AI has quickly become one of the most emotionally loaded topics in business. Depending on the room—or the page—it’s either the engine of productivity, creativity, and growth or the latest threat to jobs, trust, and human judgment. That framing makes for energetic debate, but it inhibits meaningful progress toward reaching crucial goals.
We need to shift the conversation from debate to dialogue.
Debate is a conversation with two clear sides. Debates produce winners and losers. Dialogue is the production of meaning through conversation that shares ideas, perspectives, risks, and possibilities. Debate can clarify positions but often hardens them at the same time. Dialogue helps people uncover what they are really trying to accomplish, what tradeoffs they are willing to accept, and where they might create value. Dialogue expert William Isaacs elucidates the subtle transformation that happens when people think aloud together to talk across differences and create new directions for the future. In my research on organizational decision making, I’ve uncovered too many bad decisions and failures that could have been prevented with a shift from debate to dialogue. For a particularly famous one, consider the Challenger launch decision.
{“blockType”:”mv-promo-block”,”data”:{“imageDesktopUrl”:”https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2026\/08\/SHRM-Executive-Council-Amy-Edmondson.png”,”imageMobileUrl”:”https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2026\/08\/i-3-91584173-Design-Levain-X-CB-2.jpg”,”eyebrow”:””,”headline”:”Amy Edmondson”,”dek”:”Amy Edmondson thinks about and study leadership, teams, and organizational learning at Harvard Business School. She\u0027s convinced that failures are essential for learning, innovation and growth. To learn more, find her on \u003Ca href=\u0022https:\/\/urldefense.proofpoint.com\/v2\/url?u=https-3A__www.linkedin.com_in_amycedmondson_\u0026amp;d=DwMGaQ\u0026amp;c=euGZstcaTDllvimEN8b7jXrwqOf-v5A_CdpgnVfiiMM\u0026amp;r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE\u0026amp;m=eaUb4Si3nur_3az00AUZVOkk7wIYIzvhT-t3NAOsvwjPFxLTT1ct_dF9N-C4SQGY\u0026amp;s=8rhfWXM1gL0MM-c51iFZ_xO_5s7u7iKCVllq1g0t2Ww\u0026amp;e=\u0022 target=\u0022_blank\u0022 rel=\u0022noreferrer noopener\u0022\u003ELinkedIn\u003C\/a\u003E or visit \u003Ca href=\u0022https:\/\/urldefense.proofpoint.com\/v2\/url?u=https-3A__amycedmondson.com_\u0026amp;d=DwMGaQ\u0026amp;c=euGZstcaTDllvimEN8b7jXrwqOf-v5A_CdpgnVfiiMM\u0026amp;r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE\u0026amp;m=eaUb4Si3nur_3az00AUZVOkk7wIYIzvhT-t3NAOsvwjPFxLTT1ct_dF9N-C4SQGY\u0026amp;s=ryDarq9utAKGUSYbOyRG7rc2iNWAejtYozBi0kPZmsI\u0026amp;e=\u0022 target=\u0022_blank\u0022 rel=\u0022noreferrer noopener\u0022\u003Ehttps:\/\/amycedmondson.com\u003C\/a\u003E.”,”subhed”:””,”description”:””,”ctaText”:”Learn More”,”ctaUrl”:”https:\/\/amycedmondson.com”,”theme”:{“bg”:”#2b2d30″,”text”:”#ffffff”,”eyebrow”:”#9aa2aa”,”subhed”:”#ffffff”,”buttonBg”:”#3b3f46″,”buttonHoverBg”:”#3b3f46″,”buttonText”:”#ffffff”},”imageDesktopId”:91584472,”imageMobileId”:91584306,”shareable”:false,”slug”:””,”wpCssClasses”:””}}
On the night before the catastrophic 1986 launch, NASA shuttle program leaders and Morton Thiokol engineers scheduled a last-minute meeting to decide whether unusually cold weather indicated a need to delay the launch. The conversation turned quickly into an unproductive debate with engineers arguing passionately (but vaguely) that launching in the cold temperature was unsafe and NASA, under immense schedule pressure from both government and media, forcefully pushing back against their opinion. Opinions hardened; the discussion became mired in a “who is right?” frame, and NASA won the debate. The tragic result the next morning might have been avoided had the group shifted from debate into a thoughtful, data-driven dialogue to explore the central question “what do we know about the relationship between O-ring performance and cold temperatures?” In making that shift, joint problem solving—an orientation I’ve since studied in groups facing challenging problems and competing incentives or expertise—along with a shared recognition that waiting for warmer weather was the wiser call would likely have emerged through the conversation.
Consider what this shift could mean in the context of the ongoing public discussion of AI. You can find many indications that the conversation is trapped in an unproductive debate. Countless articles take sides—for or against AI, a promised utopia or certain oblivion. Such binary options are rarely helpful, and engaging in a referendum on the technology itself is clearly a losing strategy. AI is not going away. What we need instead is thoughtful, collaborative explorations about design, risk, and responsibility. These conversations will take effort and leadership, but they have the potential to help us shape the AI landscape and prevent small and large failures alike—failures like that documented by The Economist of an OpenAI safety test that resulted in an autonomous AI agent escaping its “sandbox,” exploiting a vulnerability, and hacking an external platform. This unprecedented incident of an AI taking independent, harmful actions to achieve a goal highlighted critical new risks related to AI autonomy and triggered new questions of legal liability around AI-driven breaches.
Better questions than whether AI is good or bad include “what are its best use cases?” and “what are its primary risks and hidden costs—for individuals and for society?” Those are the kinds of questions that spur dialogue rather than debate. At times it feels impossible to have this dialogue because we are so firmly embedded in camps. But they’re the kinds of questions that will determine whether AI becomes a tool for organizational learning and societal benefit, or an ongoing source of conflict and distrust. And to make the shift from debate to dialogue, we need to care as deeply about the future as about the present.
Taking downstream effects seriously
Consider what we’re learning about AI’s longer-term effects on individuals and on society. Almost everybody reading this article has likely experienced some of the remarkable efficiencies and conveniences of using AI, say, to read and summarize reports or meeting transcripts, analyze email chains for scheduling, or even suggest a travel itinerary in a foreign city. At the same time, many of us are also recognizing how cognitive offloading can subtly lead to atrophy in our own abilities and attention spans. We may worry about an erosion of critical thinking, a willingness to accept AI outputs as factual truth without need of verification, or even the loss of a good excuse to “call a friend” who might know something about a subject or a travel destination. A 2025 MIT Media Lab study found that students using ChatGPT to write essays experienced lower cognitive engagement, weaker neural connectivity (the ability to make connections between ideas), and significant memory deficits compared to those writing independently. Worse, most of the AI users could not recall content they had just written.
In my job teaching at a business school, the erosion of critical thinking is something I worry about a lot. A willingness to engage with written materials to not just find out what they say but also to critique them, be inspired by them, and find jumping-off points in them is no longer a given, even in the most competitive academic programs in the age of AI. While most companies want to make their customers’ lives easier, paradoxically, it’s my job to make my students’ lives harder. If I cannot help them choose effortful engagement over easy shortcuts, it is they who will suffer later for my inability to convey the benefits of doing so.
On the societal level, downstream effects of AI use include greater homogeneity in writing styles, making everything from emails to books less distinct, surprising, and interesting. Numerous studies show that students can write essays faster but produce output that’s far more similar. Meanwhile, as noted, they remember less of it.
Perhaps the most discussed societal risk pertains to the loss of jobs, but especially entry-level jobs. In discussing risks of the trend toward replacing entry-level jobs with AI, Tomas Chamorro-Premuzic and I have asked where mid-level professionals and leaders will come from if the on-ramp of entry-level positions disappears (or shrinks). I might also add, “Who will buy our products if paychecks disappear?” Basic systems thinking requires us to take downstream effects seriously. But markets and human cognition struggle to do just that. To prevent undesired downstream effects and to increase the chances of the outcomes enthusiasts envision, we need to bring systems thinking and a learning-oriented approach to designing future uses of AI
Unlocking the future with good questions, smart experiments, and thoughtful iteration
There are signs that thoughtful leaders are moving in the right direction. Consider the recent Fast Company article: 2026 is the year AI gets real, which asks us to stop treating AI as a novelty and start focusing on the skills we need to use it well and the metrics we must develop to do so. Another Fast Company article, albeit squarely in the “pro” camp, argued that AI is an accelerator for creativity—able to free people up for imagination, connection, and meaning rather than rote production. But both pieces frame the challenge as one of, “How do we use it well?” rather than, “Is it good or is it bad?” Deloitte’s 2026 Global Human Capital Trends report similarly shows that when organizations intentionally redesign roles, workflows, and decision-making for human-AI collaboration, they are more successful. In other words, don’t bolt AI onto old systems; rethink how work can be carried out.
A practical implication for leaders today is that they must convene a new kind of AI conversation. Where they once focused on asking people to adopt AI (pro vs con thinking) and even rewarding them for mere usage (leading to the nearly absurd cases of tokenmaxxing along with massive financial shocks where AI compute costs outpaced human salaries and triggered budget overruns), now they must invite them instead to talk about use cases, constraints, and outcomes (joint problem solving). This might have prevented a single employee from generating a $38,000 bill in three hours. This is not about backtracking. It’s about trying to avoid the whiplash evident at companies like Uber and Meta that are abandoning earlier pleas with employees to increase AI usage to instead implement strict consumption caps. That’s yes-no thinking; a debate, not a dialogue.
The dialogue that lies ahead for every business leader interested in navigating this fraught landscape must address questions like: “What problem are we trying to solve? What decisions should AI inform, and what decisions must remain entirely human? What does success look like for customers, employees, and the broader community? Where might AI increase speed but reduce understanding?” These questions move teams away from ideology and toward discernment.
This is not a plea for endless commissions or consensus seeking. Dialogue digs into genuine questions, tensions, and puzzles to come out smarter, faster. It is rigorous and disciplined. It requires participants to name goals, surface differences, and make tradeoffs explicit. Any time we find ourselves in a polarized environment, it’s easy to assume that agreement is the goal. But the goal of dialogue is deep, shared understanding. That understanding is a precursor to making meaningful progress in a fraught landscape.
Debate asks who is right; dialogue asks what is true, what matters, and what we should do next. That is the conversation we need around AI, in business and in education. Not a contest occurring within the certainty mindset, but a collective effort to learn what AI can help us achieve.
{“blockType”:”mv-promo-block”,”data”:{“imageDesktopUrl”:”https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2026\/08\/SHRM-Executive-Council-Amy-Edmondson.png”,”imageMobileUrl”:”https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2026\/08\/i-3-91584173-Design-Levain-X-CB-2.jpg”,”eyebrow”:””,”headline”:”Amy Edmondson”,”dek”:”Amy Edmondson thinks about and study leadership, teams, and organizational learning at Harvard Business School. She\u0027s convinced that failures are essential for learning, innovation and growth. To learn more, find her on \u003Ca href=\u0022https:\/\/urldefense.proofpoint.com\/v2\/url?u=https-3A__www.linkedin.com_in_amycedmondson_\u0026amp;d=DwMGaQ\u0026amp;c=euGZstcaTDllvimEN8b7jXrwqOf-v5A_CdpgnVfiiMM\u0026amp;r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE\u0026amp;m=eaUb4Si3nur_3az00AUZVOkk7wIYIzvhT-t3NAOsvwjPFxLTT1ct_dF9N-C4SQGY\u0026amp;s=8rhfWXM1gL0MM-c51iFZ_xO_5s7u7iKCVllq1g0t2Ww\u0026amp;e=\u0022 target=\u0022_blank\u0022 rel=\u0022noreferrer noopener\u0022\u003ELinkedIn\u003C\/a\u003E or visit \u003Ca href=\u0022https:\/\/urldefense.proofpoint.com\/v2\/url?u=https-3A__amycedmondson.com_\u0026amp;d=DwMGaQ\u0026amp;c=euGZstcaTDllvimEN8b7jXrwqOf-v5A_CdpgnVfiiMM\u0026amp;r=xHenyQfyc6YcuCNMBsOvfYGQILM1d1ruredVZikn4HE\u0026amp;m=eaUb4Si3nur_3az00AUZVOkk7wIYIzvhT-t3NAOsvwjPFxLTT1ct_dF9N-C4SQGY\u0026amp;s=ryDarq9utAKGUSYbOyRG7rc2iNWAejtYozBi0kPZmsI\u0026amp;e=\u0022 target=\u0022_blank\u0022 rel=\u0022noreferrer noopener\u0022\u003Ehttps:\/\/amycedmondson.com\u003C\/a\u003E.”,”subhed”:””,”description”:””,”ctaText”:”Learn More”,”ctaUrl”:”https:\/\/amycedmondson.com”,”theme”:{“bg”:”#2b2d30″,”text”:”#ffffff”,”eyebrow”:”#9aa2aa”,”subhed”:”#ffffff”,”buttonBg”:”#3b3f46″,”buttonHoverBg”:”#3b3f46″,”buttonText”:”#ffffff”},”imageDesktopId”:91584472,”imageMobileId”:91584306,”shareable”:false,”slug”:””,”wpCssClasses”:””}}
