[ { "start": 16.54, "end": 22.24, "text": "Welcome to a new episode of BFRR, our Bitcoin fiat and rock'n roll podcast that explores the intersection." }, { "start": 42.58, "end": 48.239, "text": "All the listeners, we all know that artificial intelligence or AI is currently transforming various industries." }, { "start": 48.42, "end": 49.86, "text": "And this may on May not?" }, { "start": 50.1, "end": 52.059, "text": "We will see also hold for the payment space." }, { "start": 52.239, "end": 55.82, "text": "as we know it there was a lot of fantasy around agentic commerce." }, { "start": 55.9, "end": 60.66, "text": "so basically commerce experiences were not humans shop but agents On our behalf." }, { "start": 61.26, "end": 67.66, "text": "There's also lots of fantasy and noise about agentic payments where basically agents would pay on our behalf in the future." }, { "start": 68.32, "end": 71.5, "text": "But as for every innovation, there is a lot of substance to the topic." }, { "start": 71.58, "end": 75.38, "text": "A lot of potential interesting visions how this changes the world etc." }, { "start": 75.78, "end": 79.82, "text": "but obviously There's also a lot hype and BS around that." }, { "start": 80.32, "end": 89.76, "text": "So what we now want do in this podcast Is regularly hosting episodes with industry leaders on The topic of AI In particular agented payments." }, { "start": 89.88, "end": 94.399, "text": "so As you are used it from the podcast no hype No BS just facts." }, { "start": 95.179, "end": 99.8, "text": "That's basically what we want to offer you and do a lot of deep dive sessions into agentic payments." }, { "start": 99.96, "end": 104.24, "text": "So when I explore what the genetic payments promise, What they can really deliver?" }, { "start": 110.76, "end": 111.36, "text": "as we know it." }, { "start": 111.96, "end": 126.46, "text": "And today is the first of many episodes now around agentic payments where we will discuss basically fundamentals or basics and for this I'm super excited to have a guest in our podcast before, Sonya Davidovic back on our podcast." }, { "start": 126.58, "end": 128.0, "text": "so welcome back Sonya!" }, { "start": 128.14, "end": 129.14, "text": "It's great" }, { "start": 137.78, "end": 139.96, "text": "like your podcast, because it's a bit different." }, { "start": 140.1, "end": 143.3, "text": "It is more casual and interactive." }, { "start": 143.36, "end": 146.72, "text": "so its always a pleasure to interact with you guys." }, { "start": 147.08, "end": 147.88, "text": "Thank You for having me!" }, { "start": 148.5, "end": 149.34, "text": "That was the goal." }, { "start": 149.4, "end": 152.18, "text": "I'm really glad that you accepted the invite and Sonya is pretty crazy." }, { "start": 152.26, "end": 158.66, "text": "The last time we were here it was March twenty one if i am not mistaken." }, { "start": 158.76, "end": 163.76, "text": "So episode seventy three Now have more than four hundred." }, { "start": 164.48, "end": 165.9, "text": "Yeah, this was really quite a while back." }, { "start": 165.94, "end": 182.72, "text": "It wasn't German though I remember and i was uh struggling a little bit because all the expressions that we normally use in the industry are a bit harder to say in german just cause we were so used to um using english in our daily lives." }, { "start": 182.96, "end": 184.26, "text": "but it was fun." }, { "start": 184.32, "end": 187.44, "text": "yeah it really was and i think we talked particularly about cbdc's and michael." }, { "start": 187.48, "end": 193.3, "text": "they used to remember the german good old german podcast days or now almost like more than one-and-a-half years in English." }, { "start": 193.36, "end": 193.56, "text": "right" }, { "start": 194.38, "end": 195.5, "text": "Definitely easier, yeah?" }, { "start": 196.54, "end": 199.22, "text": "For sure, and I mean for everybody who doesn't know you Sonja." }, { "start": 199.46, "end": 202.32, "text": "You really suggest listening to the episode we recorded." }, { "start": 202.54, "end": 206.36, "text": "Episode seventy three where do an introduction about you as a person?" }, { "start": 206.42, "end": 208.52, "text": "what have been working on back that?" }, { "start": 208.7, "end": 211.82, "text": "so We will kind of skip this part or make us a little bit shorter today." }, { "start": 211.92, "end": 217.48, "text": "So in our nutshell Sonja from my perspective seems to always be at the forefront of innovation." }, { "start": 217.82, "end": 232.96, "text": "if you will also Worked at The Bank for International Settlements Also in From the Innovation Hub In Singapore was before at the international International Monetary Fund, also at the European Central Bank working on the digital euro and now back at the international monetary fund IMF." }, { "start": 233.52, "end": 235.08, "text": "Also looking into a genetic payment." }, { "start": 235.16, "end": 242.0, "text": "so what I would be very curious on you as I researched a lot of topics like crypto, CBDC very early." }, { "start": 242.28, "end": 244.56, "text": "Remember some early papers together with John Kiff?" }, { "start": 245.06, "end": 246.68, "text": "You looked into decentralized finance." }, { "start": 246.72, "end": 251.52, "text": "so as said always at the forefront of innovation but now sparked your interest about AI." }, { "start": 251.64, "end": 260.12, "text": "So why did you decide to go in to the agentic payments and not look more into DeFi or CBDC But more on the agentics side?" }, { "start": 260.76, "end": 265.3, "text": "Thanks for taking me back or taking on this trip down memory lane." }, { "start": 265.38, "end": 283.84, "text": "It's really fun, yeah I remember our first CVDC papers were when nobody really talked about CVDC but... Yeah i think a reoccurring theme in my career has always been exploring how technology changes the way it interacts with financial infrastructure." }, { "start": 284.38, "end": 289.44, "text": "we have to think about how will technology impact our membership?" }, { "start": 289.539, "end": 292.26, "text": "How it will impact the economy at large." }, { "start": 292.359, "end": 304.8, "text": "So, agentic AI is one of those technologies that are transformative in its nature and their way and the impact they'll have on payment systems or financial infrastructure and people in general." }, { "start": 304.94, "end": 315.8, "text": "so I kind of sparked interest because were thinking... We had an agentic Ai which has been around for a while now but then We have the whole world of payments and digital assets." }, { "start": 316.3, "end": 339.719, "text": "So we started seeing or observing how these two worlds are increasingly colliding, so in our mind it was clear that we had to explore some more what this collision is about... How do those two worlds recognize how they're different from seemingly different looks?" }, { "start": 341.94, "end": 347.419, "text": "What is interesting about Agente AI, it introduces a new participant into the ecosystem." }, { "start": 347.56, "end": 358.12, "text": "So all of sudden we have software that can interpret objectives and make decisions in initiate actions on behalf of users... That's really novel!" }, { "start": 358.719, "end": 360.28, "text": "And not just like on its own." }, { "start": 360.479, "end": 367.84, "text": "but then again how do these features mesh with our world payment digital assets?" }, { "start": 369.219, "end": 380.34, "text": "Previously we have seen that AI mainly supported decision-making, but now were seeing agents actually stepping in and participating actively the payment process itself." }, { "start": 380.4, "end": 400.039, "text": "And that really changes the ballgame entirely because it raises questions around trust accountability authorization settlement all these really important things that are highly relevant for central banks and payment operators, and then policymakers in general I would say." }, { "start": 400.88, "end": 419.479, "text": "So to get a better kind of understanding where yeah... A vision of how this world will evolve we had to get started on exploring how agentic AI works with payments And the paper i would say is just scratching the surface." }, { "start": 421.94, "end": 428.26, "text": "There's still a lot of work ahead of us in terms exploring and depth some of these topics that we will talk about today," }, { "start": 430.7, "end": 431.859, "text": "Yeah, and it's kind of still early." }, { "start": 431.979, "end": 432.2, "text": "Right?" }, { "start": 432.3, "end": 444.859, "text": "So if you think about that like the chat GVT moment was already I think four years ago but still we just... And everybody is talking about agenting and agentic payments in a prototype here and there But It Still Feels Very Early." }, { "start": 445.099, "end": 461.58, "text": "What i really appreciate About The Paper and its titled How Agentik AI Will Reshape Payments Of Course We will link this In The Show Note Is That You Really You Know Brought A Lot More Structure To This Whole Debate Because As I Remember Actually Estonia bit similar with the CVDC discussions." }, { "start": 461.719, "end": 469.56, "text": "You know, when you were writing the papers before people talked about it that was just lagging structure, lagging a joint terminology and everything." }, { "start": 470.02, "end": 472.859, "text": "so this is something I really appreciate about the paper." }, { "start": 473.039, "end": 476.84, "text": "also why we wanted to do an episode on this paper together" }, { "start": 484.159, "end": 490.88, "text": "We have to all make an effort collectively and this includes you guys as well, is your doing research yourself." }, { "start": 491.56, "end": 513.56, "text": "To get at the bottom of these I would say there's still a lot of challenges uh...and still a lots of analysis that are needed to understand better how agentic AI payments will work what impact it'll have on the economy." }, { "start": 514.24, "end": 520.44, "text": "And let's maybe start Sonja just first thing on, you know with a definition because as an economist I always love to." }, { "start": 520.78, "end": 524.92, "text": "You know started the definition like how would your define a genetic AI?" }, { "start": 525.04, "end": 526.22, "text": "What does it mean for you?" }, { "start": 526.32, "end": 527.86, "text": "or may be also what doesn't not me all" }, { "start": 528.16, "end": 528.34, "text": "right?" }, { "start": 528.42, "end": 531.84, "text": "i think its good to start was the definition First cuz." }, { "start": 531.94, "end": 533.56, "text": "clarity is absolutely vital." }, { "start": 534.32, "end": 541.58, "text": "our genetic AI too me is in a ice ai systems that can autonomously act beyond behalf of users." }, { "start": 542.26, "end": 546.4, "text": "So rather than merely providing recommendations, which we've had up till now." }, { "start": 546.98, "end": 563.0, "text": "These systems can interpret objectives and break them into distinct tasks And from there they then independently and autonomously interact with digital services and pursue specific goals With limited human intervention." }, { "start": 563.2, "end": 575.62, "text": "I want to say limited because as you will As we have noted in the paper and as we will discuss today, I do think that there's still human element needed especially for high-value transactions." }, { "start": 575.86, "end": 594.18, "text": "But in the payments context i would say agentiqi could move us from a world where humans explicitly initiate every transaction to a one where AI agents increasingly mediate decisions, and they also execute those decisions within predefined mandates." }, { "start": 594.68, "end": 596.1, "text": "Now what these mandates are?" }, { "start": 596.24, "end": 606.02, "text": "They're going to be different based on the entity individual who stands behind an agent in the intent of that specific transaction or purchase whatever their case may be." }, { "start": 606.8, "end": 607.3, "text": "so that's kind" }, { "start": 610.64, "end": 612.06, "text": "Sonya, if you allow." }, { "start": 612.28, "end": 637.5, "text": "So just to make sure that both I and the audience totally get it when it comes to a genetic AI what i'm understanding is What I understand from your paper and what you just said is that it's actually AI initiating payments under delegated authority at machine speed." }, { "start": 637.6, "end": 648.54, "text": "So an agent gets a budget per month, so let's say two hundred US dollars per month to be paid on whatever the prompt requests the agents to buy and pay for?" }, { "start": 649.04, "end": 650.3, "text": "Is that understanding correct?" }, { "start": 650.36, "end": 656.32, "text": "So this shift from click-to-pay to decide to pay which is the shift you're describing in the paper." }, { "start": 657.98, "end": 660.96, "text": "understanding we should now adopt and the recording here." }, { "start": 661.34, "end": 684.14, "text": "That's absolutely correct, it captures the main point or the main sort of shift because we're moving away from the world as I was saying that an agent to a situation where the agent can manage the entire payment or at your entire purchase cycle." }, { "start": 684.26, "end": 696.62, "text": "So not only can it initiate a purchase, let's say but I could also all of way too... Let's say dispute resolution and agent bought the wrong thing when wasn't within the budget." }, { "start": 696.76, "end": 702.34, "text": "like you can help resolve that issue with however is the seller online." }, { "start": 702.44, "end": 715.38, "text": "so this definitely big paradigm shift because we have, uh... We can enable an agent to have or vest an agent with full autonomy." }, { "start": 715.56, "end": 722.64, "text": "To act on our behalf and dictate how the agents' behavior will perform in our" }, { "start": 722.72, "end": 723.14, "text": "behalf.\"." }, { "start": 723.3, "end": 726.36, "text": "And so that-we haven't had up till now." }, { "start": 726.46, "end": 728.16, "text": "it was more than confined of like." }, { "start": 728.68, "end": 728.82, "text": "oh!" }, { "start": 728.96, "end": 733.16, "text": "It looks like you're looking for a specific product where you want make a specific payment?" }, { "start": 733.26, "end": 741.0, "text": "We recommend such and such and then you are still The human is still the one deciding whether to act upon it or not." }, { "start": 741.08, "end": 744.8, "text": "In this case, the agent basically you give him autonomy too." }, { "start": 745.44, "end": 761.58, "text": "when these specific criteria are met within a mandate that should provide the agent they can then autonomously be hacked on your behalf and like manage entire purchase from start to finish, if it's an e-commerce use case for instance." }, { "start": 761.78, "end": 762.28, "text": "Yeah excellent!" }, { "start": 762.32, "end": 770.26, "text": "It was really important from me before we now move on into the next section which is then on the impact of authentic AI and payments." }, { "start": 770.3, "end": 792.48, "text": "so really that the real world impact...it was super important for me that We know keep in the back off our all minds That this is really..that we are not talking about a word In which AI initiates payments under delegated authority From the human, from the prompter at machine speed which is a different world." }, { "start": 792.56, "end": 794.26, "text": "A different paradigm for payments." }, { "start": 794.48, "end": 808.06, "text": "and now really moving on to The impact of this concept On payments as we know it today I found interesting that in your paper you just introduced You argue that payments are undergoing This fundamental transition." }, { "start": 808.14, "end": 820.08, "text": "We just talked about So from being initiated through explicit human instructions to increasingly be initiated by AI agents acting behalf of users and you introduce an." }, { "start": 820.88, "end": 825.76, "text": "AI payment layered model to structure the discussion." }, { "start": 825.9, "end": 833.14, "text": "so you distinguish between first intent, formation orchestration then second authentication control and third settlement." }, { "start": 833.4, "end": 835.54, "text": "Could you maybe elaborate on this model?" }, { "start": 835.64, "end": 841.0, "text": "Why did you find it useful to structure this future world we are talking about in this way?" }, { "start": 841.12, "end": 844.04, "text": "So why three layers y not four or five can?" }, { "start": 844.08, "end": 850.14, "text": "he may be a liberate under model and while this future of payments so useful." }, { "start": 850.28, "end": 854.64, "text": "That's a good question, I think that i cannot answer why three and not four layers?" }, { "start": 854.94, "end": 860.12, "text": "because there is also a secret sauce behind our thinking on." }, { "start": 865.14, "end": 869.66, "text": "But I think it's more about the functions that these layers represent." }, { "start": 870.18, "end": 889.74, "text": "And this three-layer model, which we have developed is sort of a guiding framework for how we should think about the Gentigaean payments and its kind of centerpiece on this paper would argue It's not just an organizing framework or architectural concept." }, { "start": 890.1, "end": 897.14, "text": "It's also a very important risk mitigant and I explain shortly why that is the case." }, { "start": 898.36, "end": 911.84, "text": "First, i mentioned the collision between probabilistic-the probabilistic world of a genetic AI and the extremely deterministic world Of payment systems or you know settlement assets." }, { "start": 911.9, "end": 921.1, "text": "so one way...and we have to be creative how how we can reconcile these two worlds, because we want them to coexist." }, { "start": 921.16, "end": 943.22, "text": "We want to bridge these gaps and find a way to bring these two world together although as I said they're seemingly completely different Because As explained earlier the probabilistic reasoning is highly difficult To confine within you know A deterministic outcome." }, { "start": 943.32, "end": 951.9, "text": "so it's like you need to get to a logic for settlement where the instruction produces this same outcome over and over again." }, { "start": 952.76, "end": 957.5, "text": "And responsibilities are clearly allocated, and transactions are auditable in legally final." }, { "start": 957.58, "end": 959.62, "text": "that's absolutely vital for settlement." }, { "start": 960.14, "end": 973.58, "text": "but then AI kind of brings in this wild card because it introduces probabilistic reasoning and is highly adaptive decision-making as we all know from the AI that were using our daily lives." }, { "start": 973.66, "end": 979.54, "text": "So, It was a matter for us to reconcile these fundamentally different design logics." }, { "start": 979.68, "end": 982.92, "text": "And This Is Where The Three Layer Framework Was Born." }, { "start": 983.7, "end": 987.78, "text": "Because For Us We Quickly Understood That We Have To Separate The Functions." }, { "start": 988.5, "end": 997.0, "text": "And again, whether that happens in three layers or four layers or ten like it's really up to you know the creativity of the author." }, { "start": 997.16, "end": 999.28, "text": "Or Like The designer Of this framework." }, { "start": 999.38, "end": 1010.3, "text": "but for us we saw That In layer one We clearly need To separate the intent and orchestration Of the AI This Intent and Orchestration Layer." }, { "start": 1010.84, "end": 1022.24, "text": "We kind of leaving the AI to interpret user goals, gather all the required information evaluate different options and then propose actions." }, { "start": 1022.28, "end": 1026.22, "text": "Actions based on those are based in that due diligence so-to speak." }, { "start": 1026.939, "end": 1035.619, "text": "So this layer is responsible for reasoning an orchestration And This Is If You Will where you kind of leave it through." }, { "start": 1035.659, "end": 1038.52, "text": "the agents too do work their magic." }, { "start": 1039.26, "end": 1044.56, "text": "In Layer Two however We introduce these deterministic controls." }, { "start": 1045.04, "end": 1053.5, "text": "So here is where we add more rigidity into the mix, so when it comes to authentication control and authorization." }, { "start": 1054.1, "end": 1073.3, "text": "It's important that in this layer To verify that an agent is authorized to act check whether a payment falls within a predefined mandate or A budget that user has or liquidity needs at entity has whatever The case may be And whether any relevant compliance requirements are satisfied." }, { "start": 1073.34, "end": 1078.98, "text": "That's particularly important for the world of payments, but there a number of different compliance requirements." }, { "start": 1079.639, "end": 1085.399, "text": "So the key purpose for layer two was to place these enforceable guardrails around agent behavior." }, { "start": 1085.84, "end": 1087.58, "text": "so think of it as like in layer one." }, { "start": 1087.639, "end": 1094.02, "text": "they you know The agents are kind of at their liberty to gather all the input that- Their" }, { "start": 1094.159, "end": 1097.5, "text": "reason plan compare negotiate whatever They need to do right?" }, { "start": 1097.78, "end": 1098.02, "text": "Yes." }, { "start": 1098.12, "end": 1108.699, "text": "and then in Layer Two they get into this state where there's some guardrails that have to be enforced before, around these agent behaviors." }, { "start": 1108.8, "end": 1109.82, "text": "So here it becomes." }, { "start": 1109.879, "end": 1116.22, "text": "until okay we're confining you two specific parameter or mandate of a budget whatever the case may." }, { "start": 1117.419, "end": 1130.159, "text": "and then finally in layer three I would say if your worth think about is terms like a pyramid at the settlement layer theres very little room for any kind of probabilistic behaviors." }, { "start": 1130.28, "end": 1135.419, "text": "I was saying earlier, like the settlement layer really has to remain unchanged." }, { "start": 1135.48, "end": 1149.24, "text": "So once an instruction passes the authorization layer these existing payment rails or a settlement infrastructures will execute the transaction with finality and really determinism." }, { "start": 1149.379, "end": 1162.659, "text": "so there's Whatever has been authorized for a payment or for purchase, whatever the use case is then gets settled with an absolutely clear outcome." }, { "start": 1162.82, "end": 1187.22, "text": "And I think that if you structure this thinking in this way You're able to ridge these two worlds because your separating functions Your seperating function along those layers and allow them to bring out really though the best of the possible outcomes in this kind of framework, right?" }, { "start": 1187.26, "end": 1196.379, "text": "So we're allowing this creative outlet for the agents within the intent layer if you want to call it creative." }, { "start": 1197.179, "end": 1198.82, "text": "That's one way to think about it." }, { "start": 1199.32, "end": 1212.699, "text": "and then we are adding more controls and guardrails as a payment or purchase progresses with ultimate goal that settlement whatever comes at end is absolutely crystal clear." }, { "start": 1213.22, "end": 1233.82, "text": "There cannot be any doubt about whether a settlement, it's final or not and that is why the Settlement layer will always have to completely separate from the rest because for any participant who is privy of that transaction participating in this transaction there has to be a clear cut outcome." }, { "start": 1235.26, "end": 1238.48, "text": "you know clear kind end result" }, { "start": 1239.22, "end": 1243.98, "text": "And if we now try Sonja to make this a little bit, you know practical let's say with an example and also Let me check." }, { "start": 1244.04, "end": 1246.139, "text": "If I understood what you were explaining." }, { "start": 1246.22, "end": 1248.98, "text": "so let Me for example imagine how i'm With my." }, { "start": 1249.379, "end": 1252.06, "text": "You Know the AI of My daily life." }, { "start": 1252.12, "end": 1255.659, "text": "So whatever This is Claude, Gemini, JetGBT or something." }, { "start": 1256.399, "end": 1262.78, "text": "In this future world that you're describing I'm in there and say hey i want a nice looking t-shirt in my size to buy for example." }, { "start": 1262.84, "end": 1268.199, "text": "then the first step The AI agent would basically check okay did they understand what the person wants?" }, { "start": 1268.58, "end": 1270.0, "text": "Do we have all of their information?" }, { "start": 1270.879, "end": 1275.84, "text": "You know orchestration also going into checking different market places etc." }, { "start": 1276.179, "end": 1277.84, "text": "is This was the first steps about?" }, { "start": 1277.879, "end": 1278.74, "text": "Did it get this correct?" }, { "start": 1278.939, "end": 1286.639, "text": "Yeah so if he were take on classic ecommerce use case which be a good one to showcase how these three layers would work." }, { "start": 1286.72, "end": 1298.86, "text": "So an intent layer is then where the user slash view in this case clearly communicates, I want to buy t-shirt for, I don't know, fifty dollars from that brand." }, { "start": 1299.699, "end": 1299.919, "text": "you know?" }, { "start": 1299.98, "end": 1304.36, "text": "i have this much...I can spend between fifty and hundred dollars." }, { "start": 1304.419, "end": 1310.379, "text": "let's say..i want the color to be this or whatever your parameters are ...your preference as user preferences." }, { "start": 1311.179, "end": 1317.96, "text": "So you feed that intent to the agent and then in the intent layer, AI understands." }, { "start": 1318.04, "end": 1320.699, "text": "The agent tries to interpret your goals." }, { "start": 1321.54, "end": 1323.06, "text": "it will gather all of information." }, { "start": 1323.12, "end": 1329.1, "text": "so we'll go out there look for all possible t-shirts that might fit this bill." }, { "start": 1331.52, "end": 1334.08, "text": "Intent that you have expressed your preferences." }, { "start": 1334.32, "end": 1337.72, "text": "You do express and then it will evaluate all the options." }, { "start": 1338.1, "end": 1345.939, "text": "Oh, It looks like this t-shirt is Two, you know falls within your budget But the colors might be too bright or you know?" }, { "start": 1346.32, "end": 1357.1, "text": "He'll gather All the intelligence they can on that specific T shirt and try to Give you the closest match based on your preferences And then I will propose action." }, { "start": 1357.26, "end": 1364.0, "text": "so he would look be responsible and that layer will reason or orchestrate but not execute yet." }, { "start": 1364.32, "end": 1368.74, "text": "And so then, you know it all depends on like how you set this up as well right?" }, { "start": 1368.82, "end": 1379.179, "text": "Like somebody might prefer to still take a look at the options of AI picked out But some users would say no if my criteria are met." }, { "start": 1379.28, "end": 1386.02, "text": "Then in second layer an authentication control and authorization layer There will be another check against." }, { "start": 1386.12, "end": 1390.48, "text": "have all the parameters that Jonas has had been met, meaning does it fit?" }, { "start": 1390.56, "end": 1393.1, "text": "The budget is within the price range you determine." }, { "start": 1393.199, "end": 1394.419, "text": "Is it the color you like?" }, { "start": 1394.5, "end": 1396.399, "text": "Is It the brand of you normally buy?" }, { "start": 1396.919, "end": 1397.84, "text": "Is a different man?" }, { "start": 1397.96, "end": 1399.26, "text": "what is this shipping cost?" }, { "start": 1399.419, "end": 1400.62, "text": "What is the shipping time?" }, { "start": 1401.24, "end": 1409.879, "text": "All these are basically checks against your list of the criteria that you have defined, which is provided to the agent." }, { "start": 1410.34, "end": 1418.3, "text": "And then it will also make sure that the agent doesn't all of a sudden buy something completely different from what you've expressed." }, { "start": 1418.36, "end": 1422.02, "text": "let's say a purple dress instead of yellow t-shirt right?" }, { "start": 1422.1, "end": 1426.659, "text": "So... Then it passes layer two when it says yes!" }, { "start": 1426.919, "end": 1431.379, "text": "All checks are at all parameters have been met." }, { "start": 1431.52, "end": 1434.12, "text": "It's within budget, color brand everything fits." }, { "start": 1434.879, "end": 1441.58, "text": "and then in the third layer it will grab your credit card information or whatever is your payment instrument." }, { "start": 1441.78, "end": 1448.54, "text": "however you're paying for this purchase And it'll actually execute and buy this t-shirt for you InlayerTree." }, { "start": 1448.62, "end": 1454.78, "text": "it will settle the Purchase with using Your credit card Information as You have fed it previously?" }, { "start": 1455.199, "end": 1458.199, "text": "Then all you Have to do Is lean back and wait For your t-shirts to arrive." }, { "start": 1458.26, "end": 1461.699, "text": "so That's basically how it works." }, { "start": 1461.96, "end": 1465.439, "text": "Well, and for me this is a great use case on you I really hate shopping in particular." }, { "start": 1468.879, "end": 1473.159, "text": "does this trust worthily and is not hallucinating, I would be one of the first." }, { "start": 1473.24, "end": 1478.82, "text": "But i think it was very important what you mentioned on end-to-end settlement that... Of course!" }, { "start": 1478.86, "end": 1484.879, "text": "And this something we will also go a lot deeper into future episodes like What This Means Of Payment Will Be right?" }, { "start": 1484.939, "end": 1491.76, "text": "This can be The Credit Card A stablecoin maybe a CVDC at some point or token as deposit back account." }, { "start": 1492.28, "end": 1492.659, "text": "Whatever" }, { "start": 1492.82, "end": 1493.02, "text": "Right?!" }, { "start": 1493.56, "end": 1495.02, "text": "That's Something Again Which We May." }, { "start": 1495.939, "end": 1496.659, "text": "Maybe we also have time." }, { "start": 1496.699, "end": 1509.879, "text": "later in this episode or future episodes, people discuss because obviously different forms of money are differently well or badly suited for a settlement specific amount of payments that are initiated and driven by agents." }, { "start": 1510.139, "end": 1511.58, "text": "You know what's really cool about it?" }, { "start": 1513.04, "end": 1515.5, "text": "I'm with you...I do not like shopping!" }, { "start": 1515.939, "end": 1528.82, "text": "It is unusual to see women but i absolutely dislike shopping time-consuming and the lines, you know it's just like the bright lights." }, { "start": 1528.879, "end": 1541.5, "text": "And then I also have a problem with oh that size doesn't fit or the color whatever." }, { "start": 1542.139, "end": 1544.04, "text": "Like i don't like dealing without either." }, { "start": 1544.159, "end": 1545.32, "text": "so in this conundrum." }, { "start": 1545.52, "end": 1550.84, "text": "but what is beautiful about the Segentic AI and eCommerce for instance?" }, { "start": 1551.72, "end": 1555.1, "text": "It can handle the dispute resolution, let's say." }, { "start": 1555.26, "end": 1582.0, "text": "So it can help me also return all of the items that don't... you know if whatever t-shirt didn't fit well or just don't like the color and what ever case may be.. I mean of course we still have to walk the parcel back wherever it needs to be dropped off, but if will help you with like handling all the returns or any kind of dispute resolution." }, { "start": 1582.08, "end": 1585.96, "text": "Let's say we're charged a lot more than you had intended to pay." }, { "start": 1586.6, "end": 1602.5, "text": "It can handle that part as well If really can go as far as manage whole conversation with the seller around returning an item Or whatever case issue is what has specific items?" }, { "start": 1602.879, "end": 1608.82, "text": "So, Sonya next to describing this core tension so well that is a genetic AI." }, { "start": 1608.98, "end": 1612.86, "text": "Is probabilistic and adaptive in payment razor deterministic?" }, { "start": 1612.96, "end": 1617.56, "text": "And require legal finality and auditability and clear accountability." }, { "start": 1617.899, "end": 1634.54, "text": "I think describing discord tension itself as already a precious contribution because not everyone thinking about the genetic AI has it clearly in mind, but then next to describing this core tension you as a solution propose the three layer model." }, { "start": 1634.6, "end": 1642.48, "text": "In which you keep that probabilistic stuff upstream and the model and keep authorization layer two and settlement layer free really rule bound." }, { "start": 1643.0, "end": 1657.179, "text": "so beyond us let's say conceptual contribution video also come across Use cases in payments and there are many that lend themselves particularly well for a genetic AI." }, { "start": 1657.399, "end": 1663.84, "text": "so on the roadmap of these many payments use case is which are the first." }, { "start": 1666.52, "end": 1669.379, "text": "Agents pay, you mentioned dispute resolution." }, { "start": 1670.06, "end": 1671.46, "text": "You mentioned e-commerce." }, { "start": 1671.52, "end": 1676.0, "text": "So what are your top three if you will where we would see agents act first?" }, { "start": 1676.3, "end": 1677.46, "text": "Oh Top Three okay." }, { "start": 1677.699, "end": 1680.419, "text": "so you talked about its length about e-Commerce." }, { "start": 1680.5, "end": 1683.84, "text": "I think that one's like front left and center." }, { "start": 1683.98, "end": 1696.679, "text": "That was the number One use case And We already See a Number of agent guy protocols that Are being Developed at a rapid rampant speed like new ones being churned out as we speak." }, { "start": 1697.179, "end": 1705.04, "text": "So that's definitely an area use case, I will see the highest adoption if i may give that kind of prediction." }, { "start": 1705.34, "end": 1712.36, "text": "um...I do see a lot of potential as well for compliance and reporting automation." }, { "start": 1713.02, "end": 1715.12, "text": "We write about that in uh on the paper." }, { "start": 1715.86, "end": 1729.98, "text": "so there is definitely identify anomalies, suspicious transactions and like then instead of suggesting actions actually executing them." }, { "start": 1730.06, "end": 1739.58, "text": "And saying okay this looks kind of a suspicious transaction or something funky let's you know activate process such-and-such." }, { "start": 1740.26, "end": 1741.78, "text": "that will help the supervisors." }, { "start": 1741.86, "end": 1750.659, "text": "they'll receive those payment systems for instance to get more efficiency because obviously it is busy." }, { "start": 1751.28, "end": 1760.1, "text": "It's a busy type of work to monitor and oversee print transactions on daily basis, so Genting Act can really help." }, { "start": 1761.22, "end": 1766.26, "text": "If managed properly obviously there is risks in gaps with everything." }, { "start": 1767.379, "end": 1772.84, "text": "but if managed properly this could be very powerful too for oversight supervision." }, { "start": 1773.439, "end": 1774.179, "text": "And did you structure" }, { "start": 1774.22, "end": 1781.419, "text": "the use cases like around consumer and then point of sale?" }, { "start": 1783.08, "end": 1785.76, "text": "I don't know, transaction banks." }, { "start": 1785.8, "end": 1790.139, "text": "So do already cluster use case in a paper." }, { "start": 1790.8, "end": 1793.179, "text": "Yes we did." }, { "start": 1794.22, "end": 1802.04, "text": "for us it was less about seeing exactly who will be the participants and like for the e-commerce use case." }, { "start": 1802.08, "end": 1825.159, "text": "For instance, we didn't really you know identify specifically who the participants are whether it's an individual or entity but were explaining in such terms where its...for most part You can think about an individual purchasing something But they can very well apply to machine to machine payment Or business to business payments." }, { "start": 1826.28, "end": 1852.919, "text": "You know, that I think was like something we didn't want to get too bogged down on but rather wanted keep it a bit more general and explain what the impact is for e-commerce in general without identifying whether its individual to merchant or merchant to merchant because ultimately they will depend how you wanna draw a scope around your use case." }, { "start": 1853.0, "end": 1860.699, "text": "This works... regardless of who the participant is for a specific payment or purchase." }, { "start": 1861.22, "end": 1864.74, "text": "And I guess, because you asked me for your top three but i only gave you two..." }, { "start": 1865.5, "end": 1866.58, "text": "Then go for number three!" }, { "start": 1866.78, "end": 1867.98, "text": "...I'm paying attention here." }, { "start": 1868.06, "end": 1885.28, "text": "okay um no but i think this third one in my mind and this is where i do think we need to spend a lot more time on robust analysis treasure and liquidity optimization." }, { "start": 1886.04, "end": 1903.02, "text": "So helping entities with optimizing liquidity management, so this is really interesting because right now especially in the context of foreign exchange liquidity... Do you see that on paper?" }, { "start": 1903.5, "end": 1905.86, "text": "Foreign exchange markets are very disparate." }, { "start": 1905.899, "end": 1908.24, "text": "there's many venues where FX is traded." }, { "start": 1910.32, "end": 1912.0, "text": "Yes, these are automated still." }, { "start": 1912.1, "end": 1916.02, "text": "But like not to the point where you're actually executing transactions right?" }, { "start": 1916.06, "end": 1923.52, "text": "It's more like even if we still have the option or a choice but You can hear automate routine financial decisions." }, { "start": 1923.62, "end": 1933.06, "text": "when liquidity allocation would become more optimized and I think this is something were especially in the prosperity payments context This could be really powerful." }, { "start": 1933.52, "end": 1939.54, "text": "now mind you of course that it comes with warnings and like flags, all sorts of things." }, { "start": 1940.32, "end": 1946.84, "text": "And this is why because there's challenges around it you know we have to give a lot more attention." }, { "start": 1946.939, "end": 1956.56, "text": "We have to spend some quality research on finding out ways how can make sure that these gaps in challenge exists at the moment." }, { "start": 1956.86, "end": 1963.659, "text": "do not bring potentially very powerful use case down or they just block it." }, { "start": 1964.96, "end": 1966.84, "text": "We haven't spent enough time thinking about it." }, { "start": 1966.899, "end": 1978.98, "text": "So I think, again a collective effort to think through how we can resolve some of these challenges so that we can harness the biggest potential for this technology and liquidity management specifically." }, { "start": 1979.28, "end": 1982.04, "text": "Okay, so this was now three when I counted correctly Sonja." }, { "start": 1982.679, "end": 1984.52, "text": "So thanks for also sharing the third one." }, { "start": 1985.06, "end": 1993.5, "text": "and what was very interesting to me is always when i think about you know agents in like agenetic payments or AI in payments...I always think about the payment itself!" }, { "start": 1993.84, "end": 1994.78, "text": "The e-commerce payment." }, { "start": 1996.04, "end": 2002.36, "text": "What you said is basically that it can be way more, so they help to do compliance reporting automation." }, { "start": 2002.54, "end": 2005.179, "text": "Do even the dispute resolution which we know sucks right?" }, { "start": 2005.24, "end": 2015.6, "text": "You ordered something and then send your wrong t-shirt or package didn't really survive let's say So I think thats super good because my use case view was always too narrow." }, { "start": 2016.159, "end": 2026.659, "text": "but obviously this also means requires probably even more time to flourish because if I think about the agent would take over my dispute process, it wouldn't have to have access whatever." }, { "start": 2026.74, "end": 2035.1, "text": "To email inbox or write emails for some of the suppliers and communicate with them on behalf." }, { "start": 2035.24, "end": 2042.48, "text": "so that's a little bit more effort than just initiating payment on behalf of humans isn't?" }, { "start": 2042.78, "end": 2043.82, "text": "Oh yeah absolutely!" }, { "start": 2044.919, "end": 2051.52, "text": "basically authorizing your agent to use your credit card information, email or personal information." }, { "start": 2051.679, "end": 2059.92, "text": "I know that specifically in the EU and Germany particularly privacy is a big concern." }, { "start": 2059.98, "end": 2064.739, "text": "this something we generally have spend more time thinking about as well." }, { "start": 2066.4, "end": 2072.48, "text": "do you really want level of access to an agent." }, { "start": 2072.679, "end": 2074.02, "text": "What if the agent goes rogue?" }, { "start": 2074.08, "end": 2087.04, "text": "Like, how do you reel in the agent back into it and limit access to third parties or unauthorized individuals?" }, { "start": 2088.98, "end": 2093.719, "text": "It's it's not as easy as just saying like oh, it's super convenient." }, { "start": 2094.199, "end": 2097.2, "text": "Like I'll get my t-shirt and i will be great." }, { "start": 2097.4, "end": 2112.34, "text": "everything will be awesome if we do need to spend a lot more work on thinking about these challenges including cyber security and data privacy." }, { "start": 2112.64, "end": 2128.299, "text": "right because We don't have proper guardrails and praise just yet on how to ensure that this agent doesn't or like to pre-end cyber attacks, let's say unauthorized access to our personal information." }, { "start": 2128.7, "end": 2138.7, "text": "Our credit cards... How do we build a proper accountability framework in governance around it so there'd be limit and robust cybersecurity frameworks?" }, { "start": 2138.74, "end": 2144.52, "text": "So did we limit access to unauthorised third parties tour vital information." }, { "start": 2144.66, "end": 2154.319, "text": "So yes, it is still one of those question marks that we need to spend considerable time and energy on if we want to get this right?" }, { "start": 2154.68, "end": 2158.48, "text": "Yes, Sonja let's do so right now and really double click these challenges!" }, { "start": 2158.74, "end": 2182.759, "text": "You outlined cyber security data privacy already but then for example also there is an outline tension between on the one hand probabilistic adaptive decision making by the one or multiple agents and then on the other hand, that deterministic requirements of payment infrastructure which is why you introduced your three-layer model." }, { "start": 2184.52, "end": 2189.96, "text": "That in itself could be a challenge overcoming those or bridging these two worlds." }, { "start": 2190.54, "end": 2198.36, "text": "what are key challenges for making authentic AI happen to have its impact materialize?" }, { "start": 2199.22, "end": 2201.24, "text": "Yeah, I you mentioned one just now." }, { "start": 2201.42, "end": 2208.5, "text": "So this tension between probabilistic AI decision making and the very deterministic nature of payment systems." }, { "start": 2208.54, "end": 2213.84, "text": "so that's one that is at the core of what we're trying to conceptualize here." }, { "start": 2214.279, "end": 2218.14, "text": "like You know The three-layer framework Is Just One Way How To Think About It." }, { "start": 2218.18, "end": 2224.6, "text": "i'm Sure And Have A Big Believer In Human Ingenuity And Progress That We Will Make More Headway." }, { "start": 2228.42, "end": 2232.66, "text": "industry participants have gotten in touch to talk about developments." }, { "start": 2233.18, "end": 2252.259, "text": "And we are seeing a lot more effort on solving for some of the layer two, I mean authorization layer solutions that specifically speak to that layer because that's going to be one of the critical pieces to figure out like how do you build these guardrails properly?" }, { "start": 2252.4, "end": 2257.94, "text": "so the agent is not doesn't go rogue all but is also not too constrained." }, { "start": 2257.98, "end": 2265.1, "text": "So the balance this like need for allowing it to do what its intended to do, But within the confines that you determine." }, { "start": 2265.16, "end": 2273.48, "text": "so layer two Is really vital in that regard and I've seen an industry has been already thinking about some solutions around It." }, { "start": 2273.92, "end": 2276.9, "text": "The second one we talked About was a cybersecurity risk." }, { "start": 2278.08, "end": 2288.799, "text": "We will see higher surface attack if You Will because you know, we could all of a sudden have unauthorized or malicious actors targeting agents directly." }, { "start": 2289.48, "end": 2306.16, "text": "Or manipulate their decision processes obtaining payment credentials personal information that's all something that um...you know..we have to think about harder and I don't want us say that it would be will have a solution right away in." }, { "start": 2306.46, "end": 2308.68, "text": "this is where more work is needed to figure out." }, { "start": 2308.759, "end": 2316.2, "text": "how do we build the cybersecurity framework that limits the access of malicious actors to target agents directly." }, { "start": 2317.0, "end": 2319.48, "text": "Then, third one I would say is opacity." }, { "start": 2320.06, "end": 2322.88, "text": "there are as we know and this nothing new." }, { "start": 2323.12, "end": 2335.819, "text": "it's not just related to payment but agentic systems may create a opacity so it will become difficult to understand why a specific decision was taken or how an agent reached a particular outcome." }, { "start": 2336.299, "end": 2343.2, "text": "It might be more obvious in the case of t-shirt, we talked about where you want a Lielo T-shirt between fifty and hundred dollars." }, { "start": 2343.24, "end": 2358.4, "text": "whether that is the case maybe but it becomes infinitely harder if you have like complex cross border transaction with multiple intermediaries involved disentangle what went wrong in case something goes wrong." }, { "start": 2358.54, "end": 2363.18, "text": "So this is where opacity is problematic." }, { "start": 2363.259, "end": 2366.42, "text": "as you're trying to unravel what has happened, it's not like." }, { "start": 2366.46, "end": 2376.42, "text": "you will have clarity right away on what has occurred within that probabilistic especially with layer one and the probabilistic world." }, { "start": 2377.14, "end": 2384.44, "text": "And then finally I'd say there is a risk around these correlated behavior among many agents." }, { "start": 2384.48, "end": 2406.52, "text": "So if you have like agents all of a sudden, hundreds or thousands of them that are picking up on the market signal when it comes to liquidity for instance and then they could really amplify any volatility in create potential systemic risks or unintended market dynamics that is...could have disastrous consequences potentially." }, { "start": 2406.7, "end": 2414.1, "text": "so how do we build our structured this system in a way that it doesn't disrupt financial stability." }, { "start": 2414.22, "end": 2416.96, "text": "So, that is something we have to think about." }, { "start": 2417.04, "end": 2422.98, "text": "some more and I think what also remains largely unresolved are these legal questions." }, { "start": 2423.94, "end": 2424.94, "text": "you know who's liable?" }, { "start": 2425.04, "end": 2426.04, "text": "Who's accountable?" }, { "start": 2427.4, "end": 2435.0, "text": "What about authorization when consent with autonomous systems initiate payments like how do you define accountability?" }, { "start": 2435.08, "end": 2437.279, "text": "and liability within the framework?" }, { "start": 2437.72, "end": 2440.759, "text": "Is it the agent, but then who does the agent belong to?" }, { "start": 2440.94, "end": 2447.88, "text": "So this is why you know in the papers we have suggested some mitigants or ways how to think about these challenges." }, { "start": 2447.98, "end": 2454.7, "text": "But that is like an area that also remains largely untapped." }, { "start": 2454.92, "end": 2460.42, "text": "We have not figured out How we build a proper accountability framework around It." }, { "start": 2460.6, "end": 2465.14, "text": "and um...how do we define liability In this context?" }, { "start": 2465.38, "end": 2465.92, "text": "Yeah, excellent." }, { "start": 2465.98, "end": 2468.84, "text": "So you're already listing quite some challenges." }, { "start": 2469.36, "end": 2484.38, "text": "what I'm hearing is that authorization will be tricky That is the agent must not go rogue so needs guardrails but also have agency to take away Some of the work and actually receive delegated tasks." }, { "start": 2484.819, "end": 2486.36, "text": "then you mentioned cyber security?" }, { "start": 2486.799, "end": 2487.819, "text": "So a gentle gay eyes." }, { "start": 2487.98, "end": 2489.34, "text": "even increasing they are ready." }, { "start": 2489.5, "end": 2504.84, "text": "high attack surface of payments, a genetic AI is even increasing this attack surface in that also an agent can now falsely communicate payment details like bank account numbers." }, { "start": 2504.9, "end": 2506.54, "text": "So you mentioned opacity." }, { "start": 2506.66, "end": 2512.5, "text": "why did the agent make a buying decision and this correlation between multiple agents?" }, { "start": 2513.779, "end": 2524.42, "text": "To be even more specific and already tackle those challenges, which a subset of our audience at least will always soon try to tackle." }, { "start": 2525.44, "end": 2528.759, "text": "How about we double-click on cybersecurity?" }, { "start": 2530.02, "end": 2533.0, "text": "I mean, cybersecurity is a major challenge." }, { "start": 2533.18, "end": 2538.68, "text": "It still is of payments and particularly e-commerce—a major part of global payment occurs online." }, { "start": 2539.18, "end": 2551.42, "text": "And how can the community or industry structure then reduce this attack surface again with these increasing cyber security risks?" }, { "start": 2551.819, "end": 2554.14, "text": "Well that's a billion dollar question!" }, { "start": 2557.06, "end": 2559.66, "text": "None of us really have a proper answer to." }, { "start": 2559.9, "end": 2568.799, "text": "I don't think that there is one foolproof system, that will remedy or mitigate the potential cybersecurity risk." }, { "start": 2568.88, "end": 2580.66, "text": "so even if we were to design a system based on three-layer framework like they're still our potentially vulnerabilities that malicious actors could exploit." }, { "start": 2580.799, "end": 2591.04, "text": "So this just says When we're building a system, you have to think about how do I future-proof it?" }, { "start": 2591.279, "end": 2594.66, "text": "How do i make sure that like in Building It." }, { "start": 2595.2, "end": 2599.779, "text": "I'm trying to build something That's robust and that is impenetrable To attacks." }, { "start": 2600.279, "end": 2605.259, "text": "of course everyone would want that right but things are evolving." }, { "start": 2605.66, "end": 2608.86, "text": "technology develops systems change." }, { "start": 2608.96, "end": 2618.46, "text": "so In whatever We're building we Have to have that forward looking sort Of vision that, you know we're future-proofing." }, { "start": 2618.74, "end": 2620.58, "text": "The design leaves enough flexibility." }, { "start": 2620.68, "end": 2624.04, "text": "so then we can adapt as technology shifts and technology evolved." }, { "start": 2624.08, "end": 2642.5, "text": "because if I build a secure or resilient cyber resilience system now it might be resilient only for the next year let's say And there are new entrants New malicious actors New software That can exploit vulnerability." }, { "start": 2643.799, "end": 2645.259, "text": "safe and resilient today." }, { "start": 2645.92, "end": 2664.2, "text": "So this is my key sort of message to everyone, like in whatever you're building just make sure that there's enough design flexibility inherent changes and upgrades." }, { "start": 2664.34, "end": 2668.9, "text": "And you don't build like a monolithic structure, but something that is modular." }, { "start": 2669.0, "end": 2673.12, "text": "there's adaptable if you can change as technology evolves." }, { "start": 2673.16, "end": 2680.98, "text": "the software involves this you know, things are happening so that it stands a test of time." }, { "start": 2681.04, "end": 2689.819, "text": "So then you can adapt your system to be resilient against future attacks and this will require upgrades there's no doubt about it." }, { "start": 2689.88, "end": 2704.48, "text": "we cannot build something That would last and be impenetrable forever without the ability to adapt it or change it to insulate over time kind of goal with the time." }, { "start": 2704.6, "end": 2721.94, "text": "So that's, I would say is a key message because like again even three-layer framework if we separate functions out and it keeps harping on that point The three layer framework not just architectural concept but also risk." }, { "start": 2722.04, "end": 2733.74, "text": "mid again Because you separated function so its easier to manage any risks or issues at occur it's not going to be foolproof." }, { "start": 2733.88, "end": 2752.44, "text": "There will be vulnerabilities that can't be exploited, so we have to build something that can be upgraded adapted changed two new developments without kind of throwing away everything but like keeping the core structure and then adjusting it just adjusting it." }, { "start": 2753.0, "end": 2767.0, "text": "So yeah It could stand a test time involved with would change the technology and changes in software, protocols new things that are happening as we speak." }, { "start": 2767.38, "end": 2768.779, "text": "Space is really super dynamic." }, { "start": 2768.96, "end": 2777.84, "text": "before you ask your last questions where were asked for an outlook Some insights you have and how this looks in twenty thirty, twenty-thirty five." }, { "start": 2777.88, "end": 2787.94, "text": "Maybe one question before that because for our audience obviously they hear us a lot talk about CBDC is talking as deposits wholesale retail stablecoins credit cards all of that." }, { "start": 2788.0, "end": 2795.68, "text": "so What is now the kind of means-of payment that agents will use?" }, { "start": 2795.759, "end": 2799.66, "text": "And we would definitely use this podcast to dive down, or dive deeper into the topic." }, { "start": 2799.72, "end": 2813.22, "text": "To talk with stablecoin issuers maybe also with banks together a little bit different perspectives there but you also briefly touched us in the paper Sonia like the role that also stablecoins potential CBDCs could play in these agentic payments." }, { "start": 2813.34, "end": 2817.22, "text": "Ram Like what's your initial perspective on this?" }, { "start": 2817.84, "end": 2824.36, "text": "means off payment that Agents could maybe use in the future for our shopping experience and authentic payments of the" }, { "start": 2845.58, "end": 2848.66, "text": "one payment instrument you need to use with a Gentik AI." }, { "start": 2848.779, "end": 2855.44, "text": "So we're not prescriptor like that, really frankly I would say it could work for any payment instrument Really?" }, { "start": 2855.98, "end": 2874.54, "text": "so long as It satisfies The mandate in a parameter is that You have...you are user or an entity Have defined Frankly also depends on the use case For instance and cross-border payments you know, a high value transaction." }, { "start": 2874.759, "end": 2883.02, "text": "Maybe there's preferred payment over small scale or low-value purchases on the e-commerce side." }, { "start": 2883.08, "end": 2890.52, "text": "so that will also drive like the choice of the payment instrument but that's no different from The decision we have to make now." }, { "start": 2890.84, "end": 2894.799, "text": "So That is very much on the answer I would have for that." }, { "start": 2895.259, "end": 2907.66, "text": "When it comes to digital assets specifically We just have to be creative in how we think about Mapping these agentic mandates or the mandates of perimeters that the agents operate under, how do we map them?" }, { "start": 2907.74, "end": 2913.4, "text": "Or is there a way to map them to the smart contracts that are underlying digital assets." }, { "start": 2913.46, "end": 2918.2, "text": "So you know in terms of technical interoperabilities There's something that we can achieve there." }, { "start": 2918.799, "end": 2920.68, "text": "Is there mapping that needs to happen?" }, { "start": 2921.319, "end": 2925.52, "text": "What is the potential benefit Of enabling such mapping?" }, { "start": 2926.16, "end": 2945.38, "text": "I think That could be Something interesting too To research like a technical interoperability that we can unlock through these smart contracts and online digital assets, they will work in concert with the confines." }, { "start": 2945.52, "end": 2950.08, "text": "And then we define layer two of the agentic AI payment systems." }, { "start": 2950.759, "end": 2954.16, "text": "so yeah it's kind of something fun to explore." }, { "start": 2954.74, "end": 2955.14, "text": "I would think" }, { "start": 2956.42, "end": 2963.92, "text": "Yeah, so I know that your IMF note is deliberately not super bullish and balanced on this one though." }, { "start": 2979.86, "end": 2980.819, "text": "Probably I share two." }, { "start": 2981.02, "end": 2993.86, "text": "My first is stablecoins are programmable by default so the speed, the spend limits The conditions and guardrails live in their wallet itself not in a bank's back office." }, { "start": 2993.94, "end": 2997.94, "text": "So the controls really travel with stablecoins as they form of money." }, { "start": 2998.48, "end": 3005.12, "text": "And second argument for why stablecoins particular lend themselves to agentic payments Is that there always on?" }, { "start": 3005.259, "end": 3005.9, "text": "Always final!" }, { "start": 3005.98, "end": 3020.2, "text": "So agents operate twenty four seven across time zones, but then batch windows cutoff times and correspondent chains are really human or human scale constraints." }, { "start": 3020.36, "end": 3027.38, "text": "So I have a pretty strong opinion that is agents need money that behaves like software and stablecoins do exactly." }, { "start": 3037.48, "end": 3042.24, "text": "So, there is many ways to skin a cat." }, { "start": 3042.9, "end": 3051.54, "text": "And I think we have to keep an open mind because you cannot just assume that... There's only one right solution or one way to go?" }, { "start": 3053.52, "end": 3061.12, "text": "We will be in the transition period for awhile and so it has to work." }, { "start": 3062.38, "end": 3071.04, "text": "Frankly also e-commerce use cases nowadays works with simple credit card and credit cards like the credit card information that you provide." }, { "start": 3071.1, "end": 3074.94, "text": "And it works well, there's no reason why that shouldn't work." }, { "start": 3075.0, "end": 3076.9, "text": "Mind you if everything is managed properly?" }, { "start": 3077.02, "end": 3082.52, "text": "Like data privacy is insured and have no exploits or data breaches." }, { "start": 3082.62, "end": 3087.66, "text": "but I do think that there are more ways how to enable a Gen-TKI." }, { "start": 3088.08, "end": 3090.72, "text": "payment instruments will be very much use case dependent." }, { "start": 3090.9, "end": 3095.58, "text": "All right, I think the question around which means of payment is best suited for Agente Gai." }, { "start": 3096.38, "end": 3102.279, "text": "Is potentially write next to tackle in an upcoming episode?" }, { "start": 3102.6, "end": 3109.06, "text": "For now Sonja we would really appreciate that you spend so much time covering or researching Agente GI." }, { "start": 3110.18, "end": 3118.12, "text": "share some thinking about a future and outlook So trigger your thinking on it." }, { "start": 3119.12, "end": 3128.4, "text": "what do the share of payments really globally that will be initiated by AI agents in twenty thirty," }, { "start": 3128.66, "end": 3128.72, "text": "so" }, { "start": 3128.84, "end": 3130.94, "text": "already about three and a half years." }, { "start": 3131.36, "end": 3132.96, "text": "And then in twenty-thirty five." }, { "start": 3134.48, "end": 3139.68, "text": "you may want to use interballs instead just one percentage number to show your prediction." }, { "start": 3140.0, "end": 3140.96, "text": "Crystal ball time!" }, { "start": 3141.68, "end": 3145.5, "text": "Exactly I was going say let me get my crystal ball out." }, { "start": 3148.52, "end": 3150.34, "text": "No, I joke aside." }, { "start": 3151.92, "end": 3159.7, "text": "It's still too early to make credible numerical forecasts and now it would be right for me to do so." }, { "start": 3160.759, "end": 3166.34, "text": "But what seems clear is that the share of agent-mediated payments will grow significantly over time." }, { "start": 3166.74, "end": 3178.0, "text": "I don't know by what year exactly and what exact number there would be, but we can expect to see these payment growth in conjunction with protocols under LIDs' payments." }, { "start": 3178.299, "end": 3186.2, "text": "And so for the pace of adoption—that's a whole different story because it'll depend less on AI capabilities themselves." }, { "start": 3189.62, "end": 3198.86, "text": "all these open questions and challenges that I've mentioned in our podcast today, so question around authorization accountability governance." }, { "start": 3199.16, "end": 3201.36, "text": "And most importantly trust." }, { "start": 3202.42, "end": 3206.66, "text": "you know can i trust the agent but something as sensitive?" }, { "start": 3206.86, "end": 3211.22, "text": "As my payment information and credit card information and personal data?" }, { "start": 3212.14, "end": 3216.779, "text": "how what does it take to build that trust, earn that trust?" }, { "start": 3216.88, "end": 3220.08, "text": "and certainly that is something that cannot be done overnight." }, { "start": 3220.4, "end": 3243.08, "text": "It takes time And it takes a collective effort as I was saying earlier to tackle some of these open questions That will hopefully enable that trust so we can get to state where We absolutely feel comfortable To share our payment credentials with an agent and unleash the agent who goes shopping for us or you know carry out whatever payments we have in mind." }, { "start": 3243.68, "end": 3250.62, "text": "So I do think that twenty-thirty might be, you know how should i say?" }, { "start": 3251.54, "end": 3262.94, "text": "You know...I can't say like how long it takes to build trust but certainly for an illegal and the domain of all these open legal questions It will take a lot more time than just a few years." }, { "start": 3263.5, "end": 3264.04, "text": "I would argue" }, { "start": 3264.779, "end": 3266.18, "text": "Okay, totally understandable." }, { "start": 3266.9, "end": 3274.759, "text": "We try to nudge and trick you into some bold hot take predictions but totally understandable that it's almost impossible if not impossible." }, { "start": 3274.86, "end": 3277.14, "text": "so let's just do this too hard-to-do." }, { "start": 3278.94, "end": 3285.02, "text": "Yeah I mean the space is like as Eunice was saying in a beginning still very nascent." }, { "start": 3285.06, "end": 3287.86, "text": "we're still at the early stages." }, { "start": 3288.74, "end": 3296.12, "text": "quite frankly in the It feels a bit uncertain and strange." }, { "start": 3296.16, "end": 3317.36, "text": "And trust doesn't come easily, we just need to keep at it as the industry and policy makers like both public and private sectors to kind of join efforts and make sure that we're exploiting best possible outcome for everyone in building their trust." }, { "start": 3317.46, "end": 3324.54, "text": "so yeah there's no... We can really put you know, exact prediction on when that will happen." }, { "start": 3324.6, "end": 3334.22, "text": "But I think if we all together work on it and like we have human ingenuity in the mix We'll be able to accomplish something within." }, { "start": 3334.98, "end": 3339.12, "text": "You know next couple of years perhaps there will be more of these guardrails." }, { "start": 3339.16, "end": 3344.9, "text": "There will be clearly defined parameters Within that layer too That i spoke about." }, { "start": 3344.94, "end": 3357.06, "text": "they will make It a bit easier for people start adopting that And then we can build a trust over time, but how long that will take exactly it's really hard to tell." }, { "start": 3357.72, "end": 3358.24, "text": "We shall see." }, { "start": 3358.319, "end": 3362.72, "text": "maybe in our next podcast there is more to report." }, { "start": 3363.24, "end": 3364.44, "text": "Yes excellent Sonya!" }, { "start": 3364.56, "end": 3370.54, "text": "We really enjoyed this conversation with you about Acentig AI for multiple reasons." }, { "start": 3370.9, "end": 3376.1, "text": "first because it kicked off our thinking at BFRR on this very topic." }, { "start": 3376.18, "end": 3377.279, "text": "so more content." }, { "start": 3379.64, "end": 3383.86, "text": "Thank you also, Sonja for sharing your knowledge in the recent paper." }, { "start": 3383.94, "end": 3399.759, "text": "How a genetic AI will reshape payments and then of course to the audience for listening, and making it until the end of this very episode." }, { "start": 3400.22, "end": 3403.2, "text": "If you enjoy our podcast please recommend and rate us." }, { "start": 3403.4, "end": 3405.5, "text": "that always helps us most!" }, { "start": 3405.799, "end": 3412.34, "text": "Do connect with us on YouTube, LinkedIn & our expert community on Telegram both in German & English." }, { "start": 3413.14, "end": 3418.5, "text": "And Sonja if thats not already the case Of course your invited too." }, { "start": 3418.94, "end": 3425.16, "text": "join our institutional DLT community where, for example this episode will be shared." }, { "start": 3425.54, "end": 3436.12, "text": "And to everyone make sure to subscribe to the podcast wherever you get your podcasts and join us next week when we'll be exploring another critical development in the evolution of digital finance." }, { "start": 3436.64, "end": 3442.06, "text": "until then This is Bitcoin fiat rock'n roll bringing Europe's perspective on digital money and digital assets" }, { "start": 3450.22, "end": 3452.4, "text": "and are provided for informational purposes only." }, { "start": 3453.279, "end": 3468.02, "text": "This podcast does not offer financial or investment advice, always conduct your own research." } ]