Transcriptly

How to Stay Great When AI Is Good Enough | Matt Beane, UC Santa BarBara

EO16:39Added Aug 31, 2026

Matt Beane, Associate Professor of Technology Management at UC Santa Barbara & author of The Skill Code, breaks down why AI's endless flood of B+ work is a t...

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:002025 was the year of try everything with
  3. 00:04AI, get results with AI, do your very
  4. 00:06best, experiment, learn, licenses for
  5. 00:09everyone. 2026 is the year that the
  6. 00:11board [music] of directors is coming to
  7. 00:13the CTO and saying, "Where is our return
  8. 00:16on our investment?" Jensen Hang from
  9. 00:18Nvidia is very famously on interviews
  10. 00:21saying, "If I'm paying $500,000 for an
  11. 00:24engineer, they need to be burning at
  12. 00:25least $250,000 worth of tokens." You can
  13. 00:28burn many many many tokens very
  14. 00:31inefficiently. All AI can do most of the
  15. 00:34time is create B+ content. It will just
  16. 00:36lots of free B+ content and you will
  17. 00:39forget what an A+ looks like. A wise
  18. 00:41[music] leader in an organization will
  19. 00:44reward people with cash or promotion or
  20. 00:48visible recognition for stopping a B+
  21. 00:52idea. I'm Matt Bean. I'm an associate
  22. 00:56professor at UC Santa Barbara in the
  23. 00:58technology management department and
  24. 01:00also CEO and co-founder of Skillbench, a
  25. 01:03startup in the AI enablement space.

  26. Don't Let AI Dumb You Down

  27. 01:17I wrote an article about this called
  28. 01:18don't let AI dumb you down and I put
  29. 01:21that on my Substack. That was 2 and a
  30. 01:23half years ago. This immediately I saw
  31. 01:24this technology I went this is a major
  32. 01:27risk. I can produce many many lots of
  33. 01:29lines of code. I can produce documents
  34. 01:31for free or cheap. Should I have done

  35. The B+ Trap

  36. 01:33that? Some experts uh say that the most
  37. 01:37valuable function of expertise is to say
  38. 01:39we will not do that thing. That is a B+
  39. 01:42idea. That is not an A+ idea. We will
  40. 01:45skip the B+ ideas. unhealthy
  41. 01:47organizations, unhealthy uh you know
  42. 01:49people who create problems for
  43. 01:51themselves will see lots of ideas and go
  44. 01:53do them and it turns out they produce
  45. 01:55lots of B+ ideas. That's not too good.
  46. 01:58You have to have restraint. You have to
  47. 02:00only do the maximum quality thing. Um AI
  48. 02:04will make that harder. So then even when
  49. 02:06you're in a task, you produce some
  50. 02:08content and some output. In general, it
  51. 02:11could be good, but there are quality
  52. 02:12problems inside. If you haven't made
  53. 02:14yourself learn to think, write, and do
  54. 02:18whatever this expert process is you're
  55. 02:20trying to do very well, you will miss
  56. 02:23these quality problems. [music] And then
  57. 02:26you slowly and subtly prevent yourself
  58. 02:28from learning and you might even lose

  59. The Trillion-Dollar Deskilling

  60. 02:31skill. I said in the book that this is a
  61. 02:34trillion dollar problem and I think it's
  62. 02:36really true. It is actually true that if
  63. 02:38we don't take active steps to protect
  64. 02:41ourselves from this, we will deskkill
  65. 02:43ourselves, other people, the next
  66. 02:45generation, [music]
  67. 02:46that will harm the economy, not right
  68. 02:48now, but a few years from now. The

  69. Fighters vs. Coasters

  70. 02:52difference between uh those who care and
  71. 02:54are trying to protect their learning and
  72. 02:56those who are just pushing through and
  73. 02:58using AI, one is to your disposition.
  74. 03:02Are you just driven to make sure that
  75. 03:05you can do your own independent work and
  76. 03:07get your own independent results? This
  77. 03:08is some sometimes just something about a
  78. 03:12given individual that makes them fight
  79. 03:14and protect their learning. But a lot of
  80. 03:16it also has to do with incentives and
  81. 03:18the social structure you're embedded in.
  82. 03:21If the nature of my job is to produce
  83. 03:24more lines of code as a software
  84. 03:26engineer, I'm focused a lot on software
  85. 03:27engineering today. Or burn lots of
  86. 03:29tokens. This is another one that is very

  87. Burning Tokens, Running in Circles

  88. 03:31common now. Jensen Huang from Nvidia is
  89. 03:34very famously on interviews saying, "If
  90. 03:36I'm paying $500,000 for an engineer,
  91. 03:39they need to be burning at least
  92. 03:40$250,000 worth of tokens." You can
  93. 03:43[music] burn many, many, many tokens
  94. 03:45very inefficiently. It's like saying, "I
  95. 03:48burned a lot of uh calories today." That
  96. 03:51this could be okay, but did you just run
  97. 03:53in circles? The immediate first question
  98. 03:56is, is that a B+ idea or an A minus
  99. 03:59idea? You have to set some threshold. A

  100. Reward the B+ Killers

  101. 04:01wise leader in an organization will
  102. 04:04reward people with cash or promotion or
  103. 04:08visible recognition for stopping a B+
  104. 04:12idea from making it forward. You have to
  105. 04:15do more to create incentives for people
  106. 04:19to insist on superb ideas only. [music]
  107. 04:23And then someone who does pick an A+
  108. 04:25idea will win. and you'll go, "Wait, I
  109. 04:28didn't know you were allowed to do
  110. 04:29this." And [music] it's like, "Yeah,
  111. 04:31this is how innovation works. You have
  112. 04:32to be very patient. You have to be
  113. 04:34persistent. You have to focus on, no,
  114. 04:36that's not good enough."

  115. What Shadow Learners Reveal About How We Build Skill

  116. 04:40In 2018, I published the first paper out
  117. 04:43of my dissertation, which was on robotic
  118. 04:46surgery, uh, which generated this idea
  119. 04:48of the shadow learning process. Shadow
  120. 04:51learning is building skill through
  121. 04:53nonapproved means, inappropriate means.
  122. 04:55Why would people do this? The reason is
  123. 04:58that in current circumstances in
  124. 05:00surgery, but I have since checked this
  125. 05:02in over 35 occupations, when a new
  126. 05:04technology arrives that disrupts the way
  127. 05:06that you do work, it also disrupts the
  128. 05:09way that you're supposed to learn on the
  129. 05:10job. What most of our skill comes from
  130. 05:12doing the work, and most of that comes
  131. 05:15with as we interact with experts on the
  132. 05:17job. The new technology allows the
  133. 05:19expert to do more work independently,
  134. 05:21which means that there's less room for
  135. 05:23the novice to participate. So you show
  136. 05:26up to the work, you hope to learn and
  137. 05:28build skill just by doing, and the
  138. 05:30expert is on Zoom, never comes to work.
  139. 05:33I don't get to interact with them. What
  140. 05:35am I learning after a year, 2 years?
  141. 05:37[music] Not as much. Most people will
  142. 05:40continue to try to learn the
  143. 05:42old-fashioned way. you know I will try
  144. 05:44to find ways to uh get involved in
  145. 05:47projects or something like this
  146. 05:49[music]
  147. 05:49um and they will struggle but a very
  148. 05:52rare few uh set of individuals will
  149. 05:55[music]
  150. 05:55instead of trying to participate
  151. 05:57normally they will find norm and
  152. 06:00rulebending culturally inappropriate
  153. 06:03ways to build skill. So for example in
  154. 06:05robotic surgery medical residents who
  155. 06:08were effective at building skill given
  156. 06:10this barrier to participation they uh
  157. 06:14found ways to operate on patients
  158. 06:16without the senior surgeon in the room.
  159. 06:18It's not illegal but that is not
  160. 06:20appropriate for a junior resident to be
  161. 06:23operating on a patient alone. But this
  162. 06:25is what the residents who were very good
  163. 06:27at building skill quickly did. They also
  164. 06:30for instance spent a lot of time on
  165. 06:32YouTube watching recorded video of
  166. 06:34surgery like a hundred times as much as
  167. 06:37a normal resident. So shadow learning is
  168. 06:40I can't build skill the normal way. So
  169. 06:42now I have to invent deviant or
  170. 06:44rulebreaking bending ways to build my
  171. 06:47skill. I've found this in every other
  172. 06:50occupation I've looked at. These are not
  173. 06:52practices we should copy. Obviously it's
  174. 06:54not appropriate. However, these people
  175. 06:57are fighting to protect some things in
  176. 07:00their work experience that will allow
  177. 07:02them to build skill.

  178. The Skill Code - Three C's

  179. 07:06Figuring out what that is, what are they
  180. 07:08fighting to protect was the purpose of
  181. 07:10writing my book. The first third of that
  182. 07:12book, I looked across all the shadow
  183. 07:14learners that I found across many
  184. 07:16contexts to figure out what do they have
  185. 07:19in common. In fact, none of them
  186. 07:22explicitly identified these
  187. 07:24characteristics like I have a plan. I'm
  188. 07:26going to protect these things in my
  189. 07:28work. But if you look at their behavior,
  190. 07:30you can see what they were fighting to
  191. 07:31protect. And that's the first third of
  192. 07:33my book. Challenge, complexity, and
  193. 07:36connection. The three C's in the skill
  194. 07:38code. So the point is the shadow learner
  195. 07:40is a critical
  196. 07:42source to go to to figure out what are
  197. 07:45they fighting to protect. What can we do
  198. 07:47to keep working conditions [music]
  199. 07:49healthy for productivity but also for
  200. 07:52skill development? Not copy them but
  201. 07:55they are a diagnostic in they they give
  202. 07:57you critical diagnostic information. So

  203. Challenge - Learn at the Edge

  204. 08:00challenge is the one that is perhaps
  205. 08:02most intuitive for folks which is you
  206. 08:04need to be close to but not at the edge
  207. 08:07of your capability to learn to build
  208. 08:10skill while you're working. So it must
  209. 08:12be very difficult require a lot of
  210. 08:14focus. You're not happy and relaxed.
  211. 08:17It's very intense experience somewhat
  212. 08:19stressful and you're not performing at
  213. 08:21your best a little bit less than your
  214. 08:23best because you're straining. [music]
  215. 08:25It's it's a challenge. You must have a
  216. 08:27healthy amount of this challenge in
  217. 08:28order to really build skill wealth.
  218. 08:30Another though is that an expert is
  219. 08:32there to help you manage the frustration
  220. 08:35associated [music] with that challenge.
  221. 08:36It's natural when you try to do
  222. 08:38something that is at the edge of your
  223. 08:40capability that you will fail in small
  224. 08:41ways. This is required. When you fail,
  225. 08:45you get frustrated. If an expert is
  226. 08:47present and doing a good job, they can
  227. 08:50help you put this failure into context
  228. 08:52[music] to say, "Yeah, you tried to do
  229. 08:54this and you didn't quite succeed, but
  230. 08:57you're now capable of attempting this
  231. 08:58task. Last week, you couldn't even
  232. 09:00attempt." So, they helped you sort of
  233. 09:02understand that your frustration is
  234. 09:04natural and that in fact, you have
  235. 09:05already progressed quite a bit. The
  236. 09:07expert [music] has an important role to
  237. 09:08play in helping you process the
  238. 09:12frustration of challenge enough so that
  239. 09:14you can continue to push yourself

  240. Complexity - See the Whole System

  241. 09:16further. The next of those is
  242. 09:18complexity, which is about how you
  243. 09:21digest and engage with the broader work
  244. 09:24experience that you're embedded in, not
  245. 09:26just the focal task. So the focal task
  246. 09:28for a surgeon might be making a suture
  247. 09:30well and then tying a knot. These are
  248. 09:32very focal skills that I have to get
  249. 09:34very good at. But a good surgeon that's
  250. 09:36learning well is also engaging with the
  251. 09:39nurse, the scrub tech, the supply in the
  252. 09:42hospital, the finances for the hospital,
  253. 09:44the computer and the IT system. They're
  254. 09:47trying to understand the entire system
  255. 09:49of work that they're embedded in. And
  256. 09:51they give themselves time to reflect on
  257. 09:53the whole system so that they can
  258. 09:55process and understand it. Engaging with
  259. 09:57complexity but makes you better able to
  260. 10:00handle surprise and discover new ideas
  261. 10:03because I'm more attuned to the system.
  262. 10:05The individual, you must always preserve
  263. 10:07space and time to reflect on the
  264. 10:09circumstance and look left, look right,
  265. 10:11try to understand. You yourself can do a
  266. 10:14great deal to be curious and try to
  267. 10:17engage with the broader system. At the
  268. 10:19same time, performance pressures in jobs
  269. 10:22are very high and [music] it's very hard
  270. 10:24for individuals to devote the time and
  271. 10:27effort to do this kind of engagement and
  272. 10:28leaders [music] in organizations can do
  273. 10:30a lot to enable this kind of success. So
  274. 10:33for instance, you can have job rotation
  275. 10:35program. I've seen two warehouses where
  276. 10:38in warehouse 1, [music] people just go
  277. 10:40in, get on the line, and do their job.
  278. 10:42And they get paid the same wage as
  279. 10:44someone over here who um will learn more
  280. 10:47in uh warehouse B because they come in
  281. 10:50and for 2 to 3 days they work on the
  282. 10:52line, they put their item in the bag and
  283. 10:53then they are rotated to a different
  284. 10:55part of the line. And [music] these
  285. 10:57rotations give you an exposure to
  286. 10:59different parts of the experience. same
  287. 11:02wage, same job, same job title, but that
  288. 11:05person is going to be more adaptive.
  289. 11:07[music]
  290. 11:08Interestingly, the leaders in those
  291. 11:09organizations are not doing it to try to
  292. 11:11develop employees. When I ask them, it's
  293. 11:14very fascinating. They're doing it to
  294. 11:16create resiliency in the line. What
  295. 11:18matters to them is that the building is
  296. 11:20an effective processing unit. That means
  297. 11:22it needs to be able to handle surprise.
  298. 11:25It needs to be able to notice problems
  299. 11:28and catch quality issues. And then the

  300. Connection - Trust Is a Skill

  301. 11:30third C which is connection which is a
  302. 11:32bond of trust between human beings.
  303. 11:35trust and respect actually I should say
  304. 11:37we don't often think of bonds of trust
  305. 11:39and respect as critical for learning
  306. 11:42except actually if you reflect on your
  307. 11:43own work experience anybody and you
  308. 11:45think about when did I learn the most
  309. 11:47and why often it's a who there is a
  310. 11:51person the person who trusted you who
  311. 11:53gave you an opportunity who gave you
  312. 11:55some healthy feedback that you wanted to
  313. 11:58earn their trust and respect by getting
  314. 12:00better I can speak for myself this was
  315. 12:02part of my career for sure this kind of
  316. 12:04bond of trust and respect for the
  317. 12:06novice. It provides them with motivation
  318. 12:08to do better because you it's just
  319. 12:10important to you intrinsically. This is
  320. 12:12how humans are wired. We want to earn
  321. 12:14the trust and respect of mentors, senior
  322. 12:16people. And it goes for the other
  323. 12:17direction as well. Senior people, it's
  324. 12:20meaningful to [music] realize that you
  325. 12:22can help a junior person develop and
  326. 12:24give them challenges to help them do
  327. 12:26this. You want to earn their trust and
  328. 12:28respect in response. But for the junior
  329. 12:30person, it also means that that senior
  330. 12:31person, if you do this, will give you
  331. 12:33your next opportunity. They'll help you.
  332. 12:36So it's a functional thing as well. It's
  333. 12:38not just relational.
  334. 12:40Challenge, complexity, and connection
  335. 12:42were consistently evident across all of
  336. 12:44the shadow learning that I looked at.
  337. 12:46All these people who were breaking and
  338. 12:48bending rules, they were finding ways to
  339. 12:50protect challenge, complexity, and
  340. 12:52connection. Unconventional ways,
  341. 12:54inappropriate ways, but they were
  342. 12:56grabbing for those three things.

  343. What Leaders Must Do Now

  344. 13:02What's the leader's role in protecting
  345. 13:06healthy good quality target selection?

  346. Learn in Public

  347. 13:10Individual leadership action [music] is
  348. 13:13and has always been important. It will
  349. 13:15remain important. You have to lead by
  350. 13:17example and get primary data, personal
  351. 13:20experience with how people are doing the
  352. 13:22job in the organization. [music]
  353. 13:23most effective leaders now that I see
  354. 13:26spend personal significant time with the
  355. 13:29most advanced AI building things and
  356. 13:32showing how quickly [music] they can do
  357. 13:34that in a way that's valuable to the
  358. 13:36organization and including displaying
  359. 13:38failure and waste when they use
  360. 13:40reporting out to the organization as the
  361. 13:42leader or a leader of a function and
  362. 13:44saying you know I tried to do this thing
  363. 13:45with AI and it was very bad I decided
  364. 13:48it's a bad idea but still revealing this
  365. 13:51to people shows no one knows
  366. 13:53how to use this technology. Now, so
  367. 13:55modeling this is is very important. The

  368. Hire Juniors, Learn Both Ways - Inverted Apprenticeship

  369. 13:58general trend uh that seems to have
  370. 14:01shown up in the economy in the last year
  371. 14:02or year and a half is that firms have
  372. 14:05slowed down hiring at the junior level
  373. 14:07and tried to retain and grow talent at
  374. 14:10the more senior expertise level. It is
  375. 14:13very shortsighted. You need to be much
  376. 14:16more assertive about hiring junior
  377. 14:17people who are AI native and can do
  378. 14:20astounding things with AI even if
  379. 14:22they're professionally less experienced
  380. 14:24or have no experience and set up
  381. 14:27learning dynamics that are birectional
  382. 14:29so that the senior person can learn from
  383. 14:31the junior person and the junior person
  384. 14:33can learn from the senior person. I
  385. 14:35called this in my book and in my
  386. 14:36research inverted apprenticeship
  387. 14:38especially when new technologies arise.
  388. 14:40You [music] need a birectional learning.
  389. 14:43New employees have fresh perspective on
  390. 14:45the organization and can help improve
  391. 14:46things. So [music]
  392. 14:48healthy organizations are willing to
  393. 14:50take a shortrun uh hit to the
  394. 14:53productivity in order to ensure that
  395. 14:55they're ready for tomorrow.

  396. When AI Beats Us at Everything

  397. 14:58It's very rare in human history that we
  398. 15:00get a general purpose technology [music]
  399. 15:02that really changes the way we do
  400. 15:04things. It is going to be messy in the
  401. 15:05beginning. No one will get it right. We
  402. 15:08must all learn from each other. And so
  403. 15:10it's important to have have high
  404. 15:12standards, but also be forgiving with
  405. 15:14yourself and others and understand that
  406. 15:16everybody's going to be wasteful and
  407. 15:18make mistakes. I think it's important to
  408. 15:21be open to the possibility that AI in
  409. 15:24between 3 and [music] 40 years will be
  410. 15:28better at everything than humans are.
  411. 15:31Literally everything, all tasks,
  412. 15:33including empathy, including judgment,
  413. 15:36including creativity. This is not a
  414. 15:38crazy view to hold anymore. It is not
  415. 15:41obvious to me that our governmental,
  416. 15:43educational, institutional
  417. 15:46systems will adapt fast enough, even
  418. 15:49given a 30-year timeline, even given 50.
  419. 15:51And that means change will be more
  420. 15:53disruptive and difficult. We can either
  421. 15:56make AI part of the solution and [music]
  422. 15:58make for a better future and a better
  423. 16:00path to the future that is less painful,
  424. 16:02more beneficial for more people or
  425. 16:04change will come in more [music] painful
  426. 16:06ways. I want to live in the future where
  427. 16:09we are grateful that AI has arrived.
  428. 16:11That [music] we can look around and see
  429. 16:13that everybody is better off, not just a
  430. 16:15few, because of AI. And I think we must
  431. 16:19act immediately and together to create
  432. 16:22that kind of future.
How to Stay Great When AI Is Good Enough | Matt Beane, UC Santa BarBara — Transcriptly