Iconoclast

Kai-Fu Lee argues AI fails to boost profits due to poor measurement and organizational structure

38 min Listen to the original
Artificial IntelligenceCorporate StrategyLeadershipEnterprise SoftwareFuture of Work

Summary

Dr. Kai-Fu Lee argues that most companies fail to see bottom-line returns from AI because they measure activity rather than results. He claims only 6% of firms see financial impact, citing McKinsey research. Lee advocates for top-down transformation led by CEOs, rejecting bottom-up approaches. He introduces 'Boss AI' as a tool for executives to bypass hierarchical filters. Lee predicts human value will shift to leadership and empathy as AI handles execution.

Episode sections

  1. 0:00 – 2:04

    Introduction and AI Measurement Failures

    Dr. Kai-Fu Lee argues that measuring AI activity rather than results is ineffective. He notes that many CEOs are shocked when asked how AI impacts their bottom line, as they mistakenly believe that simply deploying agents is sufficient.

  2. 2:04 – 7:20

    Organizational Structure and CEO Leadership

    Lee advocates for top-down AI transformation led by the CEO, rejecting bottom-up approaches. He suggests flattening hierarchies and creating new roles for individuals accountable for managing AI agents, as traditional management structures are inadequate for AI-native companies.

  3. 7:20 – 20:06

    Boss AI and Executive Tools

    Lee introduces Boss AI, a tool designed for executives to access comprehensive company data, including meetings and sales calls. This allows CEOs to bypass hierarchical information filters and gain an omniscient view of operations to make better decisions.

  4. 20:06 – 29:21

    Strategic Pivot at 01.ai and Market Positioning

    01.ai shifted from building frontier models to focusing on enterprise ontology and executive tools due to resource constraints and the need for grounded, explainable AI. Lee compares their approach to Palantir, emphasizing the importance of moving beyond token economics to deliver tangible business value.

  5. 29:21 – 34:27

    Future of Work and Human Skills

    Lee posits that as AI handles execution, human value lies in right-brain leadership, creativity, and empathy. He predicts a rise in service jobs requiring adaptability and human connection, while average creative tasks will be largely automated by AI.

  6. 34:27 – 35:40

    Human Purpose and AI Limitations

    Dr. Kai-Fu Lee argues that AI surpasses humans in intelligence and repetitive tasks, implying human purpose lies elsewhere. He suggests humanity's unique value is in right-brain leadership, idea generation without data, and emotional capacities like love and family that AI cannot replicate.

  7. 35:40 – 36:27

    Lightning Round: AI and Career Insights

    In a rapid-fire segment, Dr. Lee identifies the idea that one company can build AGI and dominate as the most overhyped. He states he would never hand over decisions regarding family and love to an AI agent, and cites Google as the organization that taught him the most.

  8. 36:27 – 37:40

    Recommendations and Future Outlook

    Dr. Lee recommends Daniel Goleman's 'Emotional Intelligence' as a must-read for CEOs. He describes the next decade of AI with two words: 'exhilarating and scary,' before concluding the interview.

Assets discussed

AssetKindStanceContext
OpenAIprivate companyneutralZero one dot AI began as an answer to OpenAI. You've since stepped back from building frontier models to focus on the enterprise. What drove that decision?

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

01.ai began as an answer to OpenAI but pivoted away from building frontier models due to lack of resources compared to OpenAI, ByteDance, and Alibaba.

Dr. Kai-Fu Lee ·20:17

How to verify: Review 01.ai press releases, blog posts, or news articles from 2026-2026 detailing the company's strategic shift from foundation models to enterprise solutions. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Dr. Kai-Fu Lee wrote a letter to Carnegie Mellon in 1987 stating AI was humanity's final step towards understanding itself.

Unidentified speaker (UNKNOWN) ·34:02

How to verify: Search for archives of Dr. Kai-Fu Lee's correspondence or biographical sources citing his 2026 letter to Carnegie Mellon University. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Most middle managers do not want AI-native companies because they want to keep their jobs and power.

Dr. Kai-Fu Lee ·6:11

How to verify: This is a subjective assertion. Verify by looking for industry surveys or reports on employee sentiment regarding AI adoption, though it remains a claim of opinion rather than a hard fact. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

01.ai aims for a Hong Kong listing in 2027.

Anita Huja ·24:15

How to verify: Check for any public filings, press releases, or news reports from 01.ai regarding their IPO plans and target exchange (Hong Kong Stock Exchange). For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

AI can beat humans in decisions based on data, but transformative decisions (like the iPhone) are made without data.

Dr. Kai-Fu Lee ·19:17

How to verify: This is a philosophical/strategic claim. Verify by analyzing Lee's previous writings or interviews where he discusses the limits of data-driven decision-making versus intuition. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Frontier model companies do not want to build enterprise ontology because it slows down their token economics.

Dr. Kai-Fu Lee ·21:11

How to verify: Review public statements from major frontier model companies (OpenAI, Anthropic, Google DeepMind) regarding their focus on scaling models versus building vertical-specific ontologies. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Dr. Kai-Fu Lee held senior leadership roles at Apple, Microsoft, and Google.

Anita Huja ·1:14

How to verify: Check Dr. Kai-Fu Lee's official biography on the 01.ai website or his LinkedIn profile to confirm his tenure at Apple, Microsoft, and Google. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Dr. Kai-Fu Lee is the Founder and CEO of 01.ai.

Dr. Kai-Fu Lee ·34:27

How to verify: Check 01.ai's official website or press releases for leadership team information. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

AI is better than humans in many ways regarding intelligence and repetitive tasks.

Dr. Kai-Fu Lee ·34:27

How to verify: Review performance benchmarks of current LLMs and AI agents against human performance in cognitive and repetitive task studies. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

AI cannot do things like love, family, or right-brain leadership.

Dr. Kai-Fu Lee ·35:27

How to verify: Consult academic literature on AI capabilities in emotional intelligence, empathy, and complex social reasoning. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

The most overhyped idea in AI is that one company can build AGI and squash everyone else.

Dr. Kai-Fu Lee ·35:52

How to verify: Analyze market share and R&D investments of major AI companies; review industry forecasts for AGI timelines and competitive landscapes. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Google taught Dr. Kai-Fu Lee the most among Apple, Microsoft, and Google.

Dr. Kai-Fu Lee ·36:12

How to verify: Cross-reference Dr. Lee's professional history, which includes roles at SGI, Apple, Microsoft, and Google, with his public statements on their respective impacts. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Primary source

Daniel Goleman's 'Emotional Intelligence' is a book every CEO should read.

Dr. Kai-Fu Lee ·36:27

How to verify: Check the publication details of 'Emotional Intelligence' by Daniel Goleman and its availability in business bookstores. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Dr. Kai-Fu Lee described the next decade of AI as 'exhilarating and scary.'

Dr. Kai-Fu Lee ·36:42

How to verify: Review the transcript of the podcast episode 'Why Most Companies Fail to See Bottom-Line Returns From Deploying AI Agents' released on 2026-11-13. For recent events, verify 2026-08-13 through 2026-11-13. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify McKinsey's AI adoption statistics and methodology for the '6% bottom-line impact' claim.
  • Review 01.ai's public filings or press releases regarding their shift from frontier models to enterprise ontology.
  • Check Palantir's recent earnings reports and customer case studies to compare their enterprise AI value delivery model.
  • Assess the organizational structure of target companies to see if they have a Chief AI Officer reporting to the CEO.
  • Research Daniel Goleman's 'Emotional Intelligence' to understand the soft skills Lee recommends for leadership.
  • Investigate the market share and AGI development claims of major tech firms to evaluate the 'fragmented future' assertion.
  • Look for case studies of 'Boss AI' or similar executive tools to verify their ability to bypass hierarchical information filters.
  • Check for conflicts of interest: Lee is CEO of 01.ai, which builds the tools he recommends.
Full transcriptexpand
0:00
Dr. Kai-Fu Lee: Okay. When I sell my business, I want the best tax and investment advice. I want to help my kids, and I want to give back to the community. Ooh, then it's the vacation of a lifetime.
0:12
Unidentified speaker (UNKNOWN): I
0:12
Dr. Kai-Fu Lee: wonder if my out of office has a forever setter.
0:15
Unidentified speaker (UNKNOWN): An
0:15
Unidentified speaker (speaker_1): IG private wealth advisor creates the clarity you need with plans that harmonize your business, your family, and your dreams. Get financial advice that puts you at the center. Find your advice. advisor at IGPrivatewealth.com.
0:30
Unidentified speaker (UNKNOWN): Your
0:30
Unidentified speaker (speaker_2): Uncle Dale is the undisputed king of hot takes. Buffalo, they're a lot because his barber said so. Don't listen to Dale. Go with the score bet. Breaking news, live odds, and real time stats right at your fingertips. Everything you need for game day is one tap away. Don't go with hot takes, go with the score bet. Please play responsibly, 18 years or older.
0:49
Unidentified speaker (UNKNOWN): If
0:49
Anita Huja: gambling is affecting your mentality, mental health or well-being, help is available. In Alberta, call or text 211 or visit ab.211.ca. Standard message and data rates may apply.
1:00
Dr. Kai-Fu Lee: But if all you do is measure activity, not results, measure how many agents, not what they bring, then it serves no purpose for the company.
1:14
Anita Huja: Welcome. I'm Anita Huja, editor at large at Forbes and founder of Iconoclast, and my guest today has spent four decades at the frontier of artificial intelligence. He's held senior leadership roles at Apple, Microsoft, and Google, chairs Cinnovation Ventures, and leads 01.ai AI as
1:33
Anita Huja: founder and CEO. His best bestsellers, sellers AI Superpowers and AI 2020. 2041, change how the world thinks of this technology. And in his new book, AI Native, The Mandate to Transform Your Company, he issues a mandate to every CEO. Dr. Lee, it's a pleasure to welcome you to Iconoclass.
1:53
Dr. Kai-Fu Lee: Thank you.
1:54
Anita Huja: Dr. Lee, you argue that AI transformation cannot be delegated to the CIO. It has to start from the top down, from the CEO rather. Talk to us about that.
2:04
Dr. Kai-Fu Lee: Actually, I want I'm against the bottom-up approach, which is asking departments what budget do you need to implement AI agents, and then giving CIO the budget to help the departments do a better job with finance, legal, customer service. That should be done, but it is not going to move the dial for the company. AI is
2:24
Dr. Kai-Fu Lee: getting so smart, I'd say that they're like IQ. 300 super intelligent beings that when they enter the company, it is uh terrible uh underutilization to make them do customer service. They should help the company move the dial, make more money, make the CEO more effective, superpower the
2:44
Dr. Kai-Fu Lee: CEO, and um figure out ways to increase revenue, growth, profits, and also. in order to welcome these uh thousands of AI agents to the company where AI agents will way outnumber human uh workers, the organization chart has to change. The
3:04
Dr. Kai-Fu Lee: traditional hierarchy does not work. Um and the number of people who know how to manage AI and take accountability, there's not enough of them. So there's a huge imbalance of people who just know how to manage people, people who don't know how to manage AI. And when AI does the bulk of the work, the organization chart should be much
3:23
Dr. Kai-Fu Lee: flatter. These are all steps that a CEO needs to make. And also, as the organizational chart changes, the institutional knowledge needs to remain. So these are all tough, tough, tough problems that no one other than the CEO can handle.
3:40
Anita Huja: So what is the clearest sign that a company is running an expensive AI laboratory rather than real commercial impact?
3:48
Dr. Kai-Fu Lee: I would say about six to nine months ago, a lot of CEOs do talk to me. I'm friends with many of them. And they want to show off hey, we're K Kaifu, we're doing a lot of AI. Tell me what you're doing. Oh, we've implemented you know, 700 agents and they're being used in this department and that. And then I would turn around and say, well, how much has it moved the dial
4:08
Dr. Kai-Fu Lee: for your bottom line? Does it change your quarterly result? And comes at that silence. They're like shocked. I'm supposed to do that. I thought just having the agents would do. But if all you do is measure activity, not results, measure how many agents, not what they bring, then it serves no purpose for the company.
4:27
Anita Huja: Well you're still managing a workforce, right? You're just managing an AI workforce, right? So productivity needs to be looked at in a in a similar way, correct?
4:37
Dr. Kai-Fu Lee: Uh yes, correct. But the issue is when you think about only AI enablement, not AI native, what happens is everything works as before with human workers, and then let human workers use AI
4:51
Unidentified speaker (UNKNOWN): They
4:51
Dr. Kai-Fu Lee: can use chatbots to improve their writing and customer service and so on. And the even when they do make an impact is a very, very small, modest impact. Mackenzie showed that only 6% of all the companies that use AI see a bottom line impact. And even then the impact is not large. So it just means the
5:11
Dr. Kai-Fu Lee: CEO hasn't been driving it as use has hasn't been driving using AI to change the bottom line. So so I wrote this book to give a formulaic approach of how to change the bottom line, which I think every CEO wants to do.
5:26
Anita Huja: No, the CEO does definitely want to change the bottom line. But that disconnect that's so fascinating that you know you've had so many conversations with executives and they've taken that first step, right? They've put these AI agents into the workforce, but they didn't really have that plan mapped out. We recently just held our future of work summit
5:46
Anita Huja: with s several hundred business leaders in in the audience, all trying to figure out that problem in real time. And and uh that was also something that came up is that once they see a CEO first leading with AI, it makes the organization understand that, you know, this
6:06
Anita Huja: is a good thing and there's opportunity there to just increase productivity.
6:11
Dr. Kai-Fu Lee: Yeah. Um, I think the honest truth is most middle managers don't want AI native companies.
6:16
Anita Huja: Yes.
6:17
Dr. Kai-Fu Lee: Because they're unfamiliar with it. They want status quo, they want to keep their job, their power, their number of employees. They don't want their teams replaced with AI. They don't want to learn new skills. So that's all the more reason the CEO has to take the bull by the horn and say, hey, we're doing it. Now the question now is, what should the CEO
6:37
Dr. Kai-Fu Lee: do? And I've said the CEO can and should start with the bottom-up approach, let customer service and others use it. But very soon after that, the CEO must use AI himself or herself. Because only then can he or she set an example that I'm serious, I'm using it. And if you don't, you're not even doing what I'm doing.
6:57
Anita Huja: Right.
6:57
Dr. Kai-Fu Lee: Yet CEOs don't use much software, right?
7:00
Unidentified speaker (UNKNOWN): Right.
7:01
Dr. Kai-Fu Lee: If if anything, ERP is the most reasonable software for a CEO to use. I don't know a single CEO who does that, right? Because it's they delegate to someone, they bring the reports on paper. Right. But but AI is different. And what we've designed both in this book and in my company, 01.ai, is a
7:20
Dr. Kai-Fu Lee: product called Boss AI.
7:22
Unidentified speaker (UNKNOWN): Right. And what Boss
7:23
Dr. Kai-Fu Lee: AI is that chatbot and that AI designed for the boss. Designed for the boss to have a complete picture of what's happening in the company. So not just the databases, the sales data, the CRM. the ERP, the quarterly reports. Of course we have that. But also we
7:43
Dr. Kai-Fu Lee: want the boss to basically have every meeting captured. Uh not just the ones the boss is at, but every single meeting in the company, um, down to certain levels. If
7:53
Unidentified speaker (UNKNOWN): it's
7:53
Dr. Kai-Fu Lee: a huge company, maybe five levels, six levels. And that is the information where where where all the ideas are brainstormed. debated, discussed, decided, and strategy set. So there's so much information in the meetings. Also, all the sales calls, if the customer agrees to be recorded, then then
8:13
Dr. Kai-Fu Lee: we know if the salespeople are being effective and if the customers are being responsive, and if the claims match Matches the data. And also customer service calls, which products are broken, people complain. they get angry, and they want to refund. So we have all this data, which all always existed, but they are
8:33
Dr. Kai-Fu Lee: generally in unstructured form and way too much data for any human to digest. But now we have a superhuman. Right. Right? The AI is IQ 300. It can read, remember, all of it. And then I, as the CEO, I've been using this for nine months, I can ask questions like, uh of the thousand meetings that took place in the company last
8:53
Dr. Kai-Fu Lee: week, what are some big debates that were had that I as the CEO need to know about?
8:58
Anita Huja: Oh wow.
8:59
Dr. Kai-Fu Lee: Because my direct reports may not tell me that.
9:01
Anita Huja: Right, right. No, you're a hundred percent right. So it's able to take all of that information, synthesize it, and become like a trusted sounding board without you know, uh a relative potential bias in the mix. Right.
9:15
Unidentified speaker (UNKNOWN): I can also say
9:16
Dr. Kai-Fu Lee: I can also say what product has the most customer complaint and what's the biggest complaint about that product? And is it being fixed? I can ask who are the best managers at managing AI workers, not just human workers, because I need to elevate them because we're adding AI workers. I can also ask
9:36
Dr. Kai-Fu Lee: which of my best AI worker managers manager people are at retention risk and how can I keep them? We know that because we capture the meetings. We know someone who's turned proactive into uh passive, right? We know who signed who's someone who used to sign up voluntarily, now they stay quiet.
9:56
Dr. Kai-Fu Lee: That's a sign of retention risk. I can also ask how I can I be a more effective CEO? How can I run meetings? better? So basically, what I I will have, I do have, is a
10:09
Dr. Kai-Fu Lee: omniscient and omnipresent chief of staff who hears everything, knows everything, remembers everything, and advises me by whispering in my ears.
10:19
Anita Huja: Right. So you talked about your boss AI app, and the framework is mentioned, is detailed in depth in your book. AI Native. But drafting emails is one thing. Rebuilding the core value chain is another. How do you draw that line?
10:35
Dr. Kai-Fu Lee: Well, I think the CEO needs to jump into using this AI and then also to tell the company transparently that we're making this organizational change because I've been looking using Boss AI to understand the organizational dynamics. We need to make this change. Or we're canceling this business unit
10:55
Dr. Kai-Fu Lee: because uh Boss AI analyzed this, or we're adding resources to these people, or we're giving these great AI worker managers a big race because they're doing a super important job, or that we don't have enough people who know how to manage AI workers. So we're gonna train you middle managers, or we're gonna hire people from the outside.
11:15
Dr. Kai-Fu Lee: So the strategic decisions that are made should be transparently announced to the company. For one, to show this boss AI is not a surveillance tool. It is really helping us run the company better. You'll all get benefit as the stock goes.
11:30
Unidentified speaker (UNKNOWN): It's an
11:30
Anita Huja: efficiency tool, it sounds like
11:32
Dr. Kai-Fu Lee: it is an efficiency tool. Um, and it's boosting the boss's efficiency, which is the most valuable. It's super superpowering the boss, right? So I think that is a great first step. Then I think for the for the next layer managers, they should say, Well, can I have it too? Then, of course, a mini boss AI can
11:52
Dr. Kai-Fu Lee: be given to VPs and C-suite people. And in fact, to everybody except each person cannot look at everybody's data.
12:00
Unidentified speaker (UNKNOWN): Right,
12:00
Anita Huja: right.
12:01
Dr. Kai-Fu Lee: Their agent can only look at the data. the human can look at in a corresponding way, but that will boost the whole company's productivity. So that's an important step. And and among the and then after that, the CEO will have the HR tool, right? And also the head of HR will be able to measure people more accurately by
12:20
Dr. Kai-Fu Lee: data, who made promises and kept them, and also who's who checked in good software. that had few bugs, who sold more software and was honest about their pipeline and not sandbagging. So all these things are going to be known by metrics. So great employees will get promotion,
12:40
Dr. Kai-Fu Lee: not underperforming employees will go into PIP programs. So the company will really become higher quality. And then we need to think about re- re-reorganizing the company. We need to have, as we have more AI workers, we need to elevate people into managers of AI workers. Because AI
12:59
Dr. Kai-Fu Lee: workers on important tasks, they cannot take accountability.
13:03
Anita Huja: Right. That was a good I watched an earlier interview you did. And that was such a good point because they make a mistake. And I've even just encountered this in my own back and forth with AI, where as a journalist
13:16
Anita Huja: being accountable with wrong information is pretty serious, but there's nothing I can do about it. I can't get into an argument with my AI. I have to make sure that the proper process is in place so that you are getting the right facts in order to m create your own framework. So talk to us a little bit about that. And then also I wanted to ask
13:35
Anita Huja: that you believe the chief AI officer must report directly. to the CEO. Why does that reporting line also matter so much?
13:44
Dr. Kai-Fu Lee: Yes, because uh some CEOs are really still going to be not fully confident that they understand AI. So having a technical AI person report to the CEO as that assistant is important. Uh CIOs are an important function in the company. but they typically have
14:04
Dr. Kai-Fu Lee: IT background, not AI background. So a good AI chief AI officer is going to be more aligned with the CEO's direction. I think alignment is what is what's important. One reason the CEO is in charge is the CEO's goals are fully aligned with the company's goals. And the
14:23
Dr. Kai-Fu Lee: CIO's goals may not be fully aligned. It might be to save money, it might be to increase 24 by 7 support and so on and so forth. But AI used to transform the company is a challenging and sometimes risky proposition. And CIOs are inherently conservative. So I would argue someone who
14:43
Dr. Kai-Fu Lee: knows AI and whose job is aligned with AI transformation and aligned with the company's future that he or she re uh uh evaluates the AI opportunities the same way as a CI CEO does, that would be a good arrangement. That's not always necessary. Sometimes you have a great CIO, sometimes you have a
15:03
Dr. Kai-Fu Lee: great chief strategy officer, sometimes the CEO is pretty technical. So it's not always needed. In
15:09
Unidentified speaker (UNKNOWN): my
15:09
Dr. Kai-Fu Lee: company, I have no chief AI officer.
15:11
Anita Huja: And that's such a good point, actually, because there also is so much. apprehension and anxiety in the workforce about what this is going to mean for the greater, you know, for the greater economy. And so having somebody who truly understands AI, which is different than the CIO function, will also present more opportunity because they can
15:31
Anita Huja: leverage the tools better and safer and more efficiently.
15:35
Dr. Kai-Fu Lee: Uh Yeah, I think the new new organization uh will need to evolve towards a flatter, more delegated.
15:43
Dr. Kai-Fu Lee: These um people who manage AI is actually as important as the chief AI officer because essentially the CEO is gonna pick really important tasks. Some of them are like running business units that are long term, others might be this product needs to get growth back on track. I want I want a special force
16:03
Dr. Kai-Fu Lee: to do that. And you're appointed as the designated responsible individual, DRI, or the manager of the AI to do that. And you may have some people reporting to you, but more AI to get the growth back on track for this product. This other group really important, the DRI may be responsible for coming up with a new product
16:23
Dr. Kai-Fu Lee: that serves a particular purpose or customer sets.
16:27
Dr. Kai-Fu Lee: And this other DRI may be responsible for fixing a customer service that dissatisfaction. And some of these DRIs, their job will be done. Some of them, you'll be great, you got growth this year. Keep it, keep it for next year. So I think the main metric for these DRIs is that they're
16:47
Dr. Kai-Fu Lee: fully accountable. They're not just rubber stamping. AI decisions, but they're actively managing them, evaluating them, approving them. And when something goes right or wrong, they're accountable and they get the upside and the downside. And also the compensation structure for the DRIs needs to change because they're taking so much
17:07
Dr. Kai-Fu Lee: accountability and risk and and upside. And they got a big appointment from the company and they're empowered. The CEO will lightly manage them and empower them greatly. And as such, they should be like a general manager and
17:20
Unidentified speaker (UNKNOWN): be
17:20
Dr. Kai-Fu Lee: paid paid accordingly. So it's
17:22
Unidentified speaker (UNKNOWN): the
17:23
Anita Huja: incentives are aligned.
17:24
Dr. Kai-Fu Lee: Yeah. And also the best people will want to become DRIs and then that just like the best people want to be VPs or fellows or some such job today, uh those are job categories of the past. Now the best people want to be DRIs. So they need to be accountable end to end, understand the business process, manage AI well,
17:44
Dr. Kai-Fu Lee: and work well with other people.
17:46
Anita Huja: Yeah. And that is a good point. I recently interviewed Dean Erica James at Wharton, I think
17:52
Anita Huja: maybe just about a year ago, it feels like a lifetime ago. And I asked her, How are you adapting your coursework at Wharton for the AI age? And she said, Minnie, you'd be surprised, but so much. of it is on the soft skills. And we're creating, we're creating professionals that other people want to work with because the AI is moving so quickly, it's not even,
18:12
Anita Huja: you know, uh reasonable that we're gonna be able to adjust our curriculum by the time one class semester is over, AI will have, you know, a a new model out week over week. So what we can do is we can build people who are flexible.
18:27
Anita Huja: and and collaborative.
18:30
Dr. Kai-Fu Lee: Yeah, I think the important skill set at Warden and other great schools is really to make sure that the students are able to learn. I always believe that education is what's left when you've forgotten everything you've learned. And by now most of what's college students learn are not very useful skills anymore. But the process of learning is
18:50
Dr. Kai-Fu Lee: useful. Being adaptive, learn quickly, and also, as the dean said, build up the right brain leadership so that the
19:03
Dr. Kai-Fu Lee: taste, empathy, communication skills, teamwork, the soft skills, and also the courage, conviction to become a DRI. And the force foresight to
19:16
Unidentified speaker (UNKNOWN): make
19:17
Dr. Kai-Fu Lee: decisions that are not based on data. Because any decision based on data, AI can beat us. But there are decisions that can be made. The most dramatic, great decisions are made without data because these are transformative, like the iPhone, the recyclable rocket. There's no precedence. AI would never do that. But Steve Jobs and Elon Musk did.
19:37
Dr. Kai-Fu Lee: And of course, to learn to manage. AI. So all the students must become masters in running many AI agents to solve problems. Now, skilled with this set of capabilities, which is very different from a traditional curriculum, will really well prepare a student for the future of work.
19:57
Anita Huja: So you're not only writing about this transformation, you've lived it, including a major pivot at your own company.
20:06
Anita Huja: Zero one dot AI began as an answer to OpenAI. You've since stepped back from building frontier models to focus on the enterprise. What drove that decision?
20:17
Dr. Kai-Fu Lee: A part of it was just that uh the that building front frontier models require a lot of money and GPUs, and a um small company like us are not going to get anything close to the resources, not to mention, not compared to OpenAI, but also not compared to ByteDance or Alibaba in China. So we're
20:37
Dr. Kai-Fu Lee: bound to lose in building frontier models, even though our first models were quite good, but we just didn't have the money, the the billions of dollars needed to build the frontier models. But the other thing is that as I talk to CEOs, I find that frontier models are not enough because
20:54
Dr. Kai-Fu Lee: large companies have so much data, you can't throw the file cabinet at the model. It gets confused. So we found that what's needed are a couple of things.
21:09
Dr. Kai-Fu Lee: When that is built,
21:11
Dr. Kai-Fu Lee: it allows the frontier model to actually understand the context of the company, the org structure, the decision-making process, the workflow, the uh the company mission, and so on. So building that is something that frontier model companies don't want to do because to them it slows down their token economics.
21:31
Dr. Kai-Fu Lee: But but it's something the companies need to get in an A performance. Uh the frontier models just want to sell a C plus model. Um it's still worth selling because without it, it's at a D minus level without the AI. But now I think every company that chooses to put a resource into optimizing AI for its company can
21:51
Dr. Kai-Fu Lee: get an A solution, and that's what we provide. And we found that that was an interesting angle to be completely compatible working with large language model frontier companies and building on top of them. Also, these companies want their ontology to be hosted on their computer. They don't want to send it out because it's company secret. Right.
22:10
Unidentified speaker (UNKNOWN): If you
22:11
Dr. Kai-Fu Lee: send it to a OpenAI, they will suck it in and learn from it and maybe teach the competitors. The other thing we found is that if we want AI to be used by the CEO or senior executives, the hallucination problem must be dealt with. Right. So having the ontology.
22:27
Dr. Kai-Fu Lee: will reduce the hallucination because it's grounded on company truths and facts and important facts for the company. Secondly, we have a logical process of explainability so that every suggestion the boss AI makes to the CEO is backed up by an evidential chain that the
22:47
Dr. Kai-Fu Lee: CEO can examine and say, well, your logic is sound. that's good, or your logic is not sound, I'm not listening to you. So we build these in, and we I realize that companies actually need these two technologies, and it's built on top of large language models. Then we also saw that, uh, as I mentioned earlier, the boss uh through
23:07
Dr. Kai-Fu Lee: hierarchies, that a lot of information is uh mistransmitted because people have their own self-serving needs. people misunderstand things, and people have limited, sometimes limited capacity of knowing the company strategy. And also people don't want to tell the boss the bad news. Right.
23:25
Unidentified speaker (UNKNOWN): So they
23:25
Dr. Kai-Fu Lee: sugarcoat. So this upward downward error tr error transmission is a limiting factor for companies. And AI can blast through that by making all the data available to the boss. And then the hierarchy. can be flattened. And that leads to an effective and consistent system to manage all the AI
23:45
Dr. Kai-Fu Lee: agents and the people.
23:46
Anita Huja: Potentially the end of the middle manager.
23:49
Dr. Kai-Fu Lee: Well, the the change, right? For the middle manager who wanna become a DRI can. Some middle managers are good coaches, so because there's some young DRIs who need senior coaches. Um for the middle managers who just pass information, run meetings, and uh write people's
24:09
Dr. Kai-Fu Lee: reviews, uh, if they don't learn the new skill sets, they are not needed anymore.
24:15
Anita Huja: So you're also aiming for a Hong Kong listing in 2027, I believe. What do public market investors need to understand about enterprise AI?
24:25
Dr. Kai-Fu Lee: Yeah, well we want to file,
24:28
Anita Huja: but
24:28
Dr. Kai-Fu Lee: filing an IPO
24:29
Unidentified speaker (UNKNOWN): is we don't we
24:30
Dr. Kai-Fu Lee: don't know how long that takes in in Hong Kong. There's a very long queue. We hope we can list soon. Um, I think the public market is currently obsessed with um the token economics. And I think there's great value in token on economics, but there's not yet a sufficient understanding that you need a layer on
24:51
Dr. Kai-Fu Lee: top to harness your token economics, to make it relevant for the company, to make it provide value for the boss. So that's why writing this book is very timely. I think there are other companies who understand this. Palantir is clearly the leader in the space, and we've heard Alex Carp talk many times about it's not just token
25:11
Dr. Kai-Fu Lee: economics, that's a C plus solution. Um And I think what is needed is not for me or Alex or anyone else just to be the uh the evangelist telling the world because those are just our words against entropic and open AIs. What's important is we need to get some great customers who can showcase that
25:30
Dr. Kai-Fu Lee: this set of technologies I talked about is really moving the dial for the companies. And we're quite close to that. I think within the next six months we will have custom customer uh testimonials that shows that when you put in this much work, you get a A plus result or A minus result as opposed to the C plus token
25:50
Dr. Kai-Fu Lee: economics result. If we don't explain that to the public market, I think people will just have to evaluate our numbers and apply
25:57
Unidentified speaker (UNKNOWN): a
25:58
Dr. Kai-Fu Lee: P PE E or P PS S ratio, which may be okay anyway, but I
26:01
Unidentified speaker (UNKNOWN): But
26:01
Anita Huja: doesn't always tell the full picture.
26:03
Dr. Kai-Fu Lee: No,
26:03
Unidentified speaker (UNKNOWN): no, no.
26:04
Dr. Kai-Fu Lee: So I'm hoping we can have deliver some great customers. And we have uh many large companies starting to use our products. We have companies in manufacturing, natural gas, um uh coal coal min coal mining, um, uh iron mining, and and many other areas. Uh and traditional companies really need to do this because
26:23
Dr. Kai-Fu Lee: otherwise um small AI native companies. will start to punch holes in the um the fortress they thought was secure uh with purely human workers. It's not secure.
26:37
Anita Huja: So that's so fascinating. So from all the companies that are successful clients of yours, are you seeing some immediate use cases or industries where it naturally the technology lends itself better or is are you seeing that flat usability and value across the board?
26:57
Dr. Kai-Fu Lee: Well, the as a management tool, BOSS AI for AI transformation, and also we build tools for the top salesperson, the top um investment person, and also the HR transformation, those are fairly general. But but companies want their the CEOs want their whole company to
27:17
Dr. Kai-Fu Lee: become AI native. So there's also a need for their supply chain to become AI native. There's a need for their product and sales cycle and customer feedback to become AI native. So there are a lot of opportunities to do that. To do it one industry at a time is a little bit difficult. We prefer to look for things that are
27:37
Dr. Kai-Fu Lee: across industry by now. nature. So lots of companies have supply chains. Lots of them have uh logistics issues. So those are kind of angles we're cutting in to these. And um to the extent, and also some industries ha see more immediate benefit, right? AI is
27:57
Dr. Kai-Fu Lee: much better in things that can be applied, and then results are demonstrated, and then the cycle continues with uh better AI makes better product, makes a better user experience, they pay more money and the cycle goes on.
28:11
Anita Huja: It's
28:11
Unidentified speaker (UNKNOWN): that
28:11
Anita Huja: AI flywheel that you talk about.
28:13
Dr. Kai-Fu Lee: Yeah, that flywheel is better when it's a purely digital asset. So things like content creation, uh video creation, financial industries, um, software industries. So these are the ones that are relatively um uh instant gratification. Whereas actually the earlier
28:32
Dr. Kai-Fu Lee: examples I gave, the natural gas and iron mining, it takes a little bit longer.
28:37
Anita Huja: A lot longer, but potentially even you know, uh on a much bigger scale, you could see the end result of that productivity play out.
28:45
Dr. Kai-Fu Lee: Yes, especially if we do supply chain as as one uh virtuous cycle. and a natural gas company has a supply chain and our mining company has a supply chain, so we can take that angle as well.
28:59
Anita Huja: So we're almost out of time, but that brings me to perhaps one of the most important questions. When machines can do so much of the thinking, what remains distinctly human, right? So and you've spoken so much about this, as machine intelligence becomes a abundant, why do taste, conviction, courage become a leader's
29:19
Anita Huja: greatest advantage?
29:21
Unidentified speaker (UNKNOWN): Uh
29:21
Dr. Kai-Fu Lee: yeah, I think these are things that don't come naturally for in intelligence, right? Intelligence execution is generally regarded as left-brain. We're gonna have to rely on AI for that. Now we we are able to manage AI, so that job remain remains, but bulk of the work is done. by AI. But the right brain
29:41
Dr. Kai-Fu Lee: leadership is what brings uh ingenuity, new ideas, new product concepts, risky things, pat things that we have a passion. What separates us from AI is that when we believe in something like Elon Musk's um recycling, uh recyclable uh rockets, uh Steve Jobs' iPhone,
30:00
Dr. Kai-Fu Lee: or when I decided to work on AI for
30:02
Unidentified speaker (UNKNOWN): the
30:02
Dr. Kai-Fu Lee: forty five years ago, these are without past data. It's not like data shows AI will make me a lot of money,
30:09
Unidentified speaker (UNKNOWN): or
30:09
Dr. Kai-Fu Lee: that data shows the Nokia users want an iPhone. We don't have the data, but but just somewhere in our gut we feel this is it. This is it for me and this is for the world. And if it doesn't work out, I have no regrets. And if it works out, it would be the li opportunity of a lifetime.
30:26
Anita Huja: So
30:26
Dr. Kai-Fu Lee: I think leaders who feel that level of conviction, the willing to make the big bets, jump into it and accept the consequences if it doesn't work out, that's I think what will separate the the great leaders. And also great leaders will have charisma leadership that people will follow, that they will have
30:46
Dr. Kai-Fu Lee: taste and empathy and communication skills and teamwork skills. So those I think are also important for leaders, but also for all the employees. And if we move away from leaders, I think the new jobs I think that will emerge in quantity, or in quantity, will be service jobs, jobs that have to do with
31:06
Dr. Kai-Fu Lee: human-human connection. And there will be uh problems that are open, not closed.
31:11
Anita Huja: Yeah.
31:12
Dr. Kai-Fu Lee: For example, assembly lines close. You're just doing this all the time.
31:15
Anita Huja: Right.
31:15
Dr. Kai-Fu Lee: But a plumber, every leak is different. Right. Right.
31:19
Dr. Kai-Fu Lee: A how a house cleaner, every house is different. Right. And every homeowner may want a different type of cleaning, right? So being able to customize and deal with any environment, any house, any
31:32
Unidentified speaker (UNKNOWN): adaptability.
31:33
Dr. Kai-Fu Lee: Adaptability and also interaction with people, that's I think where the bulk of the jobs will will will will move to. And eventually AI will create new exciting jobs. We just don't know what they are yet.
31:45
Anita Huja: And so, you know, just we don't know what they are yet, but the future and the frontier models are super exciting. Um I want to touch on briefly that you've spent your career bridging Silicon Valley in China. So I want to widen that lens for a minute. It's been several years since AI superpowers, the year. your previous one of your previous books. What did that
32:04
Anita Huja: book get right? And what would you revise in the world today? I I watched 60 Minutes interview with you from a couple of years ago. You were predicting these changes that we're gonna see in the global economy and workforce, and many of them have come true.
32:19
Dr. Kai-Fu Lee: I think uh what the predictions that are right include um the challenges, uh future of work, uh it includes um China and US will co-lead this AI revolution. It includes the importance of data.
32:45
Dr. Kai-Fu Lee: Along and it has. So these are things I got right. Some of the um things that I didn't quite get right, uh, one is that I thought creativity was something that we we could hold on to. But it seems like only extreme creativity we can hold on to. Like the best architect, the best
33:04
Dr. Kai-Fu Lee: movie uh producer, the best actress, the best um uh fiction writer uh will continue to be sought after. But most of the average or above average and certainly below average will AI will do a better job for a lower price and appeal to consumer taste better. So that kind of
33:24
Dr. Kai-Fu Lee: creativity, whether it's graphic artist, drawing a picture or making a piece of art or craft or um uh making a movie or a book uh content, I think uh that kind of creativity is
33:39
Dr. Kai-Fu Lee: taken over by AI faster than I would I predicted in 2018.
33:44
Anita Huja: That makes me a bit sad. I can't lie. But before we close, I'd like to go back to where your journey began. In 1987, you wrote at to Carnegie Mellon that AI was humanity's final step towards understanding itself. Four decades later, how has your view of that destination changed?
34:02
Unidentified speaker (UNKNOWN): I
34:02
Dr. Kai-Fu Lee: think we are. at that final step, and that is the exhilarating thing. It didn't quite come about the way I thought it would. I thought it would emulate the human brain, it would be a counterpart of the human brain. We'd understand ourselves better because we were able to replicate much of the human brain, but instead we built a
34:22
Dr. Kai-Fu Lee: completely different um species that it's it ingests huge amounts of
34:26
Unidentified speaker (UNKNOWN): the human being.
34:27
Dr. Kai-Fu Lee: data. Data and compute makes it super smart in general. Uh it's better than us so much in many ways. We're still better than it in some ways. Uh, but I think that still helps us understand ourselves because that which we do that can be done by AI is not the purpose that we're on earth, right? We're put on earth
34:47
Dr. Kai-Fu Lee: by whether it's God or some other um uh um uh supernatural uh being or another AI, whoever put it whoever put us here didn't put us here for our intelligence, didn't put us here for execu executing re repetitive tasks. So I think we need to really rethink what's the reason that
35:06
Unidentified speaker (UNKNOWN): we
35:07
Dr. Kai-Fu Lee: our purpose. So maybe it is that right brain leadership. It is about being able to come up with ideas without data. Maybe it is to manage AI better, or maybe it is something we have not yet discovered, but now we'll have time to discover. Maybe it is a family, maybe it is love. And these
35:27
Dr. Kai-Fu Lee: are things that AI cannot do. So I think our humanity, understanding of ourselves on what's most important, we can now pretty much eliminate anything AI can do is not the reason why we're here. So we need to find that answer.
35:40
Anita Huja: So we can't wrap Iconoclast without the Iconoclast lightning round. Are you game, Dr. Lee?
35:46
Dr. Kai-Fu Lee: Sure.
35:46
Anita Huja: Okay. So what is the most overhyped idea in AI right now?
35:52
Dr. Kai-Fu Lee: That one company can build AGI and squash everyone else.
35:55
Anita Huja: Oh, interesting. One decision you will never hand to an agent.
36:01
Unidentified speaker (UNKNOWN): Um
36:03
Dr. Kai-Fu Lee: Family Love.
36:05
Anita Huja: Mm-hmm. Okay. Apple, Microsoft, or Google, which organization taught you the most?
36:12
Dr. Kai-Fu Lee: Google taught me the most.
36:14
Anita Huja: But
36:14
Dr. Kai-Fu Lee: Apple and Microsoft taught me a
36:15
Unidentified speaker (UNKNOWN): lot
36:15
Dr. Kai-Fu Lee: too.
36:16
Anita Huja: Taught you a lot too, yeah. It's a it's a hard uh trifecta there to beat. Every book, uh a book that every CEO should read other than your own.
36:27
Dr. Kai-Fu Lee: Um, Daniel Goldman's Emotional Intelligence.
36:31
Anita Huja: Okay, that's a good one. I'm gonna order that today. Because
36:33
Dr. Kai-Fu Lee: that's my brain.
36:34
Anita Huja: Yeah. One word for the next decade of AI.
36:40
Dr. Kai-Fu Lee: Uh
36:41
Unidentified speaker (UNKNOWN): two
36:42
Dr. Kai-Fu Lee: words: exhilarating and scary.
36:45
Anita Huja: And scary. Okay. Well, with that, Dr. Lee, thank you so much for this fascinating conversation. and congratulations on your latest book, AI Native: The Mandate to Transform Your Company. And that's Iconoclast.
37:09
Unidentified speaker (speaker_1): My teammate
37:10
Unidentified speaker (speaker_2): says he plays pickup basketball, quote, for the cardio. Well, there's more to it than that, Sean. Like with Bet365, I get paid the second my team goes at big with early payout and can cash out at any time before the game ends. Sean's getting his steps, but I'm getting paid. Bet365, the home of winning early. 18
37:30
Unidentified speaker (speaker_2): plus, please play responsibly. Upper to only terms, conditions apply.
37:33
Unidentified speaker (UNKNOWN): If
37:33
Unidentified speaker (speaker_1): gambling is affecting your mental health or well well-being, being, 211 two one one Alberta is here to help. Call or text 211 two one one or visit ab.211.ca. Standard message and data. rates may apply.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights