fallacy: Mistaking the speed of the output for the quality of the logic.

  • Information-Processing Shortcuts: Accepting the first “clean” answer as the only answer.

  • Oversimplification traps: Reducing complexity but losing sight of reality – sometimes just paying attention to where the noise is and missing the rest of the picture.

Take Back Control

The goal is to learn how to take control back. By cross-checking your instruments, your weight-bearing capacity, and your navigation approach, you can prepare to navigate the future more sustainably.

The technology provides the horsepower. The question remains: Who is steering?

PART 2: JPMorgan Chase Leadership in the GenAI Era

The Navigation Approach: How the World’s Leading Bank Used A Design Thinking Approach to Implement GenAI (Case Study)

Remember back to early 2023. ChatGPT had just exploded onto the scene, unleashing an unprecedented wave of corporate FOMO. Boards were demanding action, and senior executives everywhere were feeling immense pressure to “move fast and break things.”

At JPMorgan Chase – the world’s largest and most heavily regulated bank – the tension was at a breaking point. With a massive $15B+ technology budget and an appetite for innovation, the temptation to open the throttle and rush the runway was enormous. Yet the risks of data leakage, regulatory noncompliance and algorithmic bias posed a catastrophic threat to the bank’s structural integrity.

While competitors rushed headlong into using public tools, JPMC’s leaders under CEO Jamie Dimon made a counterintuitive executive decision: they paused, restricted public access, and systematically built a secure, closed framework. They realized that integrating Generative AI wasn’t just a tech upgrade, it was a high-stakes “wicked problem.”

JPMC’s ultimate success wasn’t determined by the speed of their computer code; it was determined by the rigor of their human decision-making processes under pressure.

Why This Case Study Matters for Leadership

You do not need to be in global banking to value the strategic takeaways from JPMC’s pivot. Our research highlights that our standard responses to complex, disruptive shifts like GenAI often lead to psychological blocks. The Harvard Case Study on JPMC’s journey can help us understand how to identify blind spots and dismantle these blocks.

In line with Part 1 of this series, we’re again using an aviation metaphor to reveal the key insights. This approach to creative detachment should trigger associative thinking, which involves making meaning through mental connections, and help you look past day-to-day operational noise to analyze the planned strategic “flight path” from a fresh perspective.

Consider how these elements of the aviation metaphor might apply to the strategic implementation of GenAI:

  • The Aircraft: Represents JPMC (the world’s most heavily regulated bank), which must be able to ‘carry the required load’ and to ‘take off’ successfully.

  • The Airspace: The need to navigate a turbulent digital landscape through critical thinking.

  • The Payload: Taking on board a volatile technology like GenAI.

  • The Airframe: Protecting the bank’s structural integrity and infrastructure.

The JPMC case study represents the ultimate “overloaded engine.” With 400+ AI use cases and a $15B+ tech budget, they faced a wicked problem that would be difficult to solve: how to integrate Generative AI to drive innovation without compromising security, ethics, or human-centric leadership?

When GenAI arrived, for example, the temptation for many was to rush the take-off. However, JPMC chose a different path. They followed a disciplined plan, that aligns wiht our Strategies for Innovative Development (SID) Design Thinking Model – which includes the following stages: 1) Enquire (identify the key challenges and frame the approach, 2) Explore (generate novel potential solutions), 3) Solve (test, analyze and prioritize these solutions), and 4) Apply (prototype and prepare for implementation).

Even with the world’s most powerful technological engine, JPMC proved that human piloting is the only way to navigate this frontier without losing altitude or risking the journey.

SID-DT-model

The AI Navigation Approach

Critical thinking can be considered a navigation approach. If your navigation path or “logic guidelines” are headed toward an unforeseen obstacle, resembling the ill-fated US Airways Flight 1549 piloted by Captain Chesley “Sully” Sullenberger, the difference between a catastrophe and a miracle lies in the recalibration framework. To combat this, you can consider the SID Model as a type of “pre-trip inspection framework” as a way of ensuring you stay in command when implementing AI:

  1. ENQUIRE: Question the Frame, Define the Intent. Critical thinking isn’t just about solving puzzles; it’s about questioning how the puzzle is framed. Before the AI “engine” is even on, you must define the intent. What is the actual problem? When you change your perspective on the wording, the impossible becomes more available.

  2. EXPLORE: Use AI as a Sparring Partner for Idea Generation. Treat technology as a sparring partner that sharpens judgment rather than a replacement for it. Challenge AI to find flaws in your logic. This “healthy friction” is what keeps the pilot engaged and alert.

  3. SOLVE: Forensic Evaluation of Potential Solutions. Rigorously analyze the solutions AI provides through the lens of human experience and strategic alignment. A clever idea is cheap; a breakthrough practical design requires disciplined, forensic reasoning.

  4. APPLY: Ensure Human Intent and Oversight Through to Implementation. The final strategic checks and sign-off must be human. Precision in defining the deliverable is the only way to prevent logic collapse. Human intent and oversight are the only ways to prevent professional obsolescence.

Intentional Piloting

JPMC’s approach demonstrates that rather than viewing AI as a simple tool, organizations must apply a disciplined process to evaluate the trade-offs between rapid automation and long-term institutional stability.

In the (initial 1st) Enquire phase, JPMC prioritized high-value problems first to avoid overload from “mental payload,” recognizing that if leaders simply “rubber-stamp” AI output, they risk a systemic logic collapse. In the (2nd) Explore phase, JPMC used GenAI to challenge traditional models, identifying where AI augments judgment and where it introduces risk. In the (3rd) Solve phase, JPMC conducted rigorous analysis to ensure AI-generated code and data are grounded in practice and ethically secure. In the (4th) Apply stage, they maintained a “human-in-the-loop” strategy. As Jamie Dimon advocated, the pilot must always be ready to overrule the autopilot to protect brand value.

The final lesson? The JPMC story proves that human intent and conscious steering are the only ways to navigate the digital frontier without crashing, even with the world’s most powerful AI engine.

Andrew and Dr. Gaia Grant (PhD) are the authors of The Innovation Race and Who Killed Creativity? and specialize in helping organizations navigate the complex tensions of AI culture. The image above captures the Grants facilitating a JPMC Leadership in the GenAI Era case study for global executives within the Harvard Business School partner ecosystem. They specialize in helping organizations navigate the tensions of AI culture.

Disclaimer: This article is for informational purposes only. JPMorgan Chase and Harvard Business School are registered trademarks of their respective owners. Their use does not imply any sponsorship, endorsement, or affiliation with this publication.