YouTube2Text

How AI agents will change payments, commerce, and fraud - Amir Sarhangi (CEO, Skyfire) — Transcript

by Wrap Up · 5,371 words · 878 segments · language en · Watch on YouTube

Full transcript

  1. 0:00The projections are by 2030, a trillion
  2. 0:04dollars worth of [music] commerce will
  3. 0:06be driven by AI agents. First step is
  4. 0:09just to even understand how often agents
  5. 0:11are coming to their website,
  6. 0:13>> [music]
  7. 0:13>> right? Today, one of the challenges
  8. 0:15merchant merchants have is that they
  9. 0:17don't actually understand what that
  10. 0:19[music] is. If you look at the
  11. 0:20statistics, the amount of traffic on the
  12. 0:22internet has 10x [music]
  13. 0:24in the last couple of years, right? And
  14. 0:26over 50% of that is
  15. 0:28>> [music]
  16. 0:28>> essentially bot traffic. Those types of
  17. 0:30actions will create a lot of problems
  18. 0:33when it comes to fraud and when it comes
  19. 0:35to security.
  20. 0:35>> Okay, guys, machine economy. Well, it's
  21. 0:38complex, right? I I totally understand.
  22. 0:41Okay, let me simplify it. AI economy,
  23. 0:43all AIs making purchase on behalf of us.
  24. 0:47It is still um quite complex. I totally
  25. 0:49get it. That's why today I'm joined by
  26. 0:52Amir Sarhangi. I'm CEO of Skyfire, and I
  27. 0:57believe this is the right person to
  28. 0:58explain us what is um
  29. 1:01machine or AI economy is, how AI is
  30. 1:04going to change how we do purchases, how
  31. 1:06we do e-commerce, and how we can
  32. 1:09safeguard ourselves when AI agents start
  33. 1:12making the payments on behalf of behalf
  34. 1:14of us. Amir, welcome to the show.
  35. 1:17>> Sam, thanks for having me. Great to be
  36. 1:19here.
  37. 1:20>> Right, Amir.
  38. 1:21Just the first question straight. What
  39. 1:24is know your agent? If you could explain
  40. 1:26a couple of words.
  41. 1:27>> Yep. Know your agent uh is actually a
  42. 1:30term that we came up with a few years
  43. 1:31ago when we first started the business.
  44. 1:34And really is synonymous with uh when
  45. 1:37you think of KYC, know your customer.
  46. 1:39So, know your know your customer is
  47. 1:41about
  48. 1:42actually doing verified check, an ID
  49. 1:45check on on a human being when you open
  50. 1:47a bank account, for example.
  51. 1:49Uh similarly uh for KYA, which is know
  52. 1:53your agent, is understanding who really
  53. 1:55this agent is representing. So, it's a
  54. 1:58combination of verified information on
  55. 2:00the human
  56. 2:02that's behind the agent, the agent
  57. 2:04developer itself. That agent developer
  58. 2:06could be a company like uh
  59. 2:08ChatGPT or Anthropic. Uh or could be an
  60. 2:12individual developer um with like
  61. 2:14OpenClaw or something like that. Uh so,
  62. 2:16it's both of the those two information
  63. 2:19combined in a token we call KYAI token.
  64. 2:22Uh and that know your agent also has
  65. 2:24information about the intent of the
  66. 2:26user. So, for example, if Sam's trying
  67. 2:29to purchase uh tickets using an AI
  68. 2:32agent, uh has what what is Sam trying to
  69. 2:35do? And then, has Sam given permission
  70. 2:38to that agent to do that on his behalf?
  71. 2:41>> Right. And
  72. 2:42I mean, look, um I want to understand um
  73. 2:45when
  74. 2:47a company develops an AI agent, I
  75. 2:49understand like for example, OpenAI or
  76. 2:51some some other big companies. But when
  77. 2:53an individual developer, let's say, I
  78. 2:55sit at home, I develop my own agent, and
  79. 2:58it starts making um
  80. 3:00a purchase on my behalf, right? At
  81. 3:02Skyfire, how do you guys understand
  82. 3:06um that this is a legitimate bot? Uh you
  83. 3:09can trust it, you can trust the
  84. 3:10information it's sending you guys. What
  85. 3:13I'm trying to say that I'm trying to
  86. 3:14understand if there are certain rules, I
  87. 3:16mean, in the industry or you guys
  88. 3:18created perhaps that says, "Yes, you can
  89. 3:20trust this agent, you can trust this
  90. 3:22AI." Or no, you cannot try trust that.
  91. 3:25How do you do that at Skyfire?
  92. 3:27>> Yeah, great question, Sam. So, that's
  93. 3:29probably one of the biggest uh problems
  94. 3:31that exist today uh is that it's very
  95. 3:35difficult to understand who these agents
  96. 3:37are. If you look at the statistics, um
  97. 3:40the amount of traffic on the internet
  98. 3:42has, you know, 10x'd in the last couple
  99. 3:45of years, right? And a lot of that is
  100. 3:48over 50% of that is essentially uh bot
  101. 3:52traffic.
  102. 3:53>> Okay.
  103. 3:53>> Now, some of this bot traffic is
  104. 3:55legitimate, some of it is actually
  105. 3:57potential revenue
  106. 3:59for a lot of the websites that these
  107. 4:01agents are trying to go to, but a lot of
  108. 4:03it is also fraudulent as as you
  109. 4:05mentioned. Like a lot of times like you
  110. 4:08need to know who these bots are. So,
  111. 4:10really what we
  112. 4:11you know, what really we have been
  113. 4:13embracing is with KYA is this concept
  114. 4:16that
  115. 4:17bot developers need to go through a KYB
  116. 4:20process so that you actually know who
  117. 4:23this is. So, in in the case of an
  118. 4:25individual or an enterprise, they would
  119. 4:28go through a KYB which is know your
  120. 4:30business and that's proving that
  121. 4:33this is really who you say you are.
  122. 4:36And then additionally
  123. 4:37when it comes to the end user having
  124. 4:39verified information on who that end
  125. 4:42user is becomes pretty critical because
  126. 4:45if you're trying to deal with a merchant
  127. 4:47or a business, that business needs to
  128. 4:49know who you are
  129. 4:50and needs to be able to in order to be
  130. 4:53able to provide you the service that
  131. 4:54you're asking for.
  132. 4:56>> Yeah, true. And um
  133. 4:58before going deeper into this topic,
  134. 5:00could you briefly explain what you what
  135. 5:02you guys are doing at Skyfire? What
  136. 5:05problems you are solving?
  137. 5:06>> Yeah.
  138. 5:07So, at Skyfire really there are three
  139. 5:10areas that we're focused on. One is
  140. 5:13agent identity.
  141. 5:15Uh secondly is around access. So, how do
  142. 5:17agents access um
  143. 5:20basically human websites, right? Like
  144. 5:22just be able to go to directly to a to a
  145. 5:25website and be able to interact. And
  146. 5:27then thirdly, ability for to be able to
  147. 5:30check out. So, for example, if
  148. 5:33an agent goes to a website, discovers a
  149. 5:35product it likes, how can it actually
  150. 5:38purchase it like a human being does?
  151. 5:40>> Right.
  152. 5:41>> The challenge today is that most of
  153. 5:43these agents get blocked uh, different
  154. 5:45layers of web security. So, for example,
  155. 5:49um, most merchants have uh, bot blocking
  156. 5:51technology like Cloudflare, Akamai. Uh,
  157. 5:54today we in fact we announced with
  158. 5:56Akamai and Experian.
  159. 5:58Um, and then additionally you have fraud
  160. 5:59checks. Uh, this is folks like Forter
  161. 6:02who are at fraud
  162. 6:04basically inspecting to make sure this
  163. 6:06is a real human being. And then, you
  164. 6:09know, one of the other big things is
  165. 6:11that it when you're purchasing, uh, a
  166. 6:13lot of times you want to be able to use
  167. 6:15your loyalty. So, you know, let's say
  168. 6:17you go to Best Buy, you want to be able
  169. 6:18to make sure that uh, purchase that
  170. 6:20you're doing also counts for your
  171. 6:23loyalty. So, ability to log in becomes
  172. 6:26another big aspect and that's typically
  173. 6:28protected by folks like Okta.
  174. 6:31>> All right. And um, where do you sit in
  175. 6:33this flow? I'm trying to understand. Do
  176. 6:35you directly work with merchants or you
  177. 6:37work with payment service providers?
  178. 6:39Where do you sit? And um,
  179. 6:42should merchants go um, make the
  180. 6:44integrations? How does it work if you
  181. 6:45could explain as well?
  182. 6:47>> Yeah, great question. So, uh, what we do
  183. 6:49is we work directly with the demand
  184. 6:51side. So, this is like an agent
  185. 6:53developer. An agent developer uses our
  186. 6:55technology. In this case, uh, what we
  187. 6:58call it a KYA token. They create this
  188. 7:00KYA token. That token has the verified
  189. 7:03information on the human and the and the
  190. 7:06uh, agent itself. Uh, and then that's
  191. 7:09used in the header when an agent makes a
  192. 7:12request to a website. That's included as
  193. 7:14part of the header.
  194. 7:15Uh, we have gone out and done uh,
  195. 7:18basically partnerships with a host of
  196. 7:20web security companies.
  197. 7:22Uh, again, this is people like Akamai,
  198. 7:24um, DataDome,
  199. 7:26>> Right.
  200. 7:26>> Cloudflare, and they uh, these web
  201. 7:30security companies accept this token and
  202. 7:32then provide access to the websites.
  203. 7:35>> Right.
  204. 7:36>> of this model is that it actually scales
  205. 7:38across the whole internet without
  206. 7:41all of the different merchants and
  207. 7:42websites having to support
  208. 7:44>> Mhm. anything new. So, there's no new
  209. 7:47requirements other than their web
  210. 7:49security company accepting this token.
  211. 7:52>> But But, you just mentioned that, for
  212. 7:54example, yeah, this is also not
  213. 7:56something new. For example, let's say
  214. 7:58um agent developer needs to go and do
  215. 8:02KYB.
  216. 8:03Oh, yeah. Agent developer. Now I Now I
  217. 8:06get it. Yeah. Yes.
  218. 8:07>> [snorts]
  219. 8:08>> And um
  220. 8:09for agent developer to do the KYB, this
  221. 8:11was my question I was thinking about.
  222. 8:13Let's say I'm individual. I don't have a
  223. 8:14business. I'm just a developer who sits
  224. 8:16at my home and just develop it. How
  225. 8:19would I do a KYB? What would you require
  226. 8:21from me?
  227. 8:21>> that's a That's a good question, Sam.
  228. 8:23So, that basically turns into a KYC. So,
  229. 8:26in the case where there is no there is
  230. 8:29no entity,
  231. 8:31uh it becomes essentially a KYC,
  232. 8:33um and that becomes an individual
  233. 8:35developer who basically is accessing uh
  234. 8:38something on his on on his his or her
  235. 8:40behalf, right?
  236. 8:42>> Okay.
  237. 8:43>> So, yes. But, there's still a
  238. 8:44verification process.
  239. 8:46>> And you still have to go through it.
  240. 8:47>> Yes.
  241. 8:48Exactly.
  242. 8:48>> Yeah.
  243. 8:49And if you talk about
  244. 8:51machine economy, where do you see at the
  245. 8:54moment AI agents are being used the
  246. 8:56most? Is it like e-commerce or
  247. 8:58subscription economy? Where is it
  248. 9:00happening the most at the moment?
  249. 9:02>> Yeah. So, look. I mean,
  250. 9:05purchasing with bots has something
  251. 9:06that's been happening for many years
  252. 9:08already.
  253. 9:08>> Okay.
  254. 9:10>> So, for example, if you if you look at,
  255. 9:12you know, the reason why when you want
  256. 9:15to go to a really popular concert and
  257. 9:17within seconds uh most of it is sold out
  258. 9:20even though you're refreshing your web
  259. 9:22page, it's not just because there's a
  260. 9:24lot of fans trying to buy. It's also
  261. 9:26there's bots that are coming in and
  262. 9:28trying to purchase on behalf of
  263. 9:30resellers. Same thing is true with Nike
  264. 9:32drops and many other places in
  265. 9:35e-commerce. So,
  266. 9:37bots have been purchasing, you know, I
  267. 9:40wouldn't say legally, but
  268. 9:42you know, in a gray zone manner.
  269. 9:45So, that's something that's already been
  270. 9:46happening it and you have companies like
  271. 9:48now that are much more, you know, front
  272. 9:51and center about it like Rye is one of
  273. 9:53our partners, Card AI. There's a bunch
  274. 9:56of players that are already now
  275. 9:58purchasing
  276. 10:00either on behalf of businesses or
  277. 10:02consumers. That's already happening. On
  278. 10:05the machine economy side of things when
  279. 10:06it comes to micro payments, I'd say it's
  280. 10:08still early days, but we're starting to
  281. 10:10see
  282. 10:12basically ability for agents on on their
  283. 10:15own to be able to pay for content and
  284. 10:18also be able to pay for APIs or access
  285. 10:21to
  286. 10:22whatever it is that they're trying to
  287. 10:23get to. But, I'd say you know, from our
  288. 10:26perspective, we see that it's still at
  289. 10:27its early days. Part of the challenge
  290. 10:29there is that
  291. 10:31a lot of the companies on the supply
  292. 10:33side of this equation continue to want
  293. 10:35to be able to sell SaaS, right? It's
  294. 10:37it's much better to sell sell a SaaS
  295. 10:39package than a on-demand per
  296. 10:43request basis. But, that is going to
  297. 10:45change.
  298. 10:46And the reason that will change is
  299. 10:48because you have many companies who will
  300. 10:51not want to pay the big SaaS dollars,
  301. 10:55especially if they're not using the full
  302. 10:56package. They'll go and try to, you
  303. 10:59know, use upstarts or even create their
  304. 11:01own, you know, stack. So,
  305. 11:04for those reasons, those pressures are
  306. 11:06going to force
  307. 11:08the those SaaS bundles to
  308. 11:10be unbundled and be offered on a more
  309. 11:13ala carte basis.
  310. 11:14>> You just you just mentioned micro
  311. 11:17payments and I often hear that it is
  312. 11:19one of the key parts of machine economy.
  313. 11:21Why are micro payments so important in
  314. 11:24this machine economy? What the role they
  315. 11:25play?
  316. 11:26>> Yeah, I mean,
  317. 11:28basically if you want to now, let's say,
  318. 11:30let's say you use HubSpot or you or
  319. 11:33agent wants to have access to some
  320. 11:36particular content
  321. 11:38for learning or for just needs access to
  322. 11:41it and wants to be able to pay for that
  323. 11:42content on a per article basis or wants
  324. 11:46to have access to an API
  325. 11:49or
  326. 11:50for whatever it is it needs.
  327. 11:51>> Yeah.
  328. 11:52>> That API call is a single call and a
  329. 11:54single access and it may come back a
  330. 11:56hundred more times and use it again, but
  331. 11:59each time it's going to pay a cent or a
  332. 12:01fraction of a cent for that access. And
  333. 12:04that's where that micro payment becomes
  334. 12:06important because
  335. 12:08in in this scenario the agent is not
  336. 12:10signing up to some big plan. It's saying
  337. 12:12I'm going to pay you every time I use
  338. 12:14your product, which means that the the
  339. 12:17provider has to be able to sell its
  340. 12:19product on a essentially on a single hit
  341. 12:22basis.
  342. 12:24>> Right. Right. And then as you know,
  343. 12:26credit rails are not best for micro
  344. 12:28payments. Those those are very
  345. 12:30expensive. Credit rails especially they
  346. 12:32are very expensive. That's why
  347. 12:34everyone's trying to avoid that. Do you
  348. 12:36think that's why
  349. 12:37stable coins would become like let's say
  350. 12:39default currency for machine economy or
  351. 12:42do you think
  352. 12:44the current rails adopting and they will
  353. 12:46still exist in one or another way?
  354. 12:50>> I think look, I think
  355. 12:52you're going to see both, but I this is
  356. 12:54where the sweet spot for stable coins is
  357. 12:57because it's not just the micro payment
  358. 12:59aspect of it. It's the fact that you're
  359. 13:02actually transferring value at the time
  360. 13:04of the service. So imagine a scenario
  361. 13:06where you hire another agent that then
  362. 13:10hires another agent to do something. In
  363. 13:12that chain, you know, you have an agent
  364. 13:15that's getting a commission or getting
  365. 13:16part of a transaction. Now imagine
  366. 13:19there's fraud or you know, this can't be
  367. 13:22done in credit because like trying to
  368. 13:24roll that back or figure out who caused
  369. 13:27what fraud becomes a big hassle, right?
  370. 13:30Whereas here with stablecoins, once that
  371. 13:33money is transferred, like
  372. 13:36you know, that that next agent owns
  373. 13:39that, you know,
  374. 13:41um
  375. 13:41that currency, which is that becomes
  376. 13:43really critical. That uh that value
  377. 13:46exchange uh at the time of service is is
  378. 13:49pretty critical when it comes to kind of
  379. 13:51this new agentic world that we're going
  380. 13:53to see.
  381. 13:54>> I see. And do you guys do transactions
  382. 13:56actually? Or you do not touch money at
  383. 13:59all? You are only working on, as you
  384. 14:01said, as a know-your-agent and basically
  385. 14:04assisting the transaction.
  386. 14:06>> Yeah, but we provide a wallet. In fact,
  387. 14:07we started with stablecoins.
  388. 14:09>> All right.
  389. 14:10>> we have with our partnerships with Visa
  390. 14:12and Mastercard, we've added uh both uh
  391. 14:16Visa tokenized cards and Mastercard, and
  392. 14:19we'll we'll be also uh adding others. Um
  393. 14:23so,
  394. 14:24our wallet basically will provide the
  395. 14:26agent all the different mechanisms uh
  396. 14:29for payment, essentially.
  397. 14:30>> Okay.
  398. 14:31Okay. And uh
  399. 14:33why should I use your wallet, for
  400. 14:34example? Why cannot I go um
  401. 14:38use only your KYA, let's say,
  402. 14:40know-your-agent, and continue using
  403. 14:42Stripe or any other service provider
  404. 14:44that I'm using at the moment?
  405. 14:46>> Yeah, I mean, an user really can. The
  406. 14:47big advantage of using us, uh other than
  407. 14:50being obviously a one-stop shop for you,
  408. 14:52is the fact that when an agent is trying
  409. 14:56to go to a website or to uh an API, one
  410. 15:01of the important things that that end
  411. 15:03destination wants to know is that
  412. 15:06there's real intent for purchasing.
  413. 15:08Meaning, we can say that like san- the
  414. 15:11same agent
  415. 15:12has a credit card on file and is a real
  416. 15:17buyer, right? And then at the time that
  417. 15:20you do try to make a purchase, um
  418. 15:23understanding not only that this is not
  419. 15:25a fraudulent card that you're using.
  420. 15:27That's something that we're we're
  421. 15:29working with Experian to unlock is this
  422. 15:31kind of real-time risk
  423. 15:33aspect.
  424. 15:35And then secondly is is the fact that
  425. 15:39we capture the intent of the user in
  426. 15:41that process, meaning that you've given
  427. 15:44authorization
  428. 15:46to the agent to make that purchase on
  429. 15:48your behalf. All that together becomes
  430. 15:50really critical for merchants who are
  431. 15:54trying to provide, you know, Sam AI, for
  432. 15:57example, AI agent,
  433. 16:00a service or a product.
  434. 16:02>> Does it mean that Well, basically, if I
  435. 16:04understand correctly, if I take the
  436. 16:06bundle from you, meaning like your
  437. 16:08wallet and other services, it will be
  438. 16:10more secure than combining it with,
  439. 16:13let's say, other providers that I'm
  440. 16:15already using. Do I understand it
  441. 16:16correctly?
  442. 16:17>> Not only more secure, but also you get
  443. 16:20ability to go to probably more
  444. 16:25destinations and websites because now we
  445. 16:27have a complete
  446. 16:29understanding of who this agent and the
  447. 16:33user behind it is and the intent of that
  448. 16:35user. So, it's really becomes a
  449. 16:37combination of access and security in
  450. 16:39that scenario.
  451. 16:40>> And
  452. 16:41what are some of the
  453. 16:43industries that are adopting AI agents
  454. 16:45in the first place?
  455. 16:47And here's an interesting question, how
  456. 16:49do you guys make money out of this?
  457. 16:51>> [laughter]
  458. 16:52>> Well, we we make money in a few
  459. 16:55different areas, but really
  460. 16:58initially really it's about adoption.
  461. 17:00So,
  462. 17:02where we make money is KY A tokens being
  463. 17:04created,
  464. 17:06and that's something that, you know, we
  465. 17:08work directly with agent developers
  466. 17:11with,
  467. 17:12but also in the future, you know, we'll
  468. 17:14hopefully also be able to uh provide
  469. 17:18more uh risk signals to the merchants.
  470. 17:21That will be another area. Um in terms
  471. 17:23of, you know, we're seeing quite a few
  472. 17:26different areas in terms of um
  473. 17:28adoptions. Uh we're working uh with uh
  474. 17:32not only kind of AI-first vertical um
  475. 17:37developers who are essentially focused
  476. 17:39on things like, whether it's fashion,
  477. 17:41shopping, uh travel, you're going to see
  478. 17:44a lot more personalized AI assistant
  479. 17:47type experiences
  480. 17:49um that are going to be um essentially
  481. 17:52uh providing a really high-end
  482. 17:54experience to uh to users.
  483. 17:57Uh and then the other area that we're uh
  484. 18:00getting some great traction around is
  485. 18:01financial institutions.
  486. 18:03Um this is card issuers, this is um
  487. 18:06folks who are essentially uh providing
  488. 18:10uh you today a wallet. Uh
  489. 18:13and and that wallet could be either
  490. 18:15through uh
  491. 18:16your your card or your bank. Uh so
  492. 18:19that's another area that we're starting
  493. 18:21to see
  494. 18:22uh quite a bit of interest.
  495. 18:23>> I mean, what do you think? Which
  496. 18:25industries are adopting AI agents the
  497. 18:27first? Where do you see the real
  498. 18:29adoption curve basically going up?
  499. 18:32>> Yeah, so in the consumer space, uh I
  500. 18:34mean, there's consumer and B2B. On the
  501. 18:36consumer side, what we're seeing is that
  502. 18:38you have uh AI-first companies that are
  503. 18:42being created now. Those companies are
  504. 18:44starting really with a completely new
  505. 18:47user experience. It's more of a, you
  506. 18:49know, kind of a chat commerce
  507. 18:50experience.
  508. 18:52That's uh you know, we're starting to
  509. 18:54see that in fashion, we're seeing that
  510. 18:56in travel, um and you know, many kind of
  511. 18:59different categories in the consumer
  512. 19:01space.
  513. 19:02You're also starting to see
  514. 19:04existing companies uh who previously are
  515. 19:08like a marketplace. Uh this is where you
  516. 19:10go to discover new things like think
  517. 19:12Pinterest or others like that who are
  518. 19:15also wanting to take the friction out of
  519. 19:18the buying experience. So, anywhere
  520. 19:20where you have a referral link where
  521. 19:23that referral link takes you to another
  522. 19:25site, that can be now more of a buy for
  523. 19:27me button. It's kind of like that
  524. 19:28instant purchasing experience where you
  525. 19:31now don't have to go to these
  526. 19:34steps and websites to to purchase. So,
  527. 19:36that's more on the consumer side. Um
  528. 19:38>> Okay.
  529. 19:41>> companies, you know, whether it's your
  530. 19:43card network, anyone anywhere where you
  531. 19:46have a wallet like think Chase Preferred
  532. 19:48or
  533. 19:50you know, Capital One.
  534. 19:52Those types of experiences will evolve
  535. 19:54in the future and you know, you will now
  536. 19:57just be able to say, "Hey, this is what
  537. 19:59I'm looking for." It'll find you the
  538. 20:01best price, apply the coupons, and be
  539. 20:04able to make that
  540. 20:05a much more of a concierge
  541. 20:07uh kind of a VIP experience for you.
  542. 20:11>> Do you do do you think
  543. 20:13I mean, we are in agent or AI economy or
  544. 20:17it's very very early stage. Why I'm
  545. 20:18asking this question is that you must
  546. 20:20have read that OpenAI and Stripe they
  547. 20:23launched a partnership so that they
  548. 20:25enable purchasing directly inside the
  549. 20:27chat. But but then they scaled back this
  550. 20:31partnership.
  551. 20:33Um there was a news. So, why do you
  552. 20:36think that happened in your opinion?
  553. 20:38>> Um
  554. 20:40There there there are many different
  555. 20:41reasons
  556. 20:42for that. You know, I obviously I'm not
  557. 20:45inside OpenAI. I cannot give you
  558. 20:47>> Sure.
  559. 20:47>> anything but from a outsider's
  560. 20:49perspective, you know, I think
  561. 20:51um the experience was not there yet, you
  562. 20:54know, from a from a consumer side. Uh
  563. 20:57and really, you know, if you think about
  564. 20:58it from a consumer perspective,
  565. 21:00when you
  566. 21:02want to use a
  567. 21:03AI agent,
  568. 21:04part of the reason you do this is
  569. 21:06because you wanted you expect that agent
  570. 21:08to do a better job than you
  571. 21:10Okay.
  572. 21:18Okay.
  573. 21:35Mhm.
  574. 21:35>> [clears throat]
  575. 21:40>> Yeah.
  576. 21:43Yeah.
  577. 21:50And I think that's where you know, at
  578. 21:52the end of the day, you know, any of
  579. 21:54these experiences for them to work
  580. 21:56really well, they need to be
  581. 22:00scaled across the internet. And then
  582. 22:02more importantly, there are other
  583. 22:04implementation aspects that have to be
  584. 22:06considered around fraud
  585. 22:08and risk that, you know,
  586. 22:12become really important in any
  587. 22:14implementation. So,
  588. 22:16you know,
  589. 22:17your question is a is a really good one,
  590. 22:20but I think a big part of the reason
  591. 22:24there was a pause in that particular
  592. 22:27instance is primarily because I think
  593. 22:29they rushed into it very quickly to to
  594. 22:31launch something.
  595. 22:32>> So, you you think we are very early
  596. 22:35stage and there is no trust because
  597. 22:37again, I have another example is
  598. 22:39Walmart. They also rolled out basically
  599. 22:42this similar types of solutions, but
  600. 22:45they rolled back and the news said that
  601. 22:47basically customers were not converting
  602. 22:50from chat to their website, you know, at
  603. 22:53all. But it was like the conversion
  604. 22:54rates were very very low. So, they said
  605. 22:58like it
  606. 22:59we are all in this back. Do you think
  607. 23:00this is like trust issue? People do not
  608. 23:03trust AI agents or there is something
  609. 23:05else?
  610. 23:06>> I think it is Trust definitely is a
  611. 23:08factor in terms of adoption of any new
  612. 23:11technology. Even if you look at the
  613. 23:13early days of uh
  614. 23:15you know, commerce on the web, people
  615. 23:18would not trust using their credit card
  616. 23:21on a website. That was not too long ago.
  617. 23:24That That was just, you know, three
  618. 23:25decades ago when, you know, Amazon was
  619. 23:28trying to push people to use their card,
  620. 23:30right? And
  621. 23:32that evolution is one where you have to
  622. 23:35ease the user into that experience. Uh
  623. 23:38you know,
  624. 23:39the way we see it is like three
  625. 23:41different steps. The first step is this
  626. 23:43experience where you see something, you
  627. 23:46like it, and you say just buy it for me,
  628. 23:49right? So, that's a place where you've
  629. 23:50already discovered what you want, and
  630. 23:53you make that purchase. So, that's
  631. 23:54that's the first way of easing the user
  632. 23:56into the experience. The second step is
  633. 23:59more of a commerce chat experience where
  634. 24:01you don't know what you want yet, um and
  635. 24:04you're trying to figure out
  636. 24:06uh what's the best option for you. So,
  637. 24:08you start chatting
  638. 24:09um kind of like what you do search, but
  639. 24:12here it's a much better experience
  640. 24:14because you can actually
  641. 24:15um do it in natural language,
  642. 24:18be able to quickly find what you want um
  643. 24:21just like you do today in Open AI. I
  644. 24:23mean, the statistics are kind of
  645. 24:25astounding how much
  646. 24:27uh now you know, uh ref- referrals are
  647. 24:30coming from
  648. 24:31uh AI companies. The numbers are pretty
  649. 24:34astounding, and the projections are by
  650. 24:382030, a trillion dollars worth of
  651. 24:40commerce will be driven by AI agents.
  652. 24:43And And those numbers will happen very
  653. 24:46quickly. That's just 3 years from now.
  654. 24:48And the reason is that you're going to
  655. 24:49have this progression where you start
  656. 24:51from this kind of buy for me experience
  657. 24:52to
  658. 24:54uh agentic kind of
  659. 24:56um
  660. 24:56chat experience, and that's still in the
  661. 24:59loop. You're You're still approving
  662. 25:01everything. So, when the agent goes and
  663. 25:03finds you what it wants, it'll ask Sam,
  664. 25:05"Hey Sam,
  665. 25:06do I have approval to buy this pair of
  666. 25:08Nike shoes for $110?" You approve it
  667. 25:11with your passkey and then goes and
  668. 25:13purchases for you. So, as you can see,
  669. 25:15it's just slowly building trust and then
  670. 25:18kind of the third phase of this is more
  671. 25:20autonomous where
  672. 25:22you want to buy a pair
  673. 25:24you know, you want to buy some concert
  674. 25:26tickets. You don't want to
  675. 25:28stay awake till midnight when they go
  676. 25:29live. You tell your agent, "Hey,
  677. 25:33if if you can find me these tickets
  678. 25:35under $200, just buy it. I approve it
  679. 25:38like beforehand with you.
  680. 25:41>> [clears throat]
  681. 25:41>> That becomes kind of the more autonomous
  682. 25:43experience and that's what we're
  683. 25:46starting to see and yes, you are going
  684. 25:48to have implementations that are not
  685. 25:51going to be the best ideal user
  686. 25:52experience and they will need to be
  687. 25:55rethought, but there's absolutely no
  688. 25:57question in my in my mind now that
  689. 26:00agentic commerce is going to happen
  690. 26:02because we're starting to see it already
  691. 26:04working
  692. 26:05and more importantly, it
  693. 26:07just gives
  694. 26:08you know, in in some cases, people want
  695. 26:10to shop. In a lot of cases, people don't
  696. 26:12want to shop and they much rather take
  697. 26:15that friction out and that's where
  698. 26:17agentic commerce will help
  699. 26:20free up time for people, make it much
  700. 26:22easier for them to do their daily tasks.
  701. 26:26>> I mean, I understand for end consumer,
  702. 26:28okay.
  703. 26:29But, I I will not go that far to the
  704. 26:32future. You are saying to certain sense,
  705. 26:34we are already in machine economy.
  706. 26:36Machines are already making purchases.
  707. 26:38Then, what do you think merchants should
  708. 26:40understand about machine economy? How
  709. 26:43they should be prepared for that? What
  710. 26:46should they think about it and consider?
  711. 26:49>> First step is just to even understand
  712. 26:51how often agents are coming to their
  713. 26:54website, right? Today, one of the
  714. 26:56challenges merchant merchants have is
  715. 26:58that they don't actually understand what
  716. 27:01that is. And and the reason I say this
  717. 27:03is that when you look at you go because
  718. 27:06obviously merchants are driven by data,
  719. 27:07right? Like data is king for them. And
  720. 27:10one of the things they care a lot about
  721. 27:12is understanding how often humans are
  722. 27:14coming to their website, what are they
  723. 27:16searching for, how often they leave a
  724. 27:18basket uh with you know, goods in it. Um
  725. 27:22and now they're starting to see a big
  726. 27:25part of that pie chart of humans being
  727. 27:28gray. That gray is basically a
  728. 27:31combination of fraudulent agents and
  729. 27:33good agents that are coming to their
  730. 27:35website to try to understand their
  731. 27:37product or try to purchase something.
  732. 27:40>> Right.
  733. 27:40>> They need to understand that that gray
  734. 27:42slice much better than they do today.
  735. 27:44And that's that's the first thing they
  736. 27:46need to be able to do is know who what
  737. 27:49type of agents are coming to them on
  738. 27:51behalf of what type of humans.
  739. 27:53And then being able to essentially
  740. 27:55unlock
  741. 27:56agents being able to purchase directly
  742. 27:58from the human website. No APIs, we're
  743. 28:01just talking about the regular old
  744. 28:02website because agents can do that now
  745. 28:05very effectively.
  746. 28:06>> Okay.
  747. 28:07>> And being able to unlock that, which
  748. 28:09then gives them much better
  749. 28:10understanding of uh what agents are
  750. 28:15driving revenue for them
  751. 28:17>> Yeah.
  752. 28:17>> and being able to kind of foster that
  753. 28:19and um and then being able to build
  754. 28:22essentially a a strategy around how they
  755. 28:25support agents going forward.
  756. 28:28>> What about financial institutions? We
  757. 28:31see you explained what should merchants
  758. 28:33do. What about financial institutions?
  759. 28:35Because they also need to be prepared,
  760. 28:37right?
  761. 28:38>> Yeah, so on the on the financial
  762. 28:39institution side, I think there's a
  763. 28:40couple of things. One is
  764. 28:42um you know,
  765. 28:44you if you're a card issuer, um if today
  766. 28:48you are offering as a financial
  767. 28:50institution you are offering an
  768. 28:51experience to end use end consumers,
  769. 28:54the thing what that's really
  770. 28:57like this AI
  771. 28:59you know, wave does for you is that it
  772. 29:01kind of evens the playing field. It
  773. 29:03allows you to provide a much better
  774. 29:05experience to your users
  775. 29:07that is very differentiated in the sense
  776. 29:09that now you can provide
  777. 29:11you know, let's say
  778. 29:13you're a card issuer you have a wallet.
  779. 29:15Now you can provide kind of this like
  780. 29:17really
  781. 29:19great concierge experience for your
  782. 29:21users to be able to use your card on the
  783. 29:24purchases that they want to do. So, it's
  784. 29:27you know, think about it as kind of
  785. 29:28Amazon Prime where like once you're
  786. 29:31locked in and starting to use Amazon
  787. 29:34Prime for one thing you'll start adding
  788. 29:36more and then it's just on automatic.
  789. 29:38It's the same story here. Once you start
  790. 29:41using an AI agent to do purchasing and
  791. 29:43you and that your card issuer is
  792. 29:46basically that's the default card you're
  793. 29:48going to be using
  794. 29:49then you are starting to really get your
  795. 29:53consumers to use your card for a lot of
  796. 29:56their purchases that are happening with
  797. 29:57it with agents. So, that's one part of
  798. 30:00the static strategy is really leaning
  799. 30:02into
  800. 30:04this new trend that will happen
  801. 30:08and and the first step is just you know,
  802. 30:10kind of that buy for me experience and
  803. 30:13then
  804. 30:14from there more of a concierge
  805. 30:15experience.
  806. 30:16The second area where financial
  807. 30:18institutions are starting to need and
  808. 30:21and needing to evolve very quickly is
  809. 30:23that their customers want to be able to
  810. 30:26access their accounts using agents. So,
  811. 30:29for example,
  812. 30:30you know, being able to
  813. 30:33access your bank account or access your
  814. 30:35financials even in a read-only fashion
  815. 30:39using Claude or ChatGPT or Gemini so
  816. 30:42that you can do analysis on your account
  817. 30:45for various right? So
  818. 30:48and similarly enterprises who are
  819. 30:50automating they want access to their own
  820. 30:53accounts using agents. So this becomes
  821. 30:55another big area for financial
  822. 30:57institutions. This is not so much
  823. 30:59agentic commerce but it's more like
  824. 31:01agentic access. So how does it
  825. 31:04How do I access my own account using
  826. 31:06agents essentially? That's the other
  827. 31:08area for financial institutions.
  828. 31:10>> I mean this is my last question. What do
  829. 31:13you think the biggest risk in the near
  830. 31:15term in machine economy? What should
  831. 31:18both merchants and financial
  832. 31:19institutions
  833. 31:21be ready?
  834. 31:23>> Yeah look you know I think in some ways
  835. 31:26if done right we can actually reduce
  836. 31:29fraud believe it or not
  837. 31:32in agent with agentic commerce because
  838. 31:35you have a much more secure way of
  839. 31:38onboarding the user and the card and
  840. 31:42then
  841. 31:43the
  842. 31:43essentially the way
  843. 31:45um
  844. 31:47everything moves from there is can be
  845. 31:50done in a very secure way. I think the
  846. 31:53challenge here and the risk is going to
  847. 31:55be if
  848. 31:56um we don't do that and we you know we
  849. 31:59have things like
  850. 32:01you know people using their you know raw
  851. 32:03pan their credit card number directly
  852. 32:06with agents or
  853. 32:09>> [clears throat]
  854. 32:09>> users trying to use an agent with their
  855. 32:11own username password um trying to
  856. 32:14access their bank and doing like two off
  857. 32:17with their own phone. That those types
  858. 32:20of actions will um
  859. 32:22will create a lot of problems when it
  860. 32:25comes to fraud and when it comes to
  861. 32:26security.
  862. 32:28So it it's more about actually
  863. 32:32you know the industry moving quickly
  864. 32:33enough to be able to provide the right
  865. 32:37tools
  866. 32:38for both agent developers
  867. 32:41and also enterprises to be able
  868. 32:44to do this much more securely.
  869. 32:47>> Right. I mean, I really enjoyed that
  870. 32:49conversation. Thank you for coming on
  871. 32:50the show and sharing these insights. I
  872. 32:52hope it was useful for my audience as
  873. 32:54well. For everyone listening to this
  874. 32:55episode, don't forget to share it with
  875. 32:57your friends and everyone interested in
  876. 32:59this and don't forget to subscribe.
  877. 33:01Thank you very much.
  878. 33:03>> Thanks, Sam.

About this transcript

This page contains the full transcript of How AI agents will change payments, commerce, and fraud - Amir Sarhangi (CEO, Skyfire) by Wrap Up, generated from the public captions YouTube serves with the video. The transcript has 5,371 words across 878 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.