<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Whee Teck Ong on Medium]]></title>
        <description><![CDATA[Stories by Whee Teck Ong on Medium]]></description>
        <link>https://medium.com/@wheeteckong?source=rss-22a11f3431cd------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*PMBMVWw4ZJzp1Bkm.</url>
            <title>Stories by Whee Teck Ong on Medium</title>
            <link>https://medium.com/@wheeteckong?source=rss-22a11f3431cd------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 10:23:19 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@wheeteckong/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Less Than One Percent]]></title>
            <link>https://medium.com/@wheeteckong/less-than-one-percent-28310b4aa40b?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/28310b4aa40b</guid>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[personal-development]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 23:13:44 GMT</pubDate>
            <atom:updated>2026-11-12T23:13:44.071Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sYepPzpj8tNiDCAkOU2VgQ.png" /></figure><p><em>The Judgment Gym · Rep 10 of 10: Build Your Judgment Gym</em></p><p>In the late 1980s, a Swedish psychologist named Anders Ericsson went to a music academy in Berlin to find out what made the best violinists the best.</p><p>He and his colleagues asked the teachers to pick out the students most likely to become international soloists, then compared them with good students who were not quite at that level, and with students training to be music teachers. They asked all of them to keep diaries and to estimate how much they had practised at every age since they started.</p><p>The answer was not talent, and it was not simply time spent with a violin. It was one particular kind of time: practice alone, aimed at the things they could not yet do, with a clear sense of whether it was working. By the age of 18, the best players had put in thousands more hours of it than the future teachers.</p><p>Ericsson called it <em>deliberate practice</em>. A popular book later turned his work into the “10,000-hour rule”, which he spent years saying was not what he found. But the core idea held up, and it is the idea behind this whole series: skill is built by the right kind of reps, not by the passing of years.</p><p>Then, in 2014, someone checked.</p><h4>The surprise in the numbers</h4><p>The psychologists Brooke Macnamara, David Hambrick and Frederick Oswald pulled together 88 studies of deliberate practice and asked a simple question: how much of the difference between people does it actually explain?</p><p>In games such as chess, about 26 percent. In music, about 21 percent. In sports, about 18 percent. Large effects, as these things go.</p><p>In education, about 4 percent.</p><p>In professions, less than 1 percent.</p><p>Read that last line again. For the kind of work most of us do, the practice the studies measured explained almost none of the difference between the people who were good at it and the people who were not.</p><p>The authors offer careful reasons. Professional work is less predictable than a chessboard or a concerto, and practice in it is harder to define and to measure. I accept all of that. But I think there is a simpler reading as well. <em>Most professionals do not practise at all. They just work.</em></p><p>A violinist knows the difference between a performance and a practice session. A chess player knows the difference between a tournament game and studying a position. Most managers, consultants and analysts have only the performance. Every decision is live, the feedback arrives months later if it arrives at all, and nobody goes back to trace the outcome to the call.</p><p>That is why judgment seems to come only with age. It is not that it cannot be trained. It is that almost nobody trains it.</p><h4>The people who kept score</h4><p>There is one field where people did start to train judgment on purpose, and the results were striking.</p><p>The psychologist Philip Tetlock spent nearly two decades collecting predictions from experts on politics and economics, and found that most of them did little better than chance. Then, from 2011, he ran a forecasting tournament funded by the US intelligence community, with thousands of ordinary volunteers.</p><p>A small group, whom he called <em>superforecasters</em>, consistently beat the field, and by a wide margin. They were bright, but they were not geniuses. What set them apart was how they worked. They broke questions into parts. They started from how cases like this usually turn out. They put numbers on their beliefs, changed them in small steps as the evidence came in, and, crucially, they kept score. Every forecast was checked against what actually happened.</p><p>Tetlock’s team also found that a short training session in a few of these habits made ordinary forecasters measurably more accurate.</p><p>That is a gym. Reps on real questions. A score that tells you the truth. And the habit of going back.</p><h4>What is in the gym</h4><p>Over the last nine articles, we have been collecting the equipment.</p><ol><li><a href="https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec">Keep the judgment</a>: decide which thinking you will not hand to the machine.</li><li><a href="https://www.linkedin.com/pulse/free-soda-whee-teck-ong-oizwc">Widen the options</a>: never decide “whether or not” between one option and nothing.</li><li><a href="https://www.linkedin.com/pulse/hungry-judges-whee-teck-ong-dp7yc">Batch the small stuff</a>: protect your best hours for the decisions that matter.</li><li><a href="https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc">Take the outside view</a>: ask how cases like this usually turn out.</li><li><a href="https://www.linkedin.com/pulse/how-series-failed-whee-teck-ong-oi7hc">Run the premortem</a>: imagine it has failed, and ask why.</li><li><a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c">Calibrate your confidence</a>: put a number on it.</li><li><a href="https://www.linkedin.com/pulse/where-did-my-sixty-come-from-whee-teck-ong-nevbc">Spot the anchor</a>: write your own number first.</li><li><a href="https://www.linkedin.com/pulse/twice-heavy-whee-teck-ong-sl5zc">Weigh the loss</a>: notice when fear of losing is weighed on an uneven scale.</li><li><a href="https://www.linkedin.com/pulse/what-my-twin-got-wrong-whee-teck-ong-tjzwc">Check the machine</a>: fluent is not the same as right.</li></ol><p>Equipment is not a gym. Plenty of people own running shoes. What turns equipment into training is a routine, a record and a review date.</p><h4>The rep: your weekly session</h4><p>Thirty minutes, once a week, on real decisions. Three reps.</p><ol><li><strong>One decision, diagnosed before any machine sees it.</strong> Pick a decision you face this week. Before you ask an AI anything, write three lines: what is at stake, which options you considered, and what would prove your view wrong. Then use the machine as much as you like.</li><li><strong>One prediction, logged.</strong> Write down one thing you expect to happen, with a percentage and a date to check it. Keep them all in one place.</li><li><strong>One past decision, traced back.</strong> Open the log, find a decision or prediction whose date has come, and ask: what happened, what did I expect, and was the gap bad luck or bad judgment?</li></ol><p>Then sign it: your name and the date of the next review. <em>If nobody owns the review, the loop never closes.</em></p><p>My own log lives in Obsidian, in the same vault as my twin. I do not write in it myself. My bots do. I give them my judgments, my thinking and my numbers, and they file them.</p><p>And I have asked them to hold me to it. Every Monday, one of them nudges me to sharpen my thinking. When my answer is vague, it pushes back until I reply with more precision.</p><p>The <a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c">60 percent</a> I put on graduate intakes is in there. But that was just the warm-up.</p><h4>If you lead a team</h4><p>The personal gym is enough for one person. If you lead others, you also own the place where they train.</p><p>In September I argued that companies should <a href="https://www.linkedin.com/pulse/dont-bolt-ai-build-judgment-gym-whee-teck-ong-6p5hc">stop bolting AI onto old structures and build a judgment gym</a>. The junior work where judgment used to be learned by accident is the work AI is now taking over. If those reps are disappearing, they have to be designed back in.</p><p>Two rules do most of the work.</p><ul><li><strong>No generation before diagnosis.</strong> On one real workflow, nobody asks the machine for an answer until they have written the stakes, the options and what would falsify their view.</li><li><strong>Owned post-mortems.</strong> When confident nonsense ships and nobody caught it, that is a training failure, owned by a person. Not “the model hallucinated”.</li></ul><p>And name the feeder path: who gets supervised reps on smaller stakes, with an experienced reviewer beside them, before they inherit the big calls.</p><h4>Back to the twin</h4><p>This series began with a twin: the vault of my writing, ten years of articles that an AI can read and search. I said then that <em>the twin keeps my past, and the gym is for my future.</em></p><p>Ten reps later, I would say it differently. It now holds every article in this series, every draft, the list of what it got wrong, and my judgment log. It asks me questions on Monday mornings. It pushes back when I am vague.</p><p><em>It is not just my past any more. It is my sparring partner.</em></p><p>But a sparring partner is not the fighter. The twin can remember every rep and push me through the next one. It cannot do one for me. It has no decision of its own to make, no prediction it will be held to, and nothing to lose by being wrong.</p><p>That is the human premium. Not knowing more than the machine. Being the one who decides, keeps score and goes back.</p><p><em>Judgment is not a gift, and it is not time served. It is reps, and reps can be designed.</em></p><p>This is the last article in the series. The gym stays open.</p><p>So tell me, in the comments: what is the one prediction you will log this week, and when will you check it?</p><h4>References</h4><ul><li>Ericsson, K. A., Krampe, R. T. and Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. <em>Psychological Review</em>, 100(3), 363–406.</li><li>Macnamara, B. N., Hambrick, D. Z. and Oswald, F. L. (2014). Deliberate practice and performance in music, games, sports, education, and professions: A meta-analysis. <em>Psychological Science</em>, 25(8), 1608–1618.</li><li>Mellers, B., Ungar, L., Baron, J. et al. (2014). Psychological strategies for winning a geopolitical forecasting tournament. <em>Psychological Science</em>, 25(5), 1106–1115.</li><li>Ong, W. T. (2026). <em>Don’t Bolt On AI. Build a Judgment Gym.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/dont-bolt-ai-build-judgment-gym-whee-teck-ong-6p5hc">https://www.linkedin.com/pulse/dont-bolt-ai-build-judgment-gym-whee-teck-ong-6p5hc</a></li><li>Ong, W. T. (2026). <em>The Machine Sees the Pattern. You Decide What It Means.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/machine-sees-pattern-you-decide-what-means-whee-teck-ong-amo2c">https://www.linkedin.com/pulse/machine-sees-pattern-you-decide-what-means-whee-teck-ong-amo2c</a></li><li>Ong, W. T. (2026). <em>Twin of What?</em> LinkedIn. <a href="https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec">https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec</a></li><li>Ong, W. T. (2026). The Judgment Gym, articles 2–9 (linked in the text).</li><li>Tetlock, P. E. (2005). <em>Expert Political Judgment: How Good Is It? How Can We Know?</em> Princeton University Press.</li><li>Tetlock, P. E. and Gardner, D. (2015). <em>Superforecasting: The Art and Science of Prediction.</em> Crown.</li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/less-than-one-percent-whee-teck-ong-b6kxc"><em>LinkedIn</em></a><em> on 13 November 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=28310b4aa40b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What My Twin Got Wrong]]></title>
            <link>https://medium.com/@wheeteckong/what-my-twin-got-wrong-d6e519e8607a?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/d6e519e8607a</guid>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[critical-thinking]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Mon, 09 Nov 2026 21:53:29 GMT</pubDate>
            <atom:updated>2026-11-09T21:53:29.685Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ArrLt_4o05B1_aenL2BSSg.png" /></figure><p><em>The Judgment Gym · Rep 9 of 10: Check the Machine</em></p><p>Every article in this series was drafted with my twin, the vault of my writing and the AI that works from it. So before the series ends, I owe you a list. Here are some of the things it got wrong.</p><p>It invented a friend. In the first draft of the first article, a friend “looked at it and asked” the question that became the title. No friend asked. I asked myself.</p><p>It gave me feelings I had not had. One draft said that reading a list of reasons my series might fail “was uncomfortable” and “the most useful twenty minutes I have spent”. I had not said either.</p><p>It described rooms I had not described. A draft about loss aversion said, “I see this in the rooms I teach in.” Maybe I do. But I never said it, and it was about to go out under my name.</p><p>It stated a rhythm I do not keep. The first article promised this series “one a week”, and the second talked about “last week” and “next week”. I publish when I am ready, sometimes twice in a week, and those lines went out before anyone caught them.</p><p>And when it wrote the description for one of my music videos, it said the song was on Apple Music. It had not checked.</p><p>Every one of those sentences was fluent. Every one sounded like me. That is exactly the problem.</p><p>And there is a fresher one, from this week.</p><p>I asked another of my AI assistants to help me track my goals, and to make sure each goal had a clear picture of its outcome. It agreed, and began with a book I am publishing with an artist friend. “Picture the book and cards finished,” it asked. “What do you see and hold in your hands?”</p><p>That was not what I meant. I meant the whole scene: the book done; a workshop to celebrate it, where everyone who comes takes home a copy and two hours of drawing instruction; a joint launch with our animal spirit cards; a pre-launch offer with a free print for early buyers; perhaps a launch with the national library.</p><p>The machine had taken “visualise” quite literally. Its question was fluent, polite and perfectly reasonable. It was also aimed at the wrong thing, and it would have stayed aimed there if I had simply answered it.</p><h4>The lawyer who did not check</h4><p>In 2023, a New York lawyer named Steven Schwartz filed a brief in an ordinary personal injury case against an airline, Avianca. It cited a series of court decisions that supported his client.</p><p>The airline’s lawyers could not find them. Neither could the judge. The cases did not exist. Schwartz had used ChatGPT for his research, and it had produced realistic case names, plausible quotations and convincing citations, all invented. When he asked it whether the cases were real, it said they were.</p><p>The judge sanctioned the lawyers and fined them. The case became famous, but the lesson is not that the tool lied. The tool did what it does: it produced fluent text. The failure was that nobody between the machine and the court asked the boring question. <em>Is this true?</em></p><h4>The jagged frontier</h4><p>The same year, researchers from Harvard Business School, Wharton and other universities ran a large experiment with Boston Consulting Group. More than 750 consultants were given realistic tasks, some with access to GPT-4 and some without.</p><p>On most tasks, the consultants with AI did better and faster. But one task was deliberately chosen to sit just outside what the AI could do well: it needed careful reading of messy data and interviews. On that task, the consultants using AI were noticeably less likely to reach the right answer than those working alone. They trusted a confident answer that happened to be wrong.</p><p>The researchers called this the <em>jagged frontier</em>. AI is excellent at some tasks and poor at others that look almost the same, and the line between them is invisible from the inside. You cannot tell which side you are on by how the answer sounds.</p><p><em>Fluent is not the same as right.</em></p><h4>The Machine Check</h4><p>So what do you do? Not stop using it. That would be the loss aversion of the last article in a new disguise. You check it, the same way, every time.</p><p>The good news is that you already have the tools. Each earlier rep in this series turns into one question.</p><ol><li><strong>Frame.</strong> Did I hand it one option to bless, or ask it to <a href="https://www.linkedin.com/pulse/free-soda-whee-teck-ong-oizwc">widen the options</a>?</li><li><strong>Base rate.</strong> Did it tell me how cases like this <a href="https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc">usually turn out</a>, and where that number comes from?</li><li><strong>Anchor.</strong> Did I write <a href="https://www.linkedin.com/pulse/where-did-my-sixty-come-from-whee-teck-ong-nevbc">my own number</a> before I saw its number?</li><li><strong>Confidence.</strong> Which parts am I <a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c">95 percent sure of</a>, which 60, and which are the kind of thing these tools get wrong?</li><li><strong>Failure.</strong> If I act on this and it <a href="https://www.linkedin.com/pulse/how-series-failed-whee-teck-ong-oi7hc">fails a year from now</a>, why did it fail?</li></ol><p>Then mark every claim in the answer with one of three words.</p><ul><li><strong>Accept:</strong> I would defend this myself.</li><li><strong>Verify:</strong> plausible, but I need a source before I rely on it.</li><li><strong>Override:</strong> I am choosing differently, and I can say why.</li></ul><p>And one rule sits above all of it. <em>If you sign it, it is yours.</em> “The AI said so” is not a reason you can give a client, a boss or a court.</p><h4>The rep: ten minutes on one answer</h4><p>Take one AI answer you are about to rely on this week: a recommendation, an analysis, a summary, a draft.</p><ol><li><strong>Write your cold verdict.</strong> Before checking anything: follow it, follow it with changes, or don’t follow it? How confident are you, as a percentage?</li><li><strong>Run the five questions.</strong> Write a one-line answer to each.</li><li><strong>Mark every claim.</strong> Accept, Verify or Override.</li><li><strong>Check the Verify pile.</strong> Find a source for each one, or drop it.</li><li><strong>Write your checked verdict and your new percentage.</strong></li></ol><h4>Back to the book</h4><p>When the assistant asked me what I could hold in my hands, the easy thing would have been to answer it. A cover, a stack of cards, a nice photograph. It would have recorded a perfectly good goal, and it would have been the wrong one.</p><p>Instead I told it what I actually meant.</p><p>That is the whole skill. Treat every answer as a first draft from a fast, tireless colleague who has never met you. Check it before it goes out under your name.</p><p><em>The machine drafts. You decide.</em></p><h4>References</h4><ul><li>Dell’Acqua, F., McFowland, E., Mollick, E. R. et al. (2023). <em>Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality.</em> Harvard Business School Working Paper 24–013.</li><li>Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023).</li><li>Ong, W. T. (2026). <em>Almost Complete.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc">https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc</a></li><li>Ong, W. T. (2026). <em>How Sure Was I?</em> LinkedIn. <a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c">https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c</a></li><li>Ong, W. T. (2026). <em>How This Series Failed.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/how-series-failed-whee-teck-ong-oi7hc">https://www.linkedin.com/pulse/how-series-failed-whee-teck-ong-oi7hc</a></li><li>Ong, W. T. (2026). <em>The Free Soda.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/free-soda-whee-teck-ong-oizwc">https://www.linkedin.com/pulse/free-soda-whee-teck-ong-oizwc</a></li><li>Ong, W. T. (2026). <em>Where Did My Sixty Come From?</em> LinkedIn. <a href="https://www.linkedin.com/pulse/where-did-my-sixty-come-from-whee-teck-ong-nevbc">https://www.linkedin.com/pulse/where-did-my-sixty-come-from-whee-teck-ong-nevbc</a></li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/what-my-twin-got-wrong-whee-teck-ong-tjzwc"><em>LinkedIn</em></a><em> on 10 November 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d6e519e8607a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Twice as Heavy]]></title>
            <link>https://medium.com/@wheeteckong/twice-as-heavy-2b053cf52279?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/2b053cf52279</guid>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[careers]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 21:04:30 GMT</pubDate>
            <atom:updated>2026-11-05T21:04:30.325Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SAntE1UgHMpFolAq5Fx31Q.png" /></figure><p><em>The Judgment Gym · Rep 8 of 10: Weigh the Loss</em></p><p>In January 2024 I published a short article with a cheerful headline: <em>Peak Anxiety? Stepping Down Might Be Your Next Power Move in Tech.</em></p><p>It told readers that leaving at the top of your career was not surrender but “strategic evolution”. It quoted Jürgen Klopp walking away from Liverpool at the height of his success. It was confident, upbeat and very sure of itself.</p><p>My twin, the vault of my writing and the AI that reads it, noticed something about that article while I was preparing this one. Three months after I published it, in April 2024, I left my firm.</p><p>So the article was not really advice to readers. It was me, talking myself into taking the leap.</p><p>And every fear on my list was true. I would lose the title. I would lose the income. I would lose the support of a large organisation behind me. But the one that weighed most was none of those. It was identity. For three decades I had been someone at somewhere, and I did not know who I would be without the somewhere.</p><p>The strange part is that I was ready. It was time, and I knew it. Even so, the fear of the unknown sat on the scale like a weight twice its size.</p><p>Once I was out, what I felt was relief. A sense of freedom. Freedom to choose what to do next, to create, to write, to have coffee with friends with no business purpose attached.</p><p>None of that had been on the scale when I was deciding. The losses were vivid and specific. The gains were vague, so they weighed almost nothing.</p><h4>The mugs</h4><p>In the late 1980s, the economists Daniel Kahneman, Jack Knetsch and Richard Thaler ran an experiment with coffee mugs.</p><p>They gave mugs to half the students in a classroom. The other half got nothing. Then they asked the owners the lowest price at which they would sell, and the others the highest price at which they would buy.</p><p>The mugs were identical. Nobody had owned one for more than a few minutes. By any sensible logic the two prices should have been about the same.</p><p>They were not. The owners asked for roughly twice what the buyers would pay. A few minutes of ownership had doubled the mug’s value in its owner’s eyes.</p><p>They called it the <em>endowment effect</em>. Underneath it is something Kahneman and Amos Tversky had described a decade earlier: <em>loss aversion</em>. A loss hurts more than an equal gain pleases. Across many studies, the ratio has tended to settle at around two to one.</p><h4>Prudence in disguise</h4><p>Once you see loss aversion, you see it everywhere, and nearly always pushing the same way: towards keeping what you have.</p><p>It is why investors hold on to falling shares, because selling would make the loss real. It is why organisations keep projects running long after they should have stopped: walking away means admitting the money is gone. It is why “let’s wait and see” so often wins the meeting.</p><p>The dangerous part is that it rarely feels like a bias. It feels like prudence. Nobody says “I am overweighting the downside by a factor of two.” They say “let’s not be reckless”.</p><p>An honest note. In recent years some researchers, notably David Gal and Derek Rucker, have argued that loss aversion has been overstated, and that the famous two-to-one ratio depends heavily on the situation and sometimes disappears. That debate is real. In the spirit of <a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c">the article on calibration</a>, I would put it this way: the tendency is strong and common, and the exact ratio is less certain than the textbooks suggest.</p><h4>The loss everyone is feeling now</h4><p>This matters more than usual right now, because most people meet AI as a loss.</p><p>What will it take from me? Which of my skills are about to be worth less? Which part of my job disappears next? Loss aversion makes every one of those threats feel about twice as large as the matching opportunity. The natural response is to defend the old way of working, and to treat the years spent building a skill as a reason not to change it.</p><p>So the first question most people ask is what they will lose. The more useful question, what they stand to gain, tends to come much later, if it comes at all.</p><p><em>The fear of loss is real. It is just weighed on a scale that is not level.</em></p><h4>The rep: reweigh one decision</h4><p>Pick one decision you are putting off: a change at work, a project you are reluctant to stop, a new tool you are avoiding, a role you are thinking of leaving.</p><ol><li><strong>Write down what you would lose.</strong> Be specific: money, status, comfort, identity.</li><li><strong>Write down what you would gain.</strong> Be just as specific. This list is usually shorter at first. Keep going until it is not.</li><li><strong>Halve the losses.</strong> Not literally, but ask of each one: would this still feel this heavy if I did not already have it?</li><li><strong>Ask the sunk-cost question.</strong> If I were starting from zero today, with no history, would I choose to be where I am?</li><li><strong>Ask the machine for the gains.</strong> Have it list what you could gain from the change. It is not loss-averse on your behalf, and that is useful here.</li></ol><p>Then copy this into the comments:</p><blockquote><strong><em>The decision I have been putting off:</em></strong><em> …</em></blockquote><blockquote><strong><em>The loss I was overweighting:</em></strong><em> …</em></blockquote><blockquote><strong><em>The gain I had not counted:</em></strong><em> …</em></blockquote><h4>Back to January</h4><p>Reading that 2024 article now, I see what it really was. Not a power move. A man reweighing a scale that had tipped too far towards what he might lose.</p><p>The title, the income and the organisation were real losses. The identity took longest to let go. But the gains I could not see from the inside, the freedom and the time and the coffee with no agenda, turned out to weigh more than all of them.</p><p>I just could not put them on the scale until I was holding them.</p><p>Next in the series: how to check the machine before you trust it.</p><p>In the meantime, post your three lines below. What is a loss you once feared that turned out lighter than you expected?</p><h4>References</h4><ul><li>Gal, D. and Rucker, D. D. (2018). The loss of loss aversion: Will it loom larger than its gain? <em>Journal of Consumer Psychology</em>, 28(3), 497–516.</li><li>Kahneman, D., Knetsch, J. L. and Thaler, R. H. (1990). Experimental tests of the endowment effect and the Coase theorem. <em>Journal of Political Economy</em>, 98(6), 1325–1348.</li><li>Kahneman, D. and Tversky, A. (1979). Prospect theory: An analysis of decision under risk. <em>Econometrica</em>, 47(2), 263–291.</li><li>Ong, W. T. (2024). <em>Peak Anxiety? Stepping Down Might Be Your Next Power Move in Tech.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/peak-anxiety-stepping-down-might-your-next-power-move-whee-teck-ong-yuerc">https://www.linkedin.com/pulse/peak-anxiety-stepping-down-might-your-next-power-move-whee-teck-ong-yuerc</a></li><li>Ong, W. T. (2026). <em>How Sure Was I?</em> LinkedIn. <a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c">https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c</a></li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/twice-heavy-whee-teck-ong-sl5zc"><em>LinkedIn</em></a><em> on 6 November 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2b053cf52279" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Where Did My Sixty Come From?]]></title>
            <link>https://medium.com/@wheeteckong/where-did-my-sixty-come-from-34998294f6da?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/34998294f6da</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[negotiation]]></category>
            <category><![CDATA[decision-making]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 20:41:05 GMT</pubDate>
            <atom:updated>2026-11-05T20:41:05.745Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5xlZFE0HC2f2YJlCmn0Vzw.png" /></figure><p><em>The Judgment Gym · Rep 7 of 10: Spot the Anchor</em></p><p>In <a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c">the last article in this series</a>, I put a number on a prediction. I said that about 60 percent of firms will shrink their graduate intakes, and that the share would rise.</p><p>I meant it. But I owe you a footnote on where that number came from.</p><p>When I was writing the piece, my twin, the vault of my writing and the AI that helps me draft, asked me for the calibrated version of my old headline. To help, it offered a couple of example sentences. One began, “I am about 70 percent sure that most large firms will keep shrinking their graduate training intakes.” Another split the claim, at 80 percent and 20 percent.</p><p>I read them, thought about it, and said 60.</p><p>Looking back, the example pulled mine. I would like to say I reasoned my way to 60 from first principles. I did not. I started near 70 and adjusted down, which is exactly how anchoring works.</p><p>I have been on the other end of this too. Some time ago I was negotiating the fee for a workshop. The person on the other side told me she had already got approval for a figure. It was 30 percent below my published rates. She did not make an offer. She announced a number, with an approval attached, and from then on every number in the conversation was measured against hers.</p><p>The deal closed at exactly her number. I have since run two of the three workshops, the client is happy, and so am I. But I know whose number it was.</p><p>That is how you negotiate. I could only admire it.</p><h4>The wheel of fortune</h4><p>In the early 1970s, Amos Tversky and Daniel Kahneman ran one of the strangest experiments in the study of judgment.</p><p>They spun a wheel of fortune in front of their subjects. The wheel was rigged to stop at either 10 or 65. Then they asked a question that had nothing to do with the wheel: what percentage of the countries in the United Nations are African?</p><p>People who had just seen the wheel stop at 10 gave a median answer of 25 percent. People who had seen it stop at 65 gave a median answer of 45.</p><p>A number they knew to be random, from a fairground wheel, moved their estimate of a real fact by 20 points.</p><p>Tversky and Kahneman called this <em>anchoring</em>. When we face an uncertain number, we grab whatever number is already in front of us and adjust away from it. And we nearly always adjust too little. We stay inside the anchor’s pull.</p><h4>The first offer</h4><p>Thirty years later, the psychologists Adam Galinsky and Thomas Mussweiler took the idea into negotiation. They set people to bargain, over the sale of a factory among other things, and varied one thing: who named the first number.</p><p>Whichever side made the first offer, buyer or seller, tended to get the better deal. In one of their studies, the opening offer and the final price moved so closely together that the first number largely decided where the deal landed.</p><p>That overturns the old advice to “never go first”. Waiting does not keep the room neutral. It hands the other side the anchor, and you spend the rest of the conversation working against its pull.</p><p>They also found the antidote. People who received a first offer and then deliberately thought about the numbers that pointed the other way, such as their own target, their walk-away point and the other side’s alternatives, were much less affected by it. The defence against an anchor is a <em>counter-anchor</em>.</p><h4>The machine is an anchor engine</h4><p>This is why anchoring matters more now than it did when the wheel was spinning.</p><p>Ask an AI for an estimate, a price, a budget, a timeline or a salary range, and it will give you a number, instantly and with complete confidence. That number anchors you exactly as an opening bid would. But it does not feel like a negotiation, so your guard is down.</p><p>The machine is not trying to win. It does not need to. Its number simply becomes the centre of gravity for everything you think next, and you adjust from it too little.</p><p>Which is what makes my own example so uncomfortable. The machine offered me an illustration, not a bid. It was trying to help. And I still had to ask myself afterwards where my number came from.</p><p><em>Treat the machine’s number as a first offer, not a fact.</em></p><h4>The rep: your number first</h4><p>Three habits, starting with the next number you need.</p><ol><li><strong>Write your number first.</strong> Before you ask anyone, or any machine, for an estimate, write down your own. Even a rough one. Then ask.</li><li><strong>Counter-anchor every number you are handed.</strong> When a figure arrives, from a counterpart, a vendor or an AI, write down three numbers that pull the other way: your target, your walk-away and the best alternative.</li><li><strong>Go first when you know the value.</strong> In your next negotiation, if you have a defensible sense of the price, name it. Make it ambitious, and make sure you can explain it.</li></ol><p>Then copy this into the comments:</p><blockquote><strong><em>The number I was handed:</em></strong><em> …</em></blockquote><blockquote><strong><em>My own number, written before I looked:</em></strong><em> …</em></blockquote><blockquote><strong><em>How far apart they were:</em></strong><em> …</em></blockquote><h4>Back to sixty</h4><p>So, where did my sixty come from?</p><p>I still think the direction is right. But the number was not entirely mine, and that is the point. Anchors do not feel like anchors. They feel like your own judgment, slightly adjusted.</p><p>The only reliable defence is order. Write your own number before you see anyone else’s, including the machine’s. Then the anchor in the room is yours.</p><p>Next in the series: why losing feels about twice as heavy as winning, and how that disguises itself as prudence.</p><p>In the meantime, post your three lines below. When did the first number in the room decide the outcome?</p><h4>References</h4><ul><li>Galinsky, A. D. and Mussweiler, T. (2001). First offers as anchors: The role of perspective-taking and negotiator focus. <em>Journal of Personality and Social Psychology</em>, 81(4), 657–669.</li><li>Ong, W. T. (2026). <em>How Sure Was I?</em> LinkedIn. <a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c">https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c</a></li><li>Tversky, A. and Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. <em>Science</em>, 185(4157), 1124–1131.</li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/where-did-my-sixty-come-from-whee-teck-ong-nevbc"><em>LinkedIn</em></a><em> on 4 November 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=34998294f6da" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Sure Was I?]]></title>
            <link>https://medium.com/@wheeteckong/how-sure-was-i-f29f921252e3?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/f29f921252e3</guid>
            <category><![CDATA[critical-thinking]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[forecasting]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[leadership]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Mon, 02 Nov 2026 01:13:21 GMT</pubDate>
            <atom:updated>2026-11-02T01:13:21.086Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fUGiixtE3QmY4HPxxjCmkg.png" /></figure><p><em>The Judgment Gym · Rep 6 of 10: Calibrate Your Confidence</em></p><p>In September last year I published an article with this headline: <em>Why Corporate Training Programs Are Disappearing Forever.</em></p><p>It was the second in a series called The Job Apocalypse. The first one had opened with a line I was rather pleased with at the time. “The headline isn’t hyperbole,” I wrote. “It’s mathematics.”</p><p>My twin, the vault of my writing and the AI that reads it, found both of those sentences while I was preparing this piece. Then it put them next to something else in my files: the folder where I keep the corporate workshops I run. In the year since I declared corporate training dead, that folder has kept filling up with new corporate programmes. I am in the middle of one of them this month.</p><p>My verdict, a year on, is that I was partly right. The pressure on junior roles and on the old model of training them was real, and it still is. But “forever” overstated it. Companies are still training their people. I am one of the people they hire to do it.</p><p>What bothers me more is this. I do not usually think in yes or no. With clients, in meetings, in my own planning, I think in probabilities. Somewhere between my head and the headline, the probability fell out.</p><p>This is not an article about whether I was right. It is about the word <em>forever</em>, and what it tells you about how sure I was.</p><h4>The dart-throwing chimpanzee</h4><p>For two decades, from the late 1980s, the political scientist Philip Tetlock ran one of the most patient experiments in social science. He recruited hundreds of experts, the kind of people who appear on television explaining what will happen next, and asked them to make specific, checkable predictions about politics and economics. Then he waited, and scored them.</p><p>The result became famous for one phrase. On average, the experts did about as well as a dart-throwing chimpanzee.</p><p>The detail that matters more is this. The most confident experts, the ones with the boldest and most quotable certainties, tended to do worst. A clear, confident story made them more memorable and less accurate at the same time.</p><p>Tetlock did not stop there. In a later project with the psychologist Barbara Mellers, the Good Judgment Project, thousands of volunteers made forecasts on world events. A small group, around the top two percent, turned out to be consistently and remarkably good. Tetlock called them <em>superforecasters</em>. In one widely reported comparison, they beat intelligence analysts who had access to classified information by about 30 percent.</p><p>They were not geniuses and they had no secret information. What they had was a habit. They were <em>calibrated</em>.</p><h4>What calibration means</h4><p>Calibration is simple to define and hard to do. It means that the things you are 70 percent sure of happen about 70 percent of the time, and the things you are 90 percent sure of happen about 90 percent of the time.</p><p>Most of us are badly calibrated, and almost always in the same direction. We are certain far more often than we are right.</p><p>The superforecasters worked differently. They spoke in numbers, not in “definitely” and “never”: not “likely” but 65 percent, then 70 when new evidence came in. They changed their minds in small steps instead of lurching from one certainty to another. And they began with the base rate, the outside view from <a href="https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc">the article on my bookshelf</a>, before they looked at the details of the case.</p><p>The hopeful part is that it can be taught. In the Good Judgment Project, even a short training module in probabilistic thinking measurably improved people’s accuracy.</p><p><em>Forever</em> is not a probability. It is a feeling dressed up as a fact.</p><h4>The machine has no calibration</h4><p>This is why calibration may be the most valuable thinking skill of the AI age.</p><p>An AI writes every answer in the same calm, fluent, confident voice. It sounds exactly as sure when it is right as when it is inventing a citation that does not exist. It gives you no honest signal of how much to trust any particular sentence.</p><p>So the signal has to come from you. The person who can read a confident answer and say <em>this part is almost certainly right, this part I would put at 60 percent and check, and this is exactly the kind of claim these systems get wrong</em> is doing the job the whole system depends on.</p><p>That is what “the human in the loop” means in practice. You are not there to do what the machine does, only slower. You are there to supply the confidence the machine cannot.</p><h4>The rep: three predictions and one answer</h4><p>Two short exercises and one check. Start both exercises before the next rep.</p><ol><li><strong>Write three predictions with numbers.</strong> Things you will know the answer to within a month: a deal closing, a project date, a hire saying yes. Give each a percentage, not a word. Write today’s date and the check date next to them.</li><li><strong>Score one AI answer.</strong> Take an answer the machine gave you this week. Before you check anything, mark each claim in it with your own confidence: 95, 70 or 50 percent. Then check the claims you marked lowest first.</li><li><strong>Check on the date.</strong> When the month is up, see how the 70s did against the 90s.</li></ol><p>Then copy this into the comments:</p><blockquote><strong><em>I am 70 percent sure that:</em></strong><em> …</em></blockquote><blockquote><strong><em>I will know by:</em></strong><em> …</em></blockquote><blockquote><strong><em>The AI claim I trusted least this week:</em></strong><em> …</em></blockquote><p>I will come back to this post in a month and ask how your predictions did.</p><h4>Back to forever</h4><p>So, how sure was I?</p><p>Reading that headline now, I can see exactly what happened. I had a real observation: entry-level roles shrinking, training budgets under pressure, AI doing the first-draft work that used to train juniors. Then I attached the most certain word in the language to it, because certainty makes a better headline.</p><p>A calibrated version would have been less quotable. It would also have been more useful, to my readers and to me.</p><p>So here is the version I would write today. I think about 60 percent of firms will shrink their graduate intakes, and I expect that share to rise as the evidence comes in. I will update it as it does, in small steps.</p><p>Not forever. Sixty percent, and rising.</p><p>Next in the series: why the first number in the room, including the one the machine gives you, drags every other number towards it.</p><p>In the meantime, post your three lines below. What is a prediction you once made with total confidence?</p><h4>References</h4><ul><li>Mellers, B., Stone, E., Murray, T. et al. (2015). Identifying and cultivating superforecasters as a method of improving probabilistic predictions. <em>Perspectives on Psychological Science</em>, 10(3), 267–281.</li><li>Ong, W. T. (2025). <em>The Job Apocalypse #2: Why Corporate Training Programs Are Disappearing Forever.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/job-apocalypse-2-why-corporate-training-programs-disappearing-ong-gegyc">https://www.linkedin.com/pulse/job-apocalypse-2-why-corporate-training-programs-disappearing-ong-gegyc</a></li><li>Ong, W. T. (2026). <em>Almost Complete.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc">https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc</a></li><li>Tetlock, P. E. (2005). <em>Expert Political Judgment: How Good Is It? How Can We Know?</em> Princeton University Press.</li><li>Tetlock, P. E. and Gardner, D. (2015). <em>Superforecasting: The Art and Science of Prediction.</em> Crown.</li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/how-sure-i-whee-teck-ong-vxa3c"><em>LinkedIn</em></a><em> on 2 November 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f29f921252e3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How This Series Failed]]></title>
            <link>https://medium.com/@wheeteckong/how-this-series-failed-65dab97bba84?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/65dab97bba84</guid>
            <category><![CDATA[project-management]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[risk-management]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Sun, 25 Oct 2026 22:58:33 GMT</pubDate>
            <atom:updated>2026-10-25T22:58:33.645Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hgpozn_JVvW5EGnAo7OutQ.png" /></figure><p><em>The Judgment Gym · Rep 5 of 10: Run the Premortem</em></p><p>It is the end of March next year, and The Judgment Gym has failed.</p><p>Not dramatically. Nobody complained. The articles just stopped being read somewhere around the seventh one, the last three were never written, and by March I had quietly moved on to something else.</p><p>That is not a prediction. It is an exercise. Before I wrote a word of this piece, my twin, the vault of my writing and the AI that reads it, was asked to assume exactly that and explain why it happened. It had the whole record to work from: the four articles so far, the dates they went out, the plan for the remaining six, and everything I have written about the series along the way.</p><p>It gave me six reasons. I will show you all of them, including the ones I did not enjoy reading. But first, why ask the question that way at all.</p><h4>Thirty percent more</h4><p>In 1989, three researchers, Deborah Mitchell, Jay Russo and Nancy Pennington, ran a set of experiments on something they called <em>prospective hindsight</em>.</p><p>They asked people to imagine a future event and explain it. Some were told the event <em>might</em> happen. Others were told it <em>had</em> happened, as a settled fact, and asked why. The second group came up with about 30 percent more reasons.</p><p>The only thing that changed was the grammar. “Might fail” lets the mind stay vague. “Did fail” sends it looking for specifics.</p><p>The cognitive psychologist Gary Klein turned this into a simple meeting ritual. He called it the <em>premortem</em>. A postmortem, he liked to point out, is very useful to everyone except the patient. A premortem moves the autopsy to before the death.</p><p>The team gathers before committing to a plan. Someone says: imagine it is a year from now and this project has failed completely. Take a few minutes, on your own, and write down every reason why. Then go round the room.</p><p>It works for two reasons at once. The first is the prospective hindsight. The second is social, and I think it matters more. At the launch of an exciting project, raising a real doubt feels disloyal. So people offer a thin, polite list of generic risks and keep their real worries to themselves. The premortem turns doubt into the assigned task. The person who privately thinks the timeline is fantasy now has cover to say so, because everyone in the room is doing the same thing.</p><p>When the researcher Beth Veinott and her colleagues later compared the premortem with other ways of testing a plan, it lowered people’s confidence in the plan about twice as much as listing pros and cons did.</p><h4>Why this matters more now</h4><p>I have written in this series that <a href="https://www.linkedin.com/pulse/hungry-judges-whee-teck-ong-dp7yc">fluency is not evidence</a>. The premortem is the practical defence.</p><p>AI will increasingly hand you plans, strategies and recommendations, and it will hand them over beautifully: structured, confident, full of plausible detail. A rough, hesitant plan invites questions. A polished one invites a signature. <em>The better a plan looks, the less we tend to interrogate it</em>, which is exactly the wrong way round.</p><p>But here is the twist. The machine is also very good at the premortem itself. Ask it to assume your plan has failed and explain why, and it will produce failure modes faster and more bluntly than most teams will in a meeting. It has no career to protect and no wish to look loyal.</p><p>What it cannot do is tell you which of those failures is likely, or what you are prepared to do about it. That part stays with you.</p><h4>My twin’s six reasons</h4><p>So here is the list it gave me, unedited.</p><ol><li><strong>Readers lost the thread.</strong> The articles came out irregularly, sometimes two in a week and then a long gap, and a ten-part series needs some rhythm to hold an audience.</li><li><strong>It became about the author.</strong> The twin, the bookshelf and the workshop units made the articles more about me than about the reader’s own decisions.</li><li><strong>The reps were not done.</strong> Each article ended with an exercise, but nothing made readers come back to say what happened when they tried it.</li><li><strong>The book never shipped.</strong> The series kept pointing to a book “currently being written”, and the book stayed in a folder called “Almost Complete”.</li><li><strong>The day job won.</strong> A full-time role and a keynote in early November pushed the writing aside.</li><li><strong>The machine wrote it.</strong> A series about not outsourcing your judgment was drafted with an AI, and readers noticed, or I let it do too much.</li></ol><p>Two of these ring true.</p><p>The first is rhythm. I publish when an article is ready, and that has meant two in one week and then a wait. A ten-part series needs better than that. So from now on The Judgment Gym goes out on Monday and Friday mornings, with a simple process behind it, so the date does not depend on my mood.</p><p>The third is the reps. Most people will read the exercise at the end of an article and move on. That is my problem to fix, not the reader’s. So the reps are getting tighter and more practical, starting with this one, which ends with a template you can fill in.</p><p>The sixth I am keeping on the list, because it is true and because it is the point of the series. These articles are written with a great deal of input: from my digital twin, which holds ten years of my writing, and from the AI that reads it. Every one is then checked, changed and approved by me before it goes out. That is the human in the loop. The machine supplies the inputs. The judgment, and the decision to publish, stay with me.</p><h4>The rep: twenty minutes before you commit</h4><p>Pick one plan you are about to approve this week: a project, a hire, a launch, or a recommendation an AI has just handed you. Then do this.</p><ol><li><strong>Say it.</strong> “It is a year from now, and this failed completely.”</li><li><strong>Write alone first.</strong> Five minutes. Everyone involved lists every reason they can think of, without discussing.</li><li><strong>Ask the machine too.</strong> Give it the same sentence and add its list to the pile.</li><li><strong>Mark the odds.</strong> High, medium or low against each reason. No more precision than that.</li><li><strong>Fix the top one today.</strong> Change one thing in the plan before you launch, not after you sink.</li></ol><p>Then copy this into the comments and fill it in. I will read every one.</p><blockquote><strong><em>My plan:</em></strong><em> …</em></blockquote><blockquote><strong><em>It failed because:</em></strong><em> 1. … 2. … 3. …</em></blockquote><blockquote><strong><em>The one I am fixing now:</em></strong><em> …</em></blockquote><p>The machine can supply the reasons. Only you can supply the odds and the fix.</p><h4>Back to March</h4><p>So that is the autopsy, held five months early.</p><p>The series has not failed. It may yet, for one of those six reasons or a seventh nobody thought of. But it is now a little less likely to fail for the reasons I have written down here, in public, with dates against them.</p><p>Next in the series: how sure should you be, and how would you know?</p><p>In the meantime, run it on something real, and post your three lines below. Which reason did you least want to write down?</p><h4>References</h4><ul><li>Klein, G. (2007). Performing a project premortem. <em>Harvard Business Review</em>, 85(9), 18–19.</li><li>Mitchell, D. J., Russo, J. E. and Pennington, N. (1989). Back to the future: Temporal perspective in the explanation of events. <em>Journal of Behavioral Decision Making</em>, 2(1), 25–38.</li><li>Ong, W. T. (2026). <em>Almost Complete.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc">https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc</a></li><li>Ong, W. T. (2026). <em>The Hungry Judges.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/hungry-judges-whee-teck-ong-dp7yc">https://www.linkedin.com/pulse/hungry-judges-whee-teck-ong-dp7yc</a></li><li>Veinott, B., Klein, G. A. and Wiggins, S. (2010). Evaluating the effectiveness of the PreMortem technique on plan confidence. <em>Proceedings of the 7th International ISCRAM Conference</em>.</li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/how-series-failed-whee-teck-ong-oi7hc"><em>LinkedIn</em></a><em> on 26 October 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=65dab97bba84" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Almost Complete]]></title>
            <link>https://medium.com/@wheeteckong/almost-complete-91574300b8ed?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/91574300b8ed</guid>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[planning]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[leadership]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Thu, 22 Oct 2026 23:19:57 GMT</pubDate>
            <atom:updated>2026-10-22T23:19:57.995Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y3he3mz4N9sgMZQOkRrM5A.png" /></figure><p><em>The Judgment Gym · Rep 4 of 10: Take the Outside View</em></p><p>In the 1970s, the psychologist Daniel Kahneman sat in a room in Israel with a small team writing a new school curriculum.</p><p>They had been at it for about a year and were making good progress, so Kahneman asked everyone to write down, privately, how much longer they thought the project would take. The answers clustered around two years.</p><p>Then he turned to the most experienced person in the room, a curriculum expert named Seymour Fox, and asked a different question. Forget this team. Think of other teams you have known that set out to build a curriculum from scratch. From the point we have reached now, how long did they take?</p><p>Fox went quiet. Then he said something he had plainly never connected to his own estimate. A large share of those teams never finished at all. The ones that did took somewhere between seven and ten years.</p><p>So the room had two answers to the same question. Two years, from looking at the project. Seven to ten, with a real chance of failure, from looking at every project like it.</p><p>The team did what almost every team does. They carried on with the first number.</p><p>The curriculum took eight more years to finish. By then the people who had asked for it had lost interest, and it was never used.</p><h4>Two ways to look at the same plan</h4><p>Kahneman and his long-time collaborator Amos Tversky went on to name what happened in that room. The <em>planning fallacy</em> is our habit of underestimating how long our own projects will take, even when we know that projects like them usually run late.</p><p>Underneath it sits a distinction I find more useful than almost anything else in the study of judgment.</p><p>The <em>inside view</em> looks at your case: your plan, your team, your progress, the particular story of how this will unfold. It feels like diligence, because it engages with the details. And it is nearly always too optimistic, because the story you tell yourself is a best-case path with no room in it for the hundred things that will actually happen.</p><p>The <em>outside view</em> ignores your details almost completely. It asks one question. When people have done things like this before, how did it turn out?</p><p>The evidence for the outside view is humbling. In the early 1990s the psychologist Roger Buehler and his colleagues asked university students to predict when they would finish their final-year theses. The students said, on average, about 34 days. They took about 55. Even when they were asked for a date they were all but certain to beat, most of them missed it. Experience does not cure this. The planner Bent Flyvbjerg has spent decades showing the same pattern in large infrastructure projects run by seasoned engineers: each team takes the inside view of its own special project and ignores the record of all the projects like it.</p><p><em>Your situation feels unique. It is unique in ways that do not matter, and ordinary in the way that does.</em></p><h4>My twin looked at my bookshelf</h4><p>I thought I understood all of this. I have written a chapter on it. Then, while I was preparing this piece, my twin, the vault of my writing and the AI that reads it, took the outside view of something of mine.</p><p>My books.</p><p>In my OneDrive there is a folder called “Completed”. It holds 16 finished books, going back to 2016. Next to it are six folders of drafts, holding about 30 more: art books, business books, a storybook, a book of walks, a photography book. Some are duplicates. The oldest is dated 2014.</p><p>And there is a fourth folder, called “Almost Complete”. Four books sit in it. One of them is the book this series draws on, <em>The Human Premium</em>. Its files were last changed in June.</p><p>That folder name is the inside view, in two words. Every book in it feels nearly done, because I know exactly what is left. The outside view asks a ruder question. Of all the books I have ever started, what share got finished, and how long did the “almost” part take?</p><p>When my twin put these numbers in front of me, my first reaction was to push back. “Almost complete” was never about time. A book sits in that folder until I feel that I have <em>it</em>. Then it gets done quite fast. I think that is what inspiration is.</p><p>I still believe that. But the outside view does not argue with it. It changes the question.</p><p>If a book gets finished when the feeling arrives, then the base rate that matters is not how long the “almost” takes. It is how often the feeling arrives at all. On my own shelf, that is 16 finished books out of roughly 50 I have started. About one in three.</p><p>For creative work, that is not a bad number. It is simply a different number from the one in my head, which told me that every book in that folder was nearly there.</p><h4>What the machine changes</h4><p>There is something almost poetic about AI and the outside view. A large language model has, in a sense, read the reference class. It has taken in records of more projects, plans, estimates and outcomes than any person could hold in mind. If you ask it what usually happens to projects like yours, it can often tell you.</p><p>But it will just as happily do the opposite. Bring it your plan and ask it to help, and it will build on your inside view with enthusiasm, filling in milestones, smoothing the risks and making your best-case story more convincing. In <a href="https://www.linkedin.com/pulse/hungry-judges-whee-teck-ong-dp7yc">the previous article</a> I wrote that fluency is not evidence. A beautifully detailed plan is the most fluent document there is.</p><p>So the order matters. Ask for the base rate <em>before</em> you show it your plan.</p><h4>The rep: two numbers on one page</h4><p>Pick one estimate you are about to commit to: a deadline, a budget, a launch date, a hiring timeline.</p><ol><li><strong>Name the reference class.</strong> What is this a case of? Not “our transformation programme” but “large system replacements in banks”, or “first books by working professionals”, or “new teams hitting their first target”.</li><li><strong>Get the base rate first.</strong> Ask the people who have seen many of these, and ask the machine, how such cases usually turn out. Do this before you describe your own case, and ask for sources.</li><li><strong>Write both numbers down.</strong> Your inside estimate and the base rate, side by side, on the same page.</li><li><strong>Adjust modestly.</strong> Move from the base rate towards your own estimate only as far as your genuine differences justify. If the only argument is “this time is different”, do not move at all.</li></ol><p>Then commit to the adjusted number, not the hopeful one.</p><h4>Back to the room</h4><p>Fox knew the answer the whole time. He had watched those other teams. The base rate was in his own head, and it still did not reach his estimate until somebody asked the question in a different way.</p><p>That is what struck me about my bookshelf. The record was sitting in my own folders, in plain sight, with dates on it. It took a machine with no stake in my story to put the two numbers side by side.</p><p>Next in the series: what happens when you imagine your project has already failed.</p><p>In the meantime, tell me: what is in your own “almost complete” folder, and what does the outside view say about it?</p><h4>References</h4><ul><li>Buehler, R., Griffin, D. and Ross, M. (1994). Exploring the “planning fallacy”: Why people underestimate their task completion times. <em>Journal of Personality and Social Psychology</em>, 67(3), 366–381.</li><li>Flyvbjerg, B. (2008). Curbing optimism bias and strategic misrepresentation in planning: Reference class forecasting in practice. <em>European Planning Studies</em>, 16(1), 3–21.</li><li>Kahneman, D. (2011). <em>Thinking, Fast and Slow</em>. Farrar, Straus and Giroux.</li><li>Kahneman, D. and Tversky, A. (1979). Intuitive prediction: Biases and corrective procedures. <em>TIMS Studies in Management Science</em>, 12, 313–327.</li><li>Ong, W. T. (2026). <em>The Hungry Judges.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/hungry-judges-whee-teck-ong-dp7yc">https://www.linkedin.com/pulse/hungry-judges-whee-teck-ong-dp7yc</a></li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/almost-complete-whee-teck-ong-fhzcc"><em>LinkedIn</em></a><em> on 22 October 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=91574300b8ed" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Hungry Judges]]></title>
            <link>https://medium.com/@wheeteckong/the-hungry-judges-1d2f87b5f017?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/1d2f87b5f017</guid>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[critical-thinking]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[leadership]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Tue, 20 Oct 2026 06:00:31 GMT</pubDate>
            <atom:updated>2026-10-20T06:00:31.773Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nQbA5IPCYuSWQbYzKlILSw.png" /></figure><p><em>The Judgment Gym · Rep 3 of 10: Batch the Small Stuff</em></p><p>In 2011, three researchers published a study of eight judges in Israel.</p><p>The judges sat on parole boards. All day, prisoners came before them asking to be released, and the judges said yes or no. The researchers collected 1,112 of those rulings and plotted them against the time of day.</p><p>The pattern was startling. At the start of a session, the judges granted about 65 percent of requests. As the session wore on, the rate fell, steadily, until just before a break it was close to zero. Then the judges stopped for a meal, came back, and the rate jumped straight back to about 65 percent. And it began to fall again.</p><p>The lesson seemed obvious. Deciding is tiring. Judgment drains like a fuel tank, and a tired judge falls back on the safe default, which for a parole board is “no”. The idea got a name, <em>decision fatigue</em>, and it travelled everywhere. It explained why you make poor choices late in the day. It explained why some famous executives wear the same clothes every morning: to save their fuel for the decisions that matter.</p><p>It is a wonderful story. I have taught a version of it myself. And the evidence underneath it is far weaker than the story.</p><h4>What happened when someone checked</h4><p>The fuel-tank idea rested on a body of research called ego depletion: use up self-control on one task and you have less for the next. For years it looked solid, with dozens of published studies behind it.</p><p>Then, in 2016, the psychologist Martin Hagger coordinated a test the original research had never faced. Twenty-three laboratories around the world ran the same experiment, with the method agreed and registered in advance, on 2,141 people. The combined effect was close to zero.</p><p>The detail I find most instructive is this. Before the results came in, every lab except one predicted that the effect would show up. These were not sceptics setting out to debunk it. They were people who expected it to work.</p><p>The hungry judges have a problem too. The cases did not arrive in random order. When two other researchers, Weinshall-Margel and Shapard, looked again at how the sessions ran, they found that prisoners without a lawyer tended to be heard last in each session, and those prisoners were less likely to be released whatever time of day it was. Much of that dramatic curve may simply reflect the order of the queue.</p><p>None of this proves that judges never tire. It does mean that one of the most repeated stories in business writing was never the clean demonstration it was sold as.</p><h4>My twin found it in my own work</h4><p>When I started on this piece, my twin, the vault of my writing and the AI that reads it, went looking for every place I had used the idea.</p><p>It found nothing in my published articles. It found it in my own workshop material. Two of the units I teach present “decision fatigue” as plain fact, one as a reason to simplify your tools and one as a reason to clear your to-do list. I had absorbed the story so completely that it had become furniture.</p><p>I am rewriting both units this month.</p><p>This is the part of the series where the machine stops being only a tool and becomes a warning. The decision-fatigue story has been repeated in so many articles, books and talks that an AI trained on them will tell it to you fluently, confidently and in full, hungry judges included. Ask it and you will get the myth with a straight face. <em>Fluency is not evidence.</em> A story that has been told a thousand times sounds true to people and to machines, for the same reason.</p><h4>What is left when the myth goes</h4><p>So should you stop worrying about how many decisions you make? No. Something useful survives, and it is simpler than a fuel tank.</p><p>Every decision has overhead. Even a trivial one costs a moment of time and a sliver of attention, and it leaves a small trace behind that pulls at the next thing you do. Make forty small decisions before lunch and you have not drained a magic battery. You have spent forty slivers of something that really is limited, which is attention, on things that did not deserve it.</p><p>That is the reason to batch the small stuff. Not to protect your willpower. To protect your attention for the decisions where your judgment is the point.</p><p>I do this in three small ways.</p><p>My work shirts are white, some with a small design in the seam. I never have to think about what to wear in the morning, and I like to think I still look good.</p><p>My sketching trips all share one online module on the kit list. It describes the basic kit every participant needs, and every participant gets it free. I made it once, and I have not had to think about kit for a single trip since.</p><p>And the footer at the end of this article was written once, handed to the AI with instructions, and that was that.</p><p>None of these saves willpower. Each one keeps a morning, a trip or a publishing day free for the decisions that actually need me.</p><h4>The rep: a decide-once list</h4><p>Before the next rep, do three things.</p><ol><li><strong>Write a decide-once list.</strong> Look back over the last few days and find five small decisions you made more than once. Lunch, the order of your morning, which meetings to accept, how you reply to a routine request. Decide each one now, write the answer down, and stop deciding it.</li><li><strong>Batch the rest.</strong> Group the small decisions that remain and deal with them together at one fixed time, instead of letting them interrupt you all day.</li><li><strong>Hand the trivia to the machine.</strong> Defaults, first drafts of routine replies, small routings. In the language of <a href="https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec">the first article in this series</a>, these belong in Offload.</li></ol><p>And one more, because of what happened to me. Pick one “fact” you repeat at work, the kind that turns up in slides and speeches. Find the original study behind it and read what it actually found. Do not ask the machine to summarise it for you. Look it up yourself.</p><h4>Back to the queue</h4><p>So here is the hungry judges study again, read a second time.</p><p>The most useful thing in it may not be about hunger at all. If the critics are right, what shaped those rulings was the order in which the cases were heard: who came first, who came last, and who arrived without a lawyer. A structural detail nobody was watching did what everyone assumed tiredness was doing.</p><p>That is the real lesson for anyone who makes decisions all day. Design the queue. Decide the small things once, batch the rest, and give the machine the trivia. Then bring your full attention to the cases that deserve it, and keep a little scepticism for the stories you like best.</p><p>Next in the series: why your own case is rarely as special as it feels, and what to look at instead.</p><p>In the meantime, tell me: what is a “fact” you have repeated for years that you have never checked?</p><h4>References</h4><ul><li>Danziger, S., Levav, J. and Avnaim-Pesso, L. (2011). Extraneous factors in judicial decisions. <em>Proceedings of the National Academy of Sciences</em>, 108(17), 6889–6892.</li><li>Hagger, M. S., Chatzisarantis, N. L. D., Alberts, H. et al. (2016). A multilab preregistered replication of the ego-depletion effect. <em>Perspectives on Psychological Science</em>, 11(4), 546–573.</li><li>Ong, W. T. (2026). <em>The Free Soda.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/free-soda-whee-teck-ong-oizwc">https://www.linkedin.com/pulse/free-soda-whee-teck-ong-oizwc</a></li><li>Ong, W. T. (2026). <em>Twin of What?</em> LinkedIn. <a href="https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec">https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec</a></li><li>Weinshall-Margel, K. and Shapard, J. (2011). Overlooked factors in the analysis of parole decisions. <em>Proceedings of the National Academy of Sciences</em>, 108(42), E833.</li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/hungry-judges-whee-teck-ong-dp7yc"><em>LinkedIn</em></a><em> on 20 October 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1d2f87b5f017" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Free Soda]]></title>
            <link>https://medium.com/@wheeteckong/the-free-soda-8d3ae3c28b1a?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/8d3ae3c28b1a</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[judgment]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[decision-making]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Tue, 13 Oct 2026 23:48:12 GMT</pubDate>
            <atom:updated>2026-10-13T23:48:12.817Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Dr7fyI9nNn3F3tZCsQlDQw.png" /></figure><p><em>The Judgment Gym · Rep 2 of 10: Widen the Options</em></p><p>This series nearly had a different name.</p><p>Earlier this month I sat down with my twin, the vault of ten years of my writing and the AI that reads it, to plan what I should write next. It did what it is good at. It scanned 189 articles, found the thread I had missed, and made a recommendation: name the series after the book I am writing.</p><p>Then it asked me a question. Should the series carry the book’s name?</p><p>It was a reasonable question. It was clearly put, and it came with a sensible argument attached. I could have answered it in one word and moved on, and I nearly did.</p><p>I said no. Then I asked for something it had not offered. <em>Give me other options.</em></p><p>In a few seconds I had three. It recommended the first. I chose the second, which is why you are reading The Judgment Gym.</p><p>I have thought about that exchange a good deal since, because of how close I came to answering the question I was given. Nothing was wrong with the analysis. The problem was the shape of the question. It was a yes or a no, about one option. And a great deal of what goes wrong in organisations starts exactly there.</p><h4>168 decisions</h4><p>In the early 1990s a management researcher at Ohio State University, Paul Nutt, did something laborious. He reconstructed how real organisations had made real decisions. He did not survey opinions or run a laboratory game. He took 168 actual decisions, about products, money, people and strategy, and traced how each one had been reached.</p><p>When he counted, he found that only 29 percent of them had considered more than one alternative.</p><p>In all the rest, the “decision” was a verdict on a single option. Should we buy this company? Should we launch this product? Should we hire this person? Yes or no.</p><p>That number is hard to judge on its own, so here is a comparison. When the writers Chip and Dan Heath described Nutt’s work in their book <em>Decisive</em>, they set it beside a study of how teenagers make decisions, by the Carnegie Mellon researcher Baruch Fischhoff. Among the teenagers, about 30 percent considered more than one option.</p><p>Put the two side by side. The organisation, with its committees and its analysts and its budget cycle, was weighing alternatives about as often as a teenager deciding whether to go to a party.</p><p>The Heaths gave the habit a name: <em>narrow framing</em>. I think of it more simply as the yes-or-no trap.</p><h4>Why the trap is comfortable</h4><p>The trap works because it feels like diligence.</p><p>The single option arrives fully formed. Often it is urgent. Usually somebody senior is behind it. So the team does the responsible thing: it researches the option, debates it, lists the pros and the cons, and builds the business case. Every hour of that work is real. All of it is aimed at a question that was too small before anyone started.</p><p><em>You can analyse one option perfectly and still decide badly, because the best choice was one nobody put on the table.</em> Nutt found the same pattern in the outcomes. As the Heaths report his results, decisions framed as “whether or not” failed 52 percent of the time over the long term. Decisions that weighed two or more alternatives failed 32 percent of the time.</p><h4>Four o’clock at the branch</h4><p>Many years ago a client told me a story I have never forgotten.</p><p>A bank’s Trade Finance application slowed to a crawl at every branch at four o’clock each afternoon. That was the hour when messengers arrived from the bank’s clients, rushing their paperwork in before the branches closed. The system became so slow it was hard to work with at all.</p><p>The person asked to investigate was handed two options. Buy new hardware, quoted at about $2 million. Or rewrite the software, quoted at about $1 million.</p><p>He chose neither. After talking to the people involved, he asked every branch to offer a free soda to any messenger who brought the papers in before three o’clock.</p><p>The congestion went away, “mysteriously”, for a great deal less than a million dollars.</p><p>Notice that he had been given two options, not one. It was still the same trap. Both options answered one narrow question, which was how to make the computer faster. Nobody had asked whether the problem was the computer at all.</p><h4>What the machine changes, and what it does not</h4><p>For most of my career there was an honest excuse for narrow framing. Producing three good options where one had been handed to you was work. It took time and imagination, and usually a meeting nobody wanted.</p><p>That excuse has gone. Ask a capable AI for alternatives and it will give you a dozen in seconds, some of them framed in ways you would not have thought of. Option generation used to be expensive. It is now nearly free.</p><p>Here is the surprise. That does not fix the problem, and it can make it worse.</p><p>The machine will widen the frame if you ask it to. It will just as happily do the opposite. Bring it your single option and ask it to evaluate it, and it will analyse that one narrow thing thoroughly, fluently and with great confidence. Your too-small question now comes back looking carefully examined. <em>The machine amplifies whatever frame you bring it.</em></p><p>Look again at what happened with my series. The machine produced a well-argued yes-or-no question. It also produced three good alternatives, the moment it was asked. What it did not do was notice, on its own, that the first question was too narrow. That was my part. So was choosing the option it had not recommended.</p><p>Last week I wrote about <a href="https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec">sorting your thinking into Offload, Share and Keep</a>. This is where that sort earns its place. Generating options goes in Offload. Refusing the frame, and making the choice, stay in Keep.</p><h4>The rep: three questions before you evaluate anything</h4><p>Pick one decision on your desk this week that is phrased as “should we” or “whether to”. Before anyone analyses it, ask three questions.</p><ol><li><strong>The vanishing options test.</strong> Suppose the option in front of you were not allowed. What would you do instead? This comes from the Heaths, and it works at once, because it shows that there were always other paths. You stopped looking when the first acceptable one appeared.</li><li><strong>Half and double.</strong> What would you do with half the budget or half the time? What would you do with twice as much? The answers are rarely smaller and larger versions of the same plan.</li><li><strong>Four more.</strong> Ask the machine for four alternatives you have not considered, and do it <em>before</em> you ask it to assess anything.</li></ol><p>Then apply one rule. No evaluation starts until there are at least three real options on the page. Real means that a sensible person could choose it. Two versions of the same plan do not count, and new hardware against new software was exactly that.</p><p>Decide which. Do not decide whether.</p><h4>The name on the door</h4><p>So, back to the name.</p><p>My twin had read everything I have published, and it used all of it to make a good case for one answer. That is what a record of the past is good at. It can tell you what follows from what you have already said.</p><p>It cannot be dissatisfied with the question. That morning, the most useful thing I did took four words and no analysis at all.</p><p>Next week: why the decisions that wear down your judgment are rarely the big ones.</p><p>In the meantime, tell me: what is the “whether or not” decision on your desk right now, and what did the vanishing options test turn up?</p><h4>References</h4><ul><li>Fischhoff, B. (1996). The real world: What good is it? <em>Organizational Behavior and Human Decision Processes</em>, 65(3), 232–248.</li><li>Heath, C. and Heath, D. (2013). <em>Decisive: How to Make Better Choices in Life and Work</em>. Crown Business.</li><li>Nutt, P. C. (1993). The identification of solution ideas during organizational decision making. <em>Management Science</em>, 39(9), 1071–1085.</li><li>Nutt, P. C. (2002). <em>Why Decisions Fail: Avoiding the Blunders and Traps That Lead to Debacles</em>. Berrett-Koehler.</li><li>Ong, W. T. (2026). <em>Twin of What?</em> LinkedIn. <a href="https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec">https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec</a></li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/free-soda-whee-teck-ong-oizwc"><em>LinkedIn</em></a><em> on 14 October 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8d3ae3c28b1a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Twin of What?]]></title>
            <link>https://medium.com/@wheeteckong/twin-of-what-ff51c6409daf?source=rss-22a11f3431cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/ff51c6409daf</guid>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[judgment]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Whee Teck Ong]]></dc:creator>
            <pubDate>Sat, 10 Oct 2026 03:22:56 GMT</pubDate>
            <atom:updated>2026-10-10T03:22:56.774Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*goJ6u4wg58izgkPfv6KNDQ.png" /></figure><p><em>The Judgment Gym · Rep 1 of 10: Keep the Judgment</em></p><p>I have spent the last two months building a twin of myself.</p><p>It is less dramatic than it sounds. I took ten years of my writing, 189 articles going back to 2016, and filed every one of them in a single vault that an AI can read, search and cross-reference. Ask it what I think about late bloomers, or boards, or why technology projects fail, and it answers in seconds, with the date and the source. It finds connections between pieces I had forgotten I wrote.</p><p>It also showed me something I did not expect. When I looked at the map of connections, the links between my articles and the material I teach were surprisingly sparse. Ten years of writing and a whole body of teaching, sitting side by side in the same vault, with only a few threads between them. I had never seen that. The machine did.</p><p>The question I keep coming back to is a simple one. <em>Twin of what?</em></p><p>When I finally answered it honestly, the answer was uncomfortable. <em>It is a twin of my past. Not my future.</em></p><p>It holds what I have already said: my recall, my vocabulary, my frameworks, my turns of phrase. What it does not hold is the thing all of those were for.</p><p>It has never been wrong in a room where being wrong was expensive. It has never recommended something to a client and then had to sit across from them a year later. It can tell you what I concluded. It cannot tell you why I hesitated, what I nearly said instead, or what I would do now that the situation has changed. It has my outputs. It does not have my judgment.</p><p>That distinction is going to matter to every professional reading this, whether or not you ever build a twin.</p><h4>The thinking does not disappear. It moves.</h4><p>When cheap calculators arrived in classrooms in the 1970s, teachers worried that children would never learn arithmetic again. They were partly right. Few of us could do long division by hand today without a struggle. That skill did wither.</p><p>But the worry missed the more important thing. The people who understood what they were doing were freed to reach for bigger problems. The thinking moved up a level: from doing the sums, to deciding which sums were worth doing and whether the answer made sense.</p><p>AI is now doing the same thing to thinking itself. When you hand a task to a capable machine, the thinking does not vanish. It relocates. From writing the first draft to judging whether the draft is any good. From finding the answer to checking it, fitting it with everything else you know, and standing behind it.</p><p>In July I argued that <a href="https://www.linkedin.com/pulse/machine-sees-pattern-you-decide-what-means-whee-teck-ong-amo2c">the machine sees the pattern and you decide what it means</a>. This is the same point from the other side. The work that remains yours is the work of deciding.</p><h4>Two findings that point the same way</h4><p>The first is about demand. The World Economic Forum’s <em>Future of Jobs Report 2025</em> puts AI and big data at the top of its list of fastest-growing skills. Rising alongside them, on the same list, are creative thinking and resilience, flexibility and agility. Employers want the machine skills and the human ones together.</p><p>The second is about erosion. A 2025 study by researchers at Carnegie Mellon and Microsoft surveyed 319 knowledge workers about how they use generative AI. The more confidence people had in the tool, the less critical thinking they reported doing.</p><p>Put the two together. The premium on a sharp mind has never been higher, and the temptation to let it go slack has never been greater.</p><p>Nobody decides to stop thinking. It happens one convenient task at a time. You let the machine summarise the report because you are busy. Then you let it suggest the recommendation. Then one day you are presenting a view you cannot defend, because it was never really yours.</p><p>This is not an argument against using AI. The clerk who refused the calculator did not win either. Use the machine boldly. <em>Let it carry the load without letting it carry the judgment.</em></p><h4>One question, three buckets</h4><p>So here is the first rep. It takes ten minutes.</p><p>Before you hand any piece of thinking to a machine, ask one question: <em>is this a muscle I want to keep?</em></p><p>Then sort what you do into three buckets.</p><ul><li><strong>Offload.</strong> Drudgery, formatting, boilerplate, computation nobody was ever wiser for doing by hand. Let it go and do not look back.</li><li><strong>Share.</strong> Do it yourself first, then let the machine check you. It checks you; it does not replace you. The order matters. If the machine goes first, you are no longer thinking. You are editing.</li><li><strong>Keep.</strong> The judgment your work turns on. The call you would have to defend. The reason anyone asks for your opinion at all.</li></ul><p>Try this before next week:</p><ol><li>List ten decisions you made or helped make in the last seven days. A hire, a price, a reply to a difficult email, a number in a forecast.</li><li>Tick every one where an AI tool was involved in any way.</li><li>Mark each one O, S or K.</li><li>Circle any K or S where the machine did the thinking before you did.</li></ol><p>That circle is a leak.</p><p>My own best example runs the other way. Last month <a href="https://www.linkedin.com/pulse/eleven-songs-country-doesnt-say-things-directly-whee-teck-ong-xpycc">I wrote about</a> <a href="https://open.spotify.com/album/4ZnyKtZQGewzCjNqqEp80r"><em>Kia Lai Kia Khee</em></a>, an album of eleven songs in Singlish that I made with AI. Before I asked the machine to draft a single line, I spent a long time thinking about the Singlish itself. Only then did I hand over the writing. That was where I led. I think many situations now call for human thinking <em>first</em>.</p><h4>Why a gym</h4><p>Last month I wrote that companies should stop bolting AI onto old structures and <a href="https://www.linkedin.com/pulse/dont-bolt-ai-build-judgment-gym-whee-teck-ong-6p5hc">build a judgment gym</a> instead. That article described the building. It did not hand anyone the equipment.</p><p>So this is the first of ten articles, one a week, each with a single rep you can do on a real decision. They draw on the research behind a book I am currently writing, <em>The Human Premium</em>, and on the workshops I run on judgment. None of the ten requires talent. Judgment is not a gift, and it is not just time served. It is reps under consequence, and reps can be designed.</p><p>Which brings me back to the twin.</p><p>It is still there, in the vault, and every article I publish from now on will be filed into it. It will keep getting better at being who I was. I am glad to have it. In the sort above, it belongs in the first two buckets: a superb memory and a tireless research assistant.</p><p>But it cannot do a single one of these reps. It has no decision of its own to make, and nothing to lose by being wrong. <em>The twin keeps my past. The gym is for my future.</em> That part of the work has no copy, and I would not want one.</p><p>Next week: why most decisions are framed as “whether or not”, and why that is nearly always the wrong question.</p><p>In the meantime, do the sort, and tell me: what did you find in your Keep bucket that you had been quietly handing away?</p><h4>References</h4><ul><li>Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R. and Wilson, N. (2025). <em>The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers</em>. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. <a href="https://doi.org/10.1145/3706598.3713778">https://doi.org/10.1145/3706598.3713778</a></li><li>Ong, W. T. (2026). <em>Don’t Bolt On AI. Build a Judgment Gym.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/dont-bolt-ai-build-judgment-gym-whee-teck-ong-6p5hc">https://www.linkedin.com/pulse/dont-bolt-ai-build-judgment-gym-whee-teck-ong-6p5hc</a></li><li>Ong, W. T. (2026). <em>Eleven Songs About a Country That Doesn’t Say Things Directly.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/eleven-songs-country-doesnt-say-things-directly-whee-teck-ong-xpycc">https://www.linkedin.com/pulse/eleven-songs-country-doesnt-say-things-directly-whee-teck-ong-xpycc</a></li><li>Ong, W. T. (2026). <em>The Machine Sees the Pattern. You Decide What It Means.</em> LinkedIn. <a href="https://www.linkedin.com/pulse/machine-sees-pattern-you-decide-what-means-whee-teck-ong-amo2c">https://www.linkedin.com/pulse/machine-sees-pattern-you-decide-what-means-whee-teck-ong-amo2c</a></li><li>World Economic Forum (2025). <em>The Future of Jobs Report 2025</em>. <a href="https://www.weforum.org/publications/the-future-of-jobs-report-2025/">https://www.weforum.org/publications/the-future-of-jobs-report-2025/</a></li></ul><p><em>Ong Whee Teck writes about the human skills that matter more as AI takes on routine work: judgment, learning, storytelling and leadership. He is the author of “The Job Apocalypse” and “Story Essence DNA”, and is now writing “The Human Premium”, on 50 research-backed skills worth more in the age of AI. He is Consulting Partner, ASEAN and Korea at Kyndryl and a Fellow of the Singapore Computer Society, with 30 years leading technology transformations across Asia Pacific, including as Managing Director at Accenture and CEO of Temasek Management Services.</em></p><p><em>Views expressed are my own and do not represent those of my employer.</em></p><p><em>Originally published on </em><a href="https://www.linkedin.com/pulse/twin-what-whee-teck-ong-ztiec"><em>LinkedIn</em></a><em> on 10 October 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ff51c6409daf" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>