Learning Library

← Back to Library

Cultivating Judgment in the AI Age

Key Points

  • The rise of cheap, abundant AI means everyone—from consultants to internal teams—must become “judgment merchants,” cultivating the hard‑to‑teach skill of good business judgment across all roles.
  • Because intelligence costs are dropping dramatically, value now comes from identifying what remains scarce (e.g., selection, sequencing, implementation, human resources, attention) and targeting those bottlenecks.
  • Effective judgment is context‑sensitive synthesis, requiring you to tailor decisions to the specific circumstances and constraints of each AI project.
  • The speaker will outline ten concrete principles (starting with the scarcity principle and context) to help professionals systematically develop and demonstrate strong judgment in the AI era.

Full Transcript

# Cultivating Judgment in the AI Age **Source:** [https://www.youtube.com/watch?v=O_VL5clgN_I](https://www.youtube.com/watch?v=O_VL5clgN_I) **Duration:** 00:14:46 ## Summary - The rise of cheap, abundant AI means everyone—from consultants to internal teams—must become “judgment merchants,” cultivating the hard‑to‑teach skill of good business judgment across all roles. - Because intelligence costs are dropping dramatically, value now comes from identifying what remains scarce (e.g., selection, sequencing, implementation, human resources, attention) and targeting those bottlenecks. - Effective judgment is context‑sensitive synthesis, requiring you to tailor decisions to the specific circumstances and constraints of each AI project. - The speaker will outline ten concrete principles (starting with the scarcity principle and context) to help professionals systematically develop and demonstrate strong judgment in the AI era. ## Sections - [00:00:00](https://www.youtube.com/watch?v=O_VL5clgN_I&t=0s) **Cultivating Judgment in the AI Era** - The speaker warns that as AI drives down the cost of intelligence, every professional must deliberately develop the hard‑to‑teach skill of good business judgment—becoming “judgment merchants”—and outlines ten principles for doing so. - [00:05:10](https://www.youtube.com/watch?v=O_VL5clgN_I&t=310s) **Judgment, Sequencing, and Deprioritization** - The speaker stresses that good judgment means ordering AI bets as thin‑sliced, trust‑building MVPs and explicitly defining non‑goals to prevent scope creep. - [00:08:35](https://www.youtube.com/watch?v=O_VL5clgN_I&t=515s) **Coalition Principle for AI Project Success** - The speaker explains how rapid feedback in AI initiatives combined with strategically mapping and sequencing stakeholder buy‑in—creating early wins to shift support from passive permission to active ownership—is essential for effective judgment. - [00:12:12](https://www.youtube.com/watch?v=O_VL5clgN_I&t=732s) **Cultivating Human Judgment in AI Era** - The speaker argues that clear, logical thinking and transparent trade‑off analysis are essential soft skills that distinguish humans from token‑spitting AI, and that encoding such judgment into scalable systems transforms personal insight into lasting organizational capability. ## Full Transcript
0:00We're all becoming judgment merchants. I 0:02know that's a new word. I'm coining it. 0:04The reason I'm saying that is that 0:06whether you work as an external 0:07consultant in AI around AI or whether 0:10you're working inside a company and you 0:12are building in AI systems, wanting to 0:15build an AI systems, leaning in on Chad 0:17GPT, everyone is going to have to start 0:20practicing judgment. This is not a 0:23taught skill. This is not something that 0:25you can just go and say, "Oh, I know 0:27judgment now. I'm good at judgment." In 0:29fact, for a long time, the ability to 0:32exercise good business judgment was the 0:34bar for principal product manager, the 0:37bar for a senior engineering leader. 0:40Now, what I'm suggesting to you is that 0:42we need to find ways to cultivate that 0:44for every level, for every job family. 0:47Why is that? Because intelligence is 0:50becoming too cheap to meter. Sam Alman 0:53was saying that intelligence is falling 0:55x a year in cost. X a year for the same 0:57intelligence level. If that is even 0:59remotely close to true, we have to 1:02double down on being good judgment 1:04merchants. And if that sounds too 1:05abstract, it's not going to be by the 1:07time you're done with this video. I'm 1:09going to go through 10 principles. As 1:11far as I know, no one has really put 1:12together a miniourse on how to have good 1:16judgment in the age of AI, even though 1:19that's an irreplaceable skill and we 1:21talk about it a lot. So, let's get into 1:23it. What are 10 ways we can show we 1:26exercise good judgment? This applies 1:27regardless of job family. And this 1:30should be specific enough for you to 1:31actually jump in and wrap your head 1:33around. That's my goal. Number one, the 1:35scarcity principle. So intelligence is 1:38abundant, right? If it's coming down at 1:40a cost of 40x a year, everybody's going 1:42to have intelligence. Value then 1:44migrates to the next bottleneck. 1:46Basically, one of the ways you show 1:49value, whether you are inside a company 1:52like Walmart or Netflix or Amazon or a 1:56tiny company like a series A or a Seed 1:59or whether you're outside as a 2:00consultant, everywhere you look on an AI 2:03project, you will see places where 2:06intelligence unlocks an enormous amount 2:08of volume and you will see places where 2:10that volume bottlenecks. Part of your 2:13value is finding the bottlenecks. It's 2:16the scarcity principle. Find what is 2:19still scarce in a world where 2:21intelligence is abundant. Get eyeballs, 2:23get get binoculars, get a microscope, 2:25whatever metaphor you want, but find the 2:28scarcity. The scarcity is there. Maybe 2:30it's scarcity of selection and sort of 2:32finding what to choose. Maybe it's 2:33scarcity of sequencing. It's really hard 2:35to know how to sequence. Maybe it's 2:37scarcity of implementation. Maybe it's 2:38scarcity of human resources in other 2:40areas. There is always going to be 2:42something scarce. Maybe customer 2:44attention is scarce, right? Find the 2:46scarcity. The fastest signal of good 2:49judgment is how precisely you can define 2:52the true current bottleneck. And that is 2:54true whether you're internal or 2:55external. Principle number two, context. 2:58Judgment is contextsensitive 3:01synthesis. In other words, you are 3:02reusing patterns with an awareness of 3:05what makes this situation unique. Good 3:08judgment is excellent pattern 3:10recognition crossed with excellent 3:12context discrimination. You know the 3:14current context and you have enough of a 3:16pattern recognition that you can 3:18actually put them together. Poor 3:20judgment is overgeneralizing on a past 3:23success or failure or just relying on an 3:25AI generated best practice. The 3:27implication is that if you are hungry to 3:30show you have good judgment, you can 3:32surface the nontransferable elements of 3:35your recommendation and put those in 3:37center stage. So when you're talking 3:39internally about a project you want 3:41done, when you are externally and you're 3:43pitching something as a consultant, what 3:45is unique about this moment, this org, 3:48this system? Let's say you're a product 3:50manager and you're trying to get 3:52something built and it's an AI native 3:54architecture. What is it about this org 3:56that makes that solution specifically 3:59correct? If you are looking to show you 4:01have good judgment, look to show you 4:04understand context deeply. That is your 4:07advantage. Get passionate about a 4:09particular corner of context. Principle 4:11number three is the constraint 4:12principle. You are analyzing what's 4:15possible and you are judging what's 4:17possible. Now, analysis by itself is 4:21paralyzing. analysis will tell you all 4:23of your options. But a great business 4:27builder will think in terms of 4:29constraints, will think in terms of what 4:32is the possible build that we can 4:34execute today. And so an excellent 4:37judge, someone who shows good judgment 4:40in the middle of the AI age, understands 4:43intuitively within a given context what 4:46is possible. Now, do you see how these 4:48principles build on each other? I've 4:49sequenced them carefully. These are not 4:51randomly allocated principles. They 4:53build on each other so that you actually 4:55develop a cleaner and cleaner 4:57understanding of judgment as we go 4:58forward. Number four is the sequencing 5:01principle. Most insights fail because 5:04timing is bad or sequencing is bad. 5:06Judgment shows up in ordering your bets 5:10to create momentum and proof before 5:12resistance starts to mount. In other 5:14words, if you are internally trying to 5:16champion an AI system and there is some 5:19skepticism about how this system will 5:21work, order your bets carefully. Show 5:24good judgment by showing you know how to 5:26sequence bets and thin slice your value 5:29so that you can deliver something that 5:31people can believe in. Thin slicing 5:34value sounds abstract until you 5:35basically have to look at an MVP of a 5:38buildable system and say this little 5:40piece that's what I'm going to deliver 5:42because that earns trust. Maybe 5:44initially we're only going to deliver a 5:46chatbot on this particular page on the 5:48website. Maybe initially we're only 5:51going to convert this part of the wiki 5:53into a rag system internally. Maybe 5:55initially I'm only going to offer this 5:58particular piece of value in my 6:00consultancy as an AI consultant because 6:03I know I can deliver it and I know I can 6:04deliver it fast. That will earn trust 6:06and then I can deliver more. Sequencing 6:08matters and good judgment shows up in 6:11knowing what is possible. Now, principle 6:14five is dep prioritization. Everyone's 6:16favorite word. Know what not to do. Have 6:19non- goals. Explicitly list and defend 6:21ideas that you are not going to go after 6:23and not going to pursue. Have a 6:25rationale for that. That is something I 6:27will guarantee you that AI is not so 6:29great at. AI loves to expand scope. One 6:32of the signals of good judgment in 2025 6:35of someone who is accountable for their 6:37work is they will be honest with you at 6:40a moment's notice about what they are 6:42saying no to what they are 6:43deprioritizing what they cannot do. That 6:46is irreplaceable. And that is something 6:49honestly that is an excellent piece of 6:51career advice in an age littered with do 6:53more, do more, do more. What are you 6:55doing that is too much? What are you 6:56doing that you need to let go of so you 6:58can focus more effectively? And if you 7:00really commit to that, if you choose 7:02what you're going to depprioritize when 7:04you look at a project, you open the door 7:06to disproportionate leverage. And so 7:08when you see projects that are runaway 7:10successes, part of why they're runaway 7:12successes is someone in that project 7:15said, "This is the goal. Not this, not 7:18this, not this. We are building a 7:20chatbot for the customer. It will not 7:22have the ability to use images. It will 7:24not have voice initially. It will not 7:26give you the ability to upload files. it 7:28will be extremely good at talking about 7:30the products on our website. That is all 7:33focus have non- goals and that's true 7:35whether you're trying to define what to 7:37build or whether you're trying to define 7:39what you're doing next. It is it is a 7:41universal truism. It's absolutely 7:44essential. You you have to say no in 7:47order to be able to say yes. And it's 7:48something that I I keep beating the drum 7:50on it because AI is so bad at it. And I 7:53want to call out like the there are 7:54things AI is terrible at and this is one 7:56of them. like specialize in this, right? 7:58Dep prioritization. Number six, the 8:01calibration principle. Judgment 8:03compounds through feedback on accuracy. 8:06In other words, your judgment gets 8:09better as you get feedback on what goes 8:11right and what does not go right. You 8:14start to anticipate more correctly as 8:16you get feedback on what works and 8:19doesn't work. In a sense, this principle 8:21is fractal. It works for your career in 8:24the sense that you try things on and you 8:25say that worked, that didn't work. But 8:27that takes period of years, right? That 8:29takes time. It also works on projects. 8:33It works on AI projects especially well 8:35because AI projects are typically run 8:37fast and they're run hard and you get 8:39feedback quickly. And so if you set up 8:41an AI agent and you see nobody is using 8:43it internally, you get feedback on how 8:45you scope that agent. Maybe you 8:47prioritize incorrectly. Maybe you depp 8:49prioritize the one thing that the 8:52business actually wanted. You can learn 8:54from that. You can calibrate. You can 8:56get better. Principle number seven. This 8:58is another key element of good judgment. 9:00It's a coalition principle. Good 9:02judgment includes mapping the social 9:05graph of decision makers and sequencing 9:07their buyin correctly. You need to be 9:09planning conviction moments where 9:11stakeholders experience early wins and 9:13shift their stance from merely 9:15permitting something to happen to active 9:17ownership. And if that sounds too 9:19abstract, if you're trying to get 9:20something over the line internally as a 9:22project, it only happens if you get your 9:25director and then your director's peer 9:27and then your senior vice president and 9:29then finally the seauite on board. 9:31That's exactly what I just described. I 9:33just used abstract terms for it so 9:35everybody could understand it. It's the 9:37same thing. If you're a consultant, you 9:38run through the same process. Do you see 9:40that? One of my larger thesis is that we 9:44are overblowing the death of 9:45consultants. It is it is actually I 9:48think more correct to say that a lot of 9:52the busy work in producing decks that 9:54consultants have done for a long time is 9:56going to go away. But the idea of 9:59someone who offers good judgment is 10:01becoming more and more important because 10:03AI is bottlenecked on good judgment. And 10:05so excellent consultants that offer good 10:07judgment are priceless. But so are 10:11excellent internal teammates who offer 10:14good judgment. In a sense, we are all 10:15becoming consultants. So the coalition 10:18principle matters. Figure out who needs 10:20to go from permission to excitement to 10:23ownership to enthusiasm and sequence 10:26your conviction moments, your aha 10:27moments to get them there. That's good 10:29judgment. Again, AI can't help you 10:31there. Principle eight is the 10:33responsibility principle. Judgment 10:35carries ownership. The clearest tell of 10:38good judgment is a willingness to say, 10:40"If I'm wrong, here's how we'll know it, 10:42and here's how what we'll do." You don't 10:44have to be right all the time to have 10:47good judgment. People sometimes think 10:48you do, but you have to have 10:50accountability and you have to be 10:52willing to say if I made a mistake, 10:54here's what we're going to do about it. 10:55This is where I often tell people the 10:57fastest way to fix AI slop at work is to 11:00tell everyone, you are accountable for 11:03every word you write. It can be with AI, 11:05but you're still accountable. That 11:06accountability is a sign of good 11:08judgment. Own the consequences. Own the 11:11consequences. Again, not something AI is 11:14good at. Number nine, the transparency 11:16principle. In the old model, assess 11:19opacity, signal value. Trust the deck. 11:21That's a fancy way of saying just trust 11:23our deliverables. A shiny report. Now, 11:25people actually trust transparent 11:27reasoning more. This is a trend that's 11:29opening up because intelligence is so 11:31available. When intelligence was was 11:33expensive, people trusted that a wellp 11:36polished deck was a sign that you'd put 11:38a lot of good thought into it. Now, a 11:40wellp polished deck is just, you know, a 11:42chat with Kimmy K2 away, but a 11:44thoughtful deck is not. A thoughtful 11:47deck still requires good old-fashioned 11:49brain power. And so, being more 11:51transparent with your reasoning is 11:53coming into vogue. You want to be honest 11:55about the options. You want to be honest 11:56about your depp prioritization logic, 11:58about your assumptions, how you think it 12:00through, what trade-offs you're 12:01negotiating. That is something that 12:03wasn't true 5 years, 10 years ago in the 12:06same way that it is today. that is 12:08somewhat new and it's a response to AI. 12:10It is a way of showing you know how to 12:12think clearly, logically in an age where 12:15AI just wants to spit tokens because the 12:17difference between intelligence that 12:19just says everything is on the table. We 12:21won't have non- goals. We're going to be 12:23super aggressive. An intelligence that 12:24can lay out hard trade-offs really 12:26clearly is real. If you've seen good 12:29thinking, you know the difference 12:31between really good quality thinking and 12:33the first draft out of chat GPT. Don't 12:35just ship the first draft out of Ched 12:37GPT. Good judgment is showing 12:39transparently how you are actually 12:41wrestling with the problem. Principle 10 12:43is the compounding principle. Judgment 12:45creates leverage when it's encoded into 12:47systems that last. And so a lot of good 12:50judgment is figuring out how to solve a 12:53problem by scaling out a playbook or an 12:55automation that others can run with. So 12:57your judgment shifts then from personal 12:59heroics to true organizational 13:02capability. I want to give you some 13:04encouragement. The 10 principles I've 13:06outlined are all things that AI is not 13:09getting better at all that fast. So if 13:11you're sitting there wondering why is 13:13Nate talking about this soft skill 13:15stuff? Isn't AI going to be good at all 13:17this stuff? The answer is these are the 13:19soft skills that distinguish what humans 13:22are good at in the age of AI. I want you 13:25to know it because I think it's not 13:27something that is taught well because we 13:29didn't have to learn it except by 13:31osmosis before. Principal engineers 13:34learn this from other principal 13:35engineers for their discipline. Now 13:37someone has to start teaching it for all 13:40of us because suddenly judgment is all 13:43of our job. Good judgment is something 13:45that AI can't take away. So we'd better 13:47get good at it. So we'd better learn it. 13:49So we better learn how to teach it. And 13:50that's where this is coming from is I 13:52want everyone to start to get into the 13:54idea of having good judgment. And I want 13:56you to start to break down the mental 13:58wall between being inside the company 14:00and outside the company because I think 14:02a lot of the traditional roles that 14:04consultants had are now being occupied 14:06by people inside the company as well. 14:09Those lines are starting to blur. And 14:10similarly, you see consultants doing 14:12software work that in the old days would 14:14have been done by internal employees. 14:16It's another example of AI pushing the 14:18lines. And that's all happening because 14:20intelligence is becoming cheaper and 14:22cheaper and judgment is becoming more 14:24valuable. Judgment is the new 14:25bottleneck. Judgment is what is becoming 14:28scarce when analysis is free. I hope you 14:30enjoyed this. I hope you feel like you 14:32have a better sense of what actually 14:33goes into the idea of good judgment. And 14:35I hope you're able to apply it at work 14:36because that's the whole point. I wrote 14:38this up more on the blog. I have a 14:40prompt to help you think through where 14:42you need to learn and grow on judgment. 14:44Dig in and have fun.