The Secret Problem With AI Agents and How To Solve It — With Guy Tytunovich
Channel: Alex Kantrowitz
Published at: 2025-03-31
YouTube video id: Ni8BPy9uK6o
Source: https://www.youtube.com/watch?v=Ni8BPy9uK6o
let's talk about what happens to the internet when it fills with AI agents and how prevalent are these robots already we're joined today by Guy ta tunich the CEO and founder of Czech who is working on the front lines to solve the secret problem with agents and he's here in a conversation presented by check guy great to see you welcome to the show great to see you Alex thanks so much for having me it's great to have you I want to just go to the prevalence question first because we talk about AI agents all the time here and I don't really have a sense as to how prevalent these agents actually are you're looking at traffic from Agents all the time so can you give us a sense as to whether people have actually deployed these already or whether this is more marketing than real life of course so I started check about eight years ago and we actually originally called it check. it's also our domain because we wanted to check your artificial intelligence we had no idea when you know AI agents would come especially eight years ago or that chat GPT and open AI would be so proud but we were hopeful and what we started with is looking at Bots Bots the old school bots are also AI they're not gen AI they weren't working like AI agents are today but essentially those Bots back then and still today are between 30 to 50% of the internet population what's different now is that in the past those AI tools or Bots or automation tools were um almost always malicious it was very easy to understand that if a bot comes in or if a bunch of bots comes come in to your website it's not because they have a b High School reunion they're not coming there to high five each other they're there to do something nefarious in most cases but we believe that we are now in the early stages of the most significant digital disruption since.com boom uh we might be a bit biased but we certainly believe so AI agents and algorithms increasing replace what was once possible through genuine human interaction and what this means is that now Bots or AI agents are not just going to be clearly nefarious in most cases I think Gardner who I like to quote say that up in that by 2028 or there about about 30% I might be misquoting but it's it's roughly there about 30% accurate yeah directionally accurate exactly um is going to be a driven by by AI agents now Bots or AI agents are going to be able to do positive things they're going to be able to book an appointment with your Barber on your behalf cancel an order on Amazon buy shoes now the prevalence of those kind of tools of the positive Bots is still very small but we believe that once open AI releases ched the operator for instance from the 200 bucks a month package to the regular paid package and once Opera who just announced they're launching an AI agent of themselves within the browser once that happens once Gemini and Google goes there once entr Tropic goes there we do believe it will be incredibly massive and again double digit percentages of the uh positive internet economy so 50% of the internet's traffic today if I heard you right is already bought but a lot of that is bad stuff and you were expecting we're going to see a much larger increase in terms of good Bots hitting the internet uh doing things like completing tasks for you like operator does today exactly and today AI agents tiny percentage are they even a single digit percentage of web traffic yet or what's your what's your read on that not even remotely close it's way less than 1% It does vary and we do see for instance we've got one customer who's a B2B Marketplace think eBay for machines and printers and so on and so forth they have about 10% of their traffic of their positive traffic coming from AI agents but they are an out a massive outlier we're still seeing with massive massive customers that receive billions of visits every month we're seeing few hundreds of visits from AI agents we again we believe it will become way more prevalent and we believe it's a question of when and not if and that when is We Believe within a few months right yeah I signed up for the $200 a month plan from open aai to be able to try out operator and I I unsubscribed I I think it's useful uh but I definitely am going to wait uh it's like $20 a month of value for me not 200 uh so I will wait until it comes out and I expect a lot of people will be using this um as it becomes more available like you're saying and then it's going to create a very interesting situation for companies companies selling online companies who are doing anything online because I guess they've been filtering out all bot traffic thinking that it's you know maybe I don't know I'm going to throw stuff out there dos or some sort of AD fraud they don't want bot traffic uh but now they're going to want it because it's going to be agents working on behalf of their customers and well that's going to be an interesting thing for them to solve because you're not going to be able to figure out who's who or maybe you Willa so kind of talk us through a little bit about how companies and how you're thinking about this I mean do you have to like have your own terms like a machine customer for an AI coming in tell me a little bit about how that works first of all we believe a check that it is um a situation that's going to offer a lot of problems and with a lot of opportunities obviously we believe that Commerce is going to grow because of it that human productivity is going to grow because of it but we also believe that problems are going to erupt both because of how you know Bob management platforms and identity and access management platforms have been operating thus far which is completely irrelevant for the world of Venta Ki and and we believe that the solution for it is to be able to um understand not just whether a user is a human being or not but pair it with and that a a continuous get mouse game but also pair it with the ability to understand the identity of the user whether this AI agent is indeed operating on behalf of who they claim they operate for is this the AI agent of Alex or is this my AI agent trying to mask itself is Alex logging into Alex's bank account a and even more so the subject matter expertise you need to understand the world of the CMO of the chief digital officer of the customer journey in order to be able able to play in the world of agentic AI and to help companies really figure out what they need to do both from the customer Journey side of things and the digital experience side of things as well as from the security side of things you want both to not just a not block or a enable positive AI agents good AI agents you want to even complement that with a digital experience that fits them it's not anymore the digital experience world of you know this bot needs to be green and needs to needs to be yay big and it needs to be here and not here in order for the human being to be able to click it and to notice it now the AI agent is a completely different being and it's still a being Not a Human Being but a being and it has its behavior and there's still a point in optimizing your website or your app in order to cater for that user because it's a good user now this world is going to change and at the same time the security side of things or the anti fraud side of things is also going to change because the opportunity that AI agents create for cyber criminals is immense and currently very few companies are able to not just decipher whether user is a human being or not whether it's an AI agent and so on almost no company knows how to decipher the intent and the identity again is it really Alex's AI agent or is it the criminal's AI agent trying to m ask itself is is Alex right and very quickly I mean people have been talking about how uh AIS are going to maybe find their own more efficient way of communicating and maybe create new languages to communicate uh information to each other and we haven't not yet seen that from generative AI but I wouldn't be surprised if we saw it but what you're saying here is kind of interesting because um if you are a human user you're right you need some signals you need human language basically to be able to give you um you know a way to guide through a normal way to guide through a website but if you're a bot you can maybe do that very differently it's almost it's it's almost like you're creating in some way like a custom API for every visitor because you're again you're machine to machine and you're going to have to find new ways of communicating to each other so how how's that going to play out it's exactly what it is and the answer is the real answer the honest answer is nobody knows there are a lot of guesses out there a lot of guess estimates even but no one really knows I've been doing artificial intelligence for about 20 years now and to like when I remember when Sora by open AI came out I could not have envisioned a day prior to that that AI is already there so the the understanding what where the world of AI and is going and the pace in which it's going is incredibly difficult to understand that being said it is um uh exactly what you outlined it's it could be a completely new language it could be a custom API for every specific visitor we don't know yet how many AI agents or companies that are Proprietors of AI agents are going to be there in six months a year 10 years it's a crazy world and what you need is the Adaptive and dynamic ability to always be a step of a step ahead both a step ahead of the AI agents and how they operate as well as the Cyber criminals that might use them now think let's I mean I love adobe's uh digital experience Cloud I love Aqua I love salesforce's digital experience cloud and there are quite a lot of companies operating in this massive Tam of a market they now need to be like they've invested bajillions of dollars in building the perfect Suite to C for digital experience for human being users now what do you do in a world where it's not just human beings anymore what what do you do there so I think a lot of the answers other than detecting the user and it's humanity and its verified or valid identity it's also trying to figure out that very question there are billions of dollars invent invested in building digital experiences well dozens of billions of dollar every single year now what this is the question how do you change it how do you make it enabled for non-human users as well I want to talk about that and first let's just talk about detection we'll go through the path detection experience and cyber security um so let's talk about detection first I was just uh doing a lot of research for a trip I have coming up where I needed to rent a car in Milwaukee and drop it off in St Louis and and it was a pretty exhaustive search cuz I was like thinking all right am I going to fly into a different airport am I going to fly into these airports what's the cheapest car um and I'm going back and forth I'm on kayak looking at flights I'm at hot I'm hot wire looking at cars then I'm back at kayak changing flights I'm back at hot wire changing cars now I'm changing looking at cars on kayak I think this is like a pretty normal experience when it comes to travel booking I would have loved a bot to go do that for me and and because my activity is multi-site um It's probably hard for those websites I would imagine to figure out whether I'm the human me or the let's say open AI operator version of me going out and doing these searches so how would a company even begin to try to detect the bot behavior in this situation 100% it's incredibly difficult again you need to hire a company like check which me saying this is very self-serving but that being said it is the truth there's a very high likelihood in my opinion that the Proprietors behind the main AI agents like open AI like Google would find a way to declare um an AI agent as one operating on their behalf that being said that is the very definition of a loophole such a thing kind of exists with Google's robots Rob robots.txt Google many years ago declared its spot that is a aiming at indexing side so they can appear in in Google search and it has a signature and it's called robot robots dxt and that signature has enabled cyber criminal Proprietors of other Bots to mimic that bot and G gain access to places that would have otherwise blocked them I believe that this is the same place the AI agents declaring themselves is going to go to and then you really need a company that can very clearly decipher if a user is an AI agent a regular b a human being a human being with malicious intentions or one with positive inent intention same as the AI agent whether it's really operating on the user's behalf and so on so forth and here you need to marry a that expertise with identity expert and finally it with the understanding of what's called the digital experience and the customer Journey because you need to enable granularity if you're kayak for instance that you had mentioned it's not necessarily the same behavior for every single AI agent that you want to enable it's not depending on the risk depending on the identity depending on the digital experience considerations depending on the stage of the acquisition if you're just about to insert input your credit card number there's a higher risk associated with that and could be that the that point you'd want to um Force the a agent if it's such detected as such to invoke human intervention and the proprietor of the bot to say okay hey this is indeed me operating on my behalf here's a biometric proof for it and yes I I approved this ER purchase and so on so all that identity the subject matter expertise in anything digital experience in in in customer Journey um the capabilities in human nonhuman behavior and finally the enablement of granularity and multi- actions in order to allow the sides and apps to decide what they do in every single instance I guess I'm curious whether there's going to be situations where sites say some Bots are allowed into the club and some Bots got to stay out I mean is it going to be a situation where um you know I'm using operator and kayak will be like all right come on in and you go ahead and um you know get the information you need and distribute it back to a human customer but if I'm using some you know potentially off-brand agent they might be like we don't necessarily trust you um because that's going to really create a whole new Dynamic I think on the internet in terms of um what what services are allowed to work and which ones don't without a doubt and that's where it becomes tricky SL interesting um I have no doubt that there are going to be hundreds of of different AI agents leading the pack then there's going to be ones that um revolve around open sources maybe open sources by open Ai and the like and you need to be very selective and Vigilant in terms of what you enable what you allow in what you block and so on and it's it's going to be very complex you're also going to have you mentioned machine customers earlier so the a lot of machine customers are AI agents and some are not for instance um a printer that knows when the ink is going to uh run out and at the right time it ER purchases new toner from Amazon or whatever and in those cases it's not really it's it's artificial intelligence to a certain extent but it's not really an AI agent it's a machine Customer because the machine is making the purchase now do you want as Amazon or as HP or whatever do you want to allow every such machine customer to to to to buy to acquire to purchase not necessarily it also goes to different regulations let's say you're financial institute a bank can you allow AI agents to transfer money from one user's account to another an anti-money laundering issues and so on there's so much to it but yeah a a new and very selective world of digital experience and security yeah this is so crazy because I'm thinking about it and there's just going to be a whole new genre of people buying things and then asking for refunds blaming their AI agents exactly exactly and again the opportunity companies to deal with especially the pcis yeah the the credit card companies yeah it's going to be massive and crazy I think also for philosopher philosophers by the way I me when you think about it it's a new being that's not a human being and it's got your I think agency would be the the the best word kind of to describe it and it certainly has your credit card and login details and would privacy regulations apply for your agent like it does for you as a human being for instance and a ton of headaches to to try and ready your side forth guy can you answer that question I mean should privacy regulations apply to bots first of all the answer is the world doesn't know yet but my opinion is one that I'm I'm certain about and in my opinion it needs to be exactly the same it's it's a tool operating on your behalf it's not a different uh it's it's it's got your permissions it's an automation tool so yeah I think or my opinion is and I think it's also going to be the case of of what's going to happen is the same privacy rules and regulations that apply to you as a human being users user sorry would apply to your AI agent as well yeah and I guess they'll have to pick whether to opt in or opt out to allow the cookies to track them and exact whole new era of training them exactly and by the way that's uh I I'll give you two uh stories about it from the traditional bot world first of all most Bots operate on behalf of criminals they have an aerious uh purpose and as such they do not click on the consent uh pop up is it legit then to try and fingerprint them in order to understand if they are indeed nefarious is it I mean they didn't consent so the courts currently seem and regulations like gdpr and CCPA seem to think that well they're not human beings they don't need to consent and and the second story is a and I don't remember the details exactly but there was a precedent in uh in the US where the judge decided that the bot is not is not entitled to the same protections from for free speech based on the U uh the American Constitution that a human being is simply because it's not a human being and as such you can a unmask it and prove who is the proprietor of the B say if they bmer merged a competitor or or or a person and so on so forth so they actually decided that the B is not entitled to the same constitutional rights as as the human being behind it and the same PRI privacy protection as the human being behind it so very interesting question again I think with AI agents it's going to be a bit different yeah that is fascinating um you know as we're talking I'm just thinking again about the cyber security we said we talk about the cyber security side of things um could you see these new agents opening up new cyber security challenges for companies of course so for instance there's a $35 billion T Market behind uh anything to do with identity it's it's called I am or Siam so customer identity access management and you've got companies there that are behemat that deal with identity but don't know how to cope with machine customers and and AI agents cuz they can't decipher if a user is indeed an AI agent or not so where does this world cup go to and it presents a huge opportunity for cyber criminals again around fake identities it also Imagine you you were a bank robber imagine that was your career choice but but a real bank robber in the physical world and imagine you could use a robot that look looks and behaves and uh Walks Like a real human being and that robot cannot be associated with you or it's going to be very difficult to ass associate that robot with you and you would send that robot to H rob a bank on your behalf Perfect Crime I mean if the but is caught how can you be associated with it it's it's it's crazy similar to that a lot of new uh Criminal Enterprise and criminal opportunities are going to arise from the fact that you can use a proxy basically an AI agent to to to work on your behalf to commit crimes on your behalf and much much more so how do you then work on matching the identity uh between let's say Bots customers um whoever's visiting a website so we've done Bots and nonhuman detection again originally that was the plan also to to check your your AI enhance check. a and we've been doing it for years now um a couple of months ago we acquired a company called d.com which H produces what I believe is the best identity graph uh out there it's certainly the best I've come across too across startups as well as incumbent players together marrying the identity graph together with human not human and obviously it's way more granular than that it's not just not human it's AI agents that are good versus AI agents that are bad and so on so forth marrying those two allows us to not just understand that this is Alex's uh AI agent it's also legit working on behalf of Alex from your device and so on so forth so that's how we're looking at it all right guys so we've been talking a lot about you know AI agents through this conversation and almost assuming like they're going to be a fact on the ground uh I'm just curious what you think the state of the current capabilities of Asians are because again like I mentioned I've tried operator uh is it worthwhile as part of the package for $20 a month uh from open AI for sure uh but I don't know again if it's um if it's really up to the point where I'd be using it uh every day so I am kind of curious what your experience has been with these Bots and how how you know how they perform today and what you expect them to do moving forward like are they just going to keep improving without a doubt without a doubt I've got no question that AI agents are going to be able to do every single thing that a human being can do I think and and that they're going to improve massively I don't think people are going to be able or willing to pay so much for an AI agent as they do certainly everyone who can afford a personal assistant would go the AI agent route but and I think the real questions there are not if and do not revolve around whether the technology is going to be good enough nor whether it's going to be too expensive because it's not there's a 100% chance an open AI already said it and the same's going to go with Google Gemini and and chrome their browser and many others it's going to be commoditized in terms of pricing it's going to be either free or costing a modest price kind of like the base the the the pro what is going the plus package of open AI the questions around it are more the psychological questions of whether the human being would trust that AI agent to do what it needs for that to do and across many aspects whether the human users would entrust a third party to hold their login details to their bank account would they trust them to really choose the right you mentioned kayak earlier and and renting a car and you know whether they really trust them I think history shows us that the answer is a clear clear clear yes you know my wife is a clinical psychologist and and as such she H worries on whether or not AI is going to replace her as well and we've kind of we have an argument about it but ER I think I think both of us understand that our generation at least is not going to trust an AI agent even if it looks and behaves and like a real human being to be our our therapist our psychologist but I think short of that pretty much everything would be able would would be enabled by AI agent and replaceable and trustable by human beings and so yeah fascinating stuff all right what before sign off why don't you tell everybody um what type of companies that you work with that check and if people are interested in hearing more how they can get in touch sure and so we currently have about 15,000 uh customers worldwide um of which about a thousand are large Enterprises um across every industry we work with the top five banks in the US and top five cyber security companies in the world and cess companies and e-commerce companies and and we protect their entire customer journey and go to market from anything that's perilous in artificial intelligence I think now is the time again if you've got a website if you've got an app now is the time to try and figure out how much of your user base is and is going to be made of non-human beings and again not just for cyber criminals or or or fraud prevention purposes but also even more so for the purposes of of customizing your digital experience to match the new internet world if they want to um if they want our help in doing that they should just go to check cq. and yes sign up and we'll get back to them momentarily great well guy I have to say this has definitely opened up my mind in terms of what we can expect with AI agents I think so much of the focus has been on capabilities and when operators going to be released to the public and so few people actually think about what happens to the internet once this stuff gets in the hands of let's say the 400 million people who are using open AI or we even saw a new uh agent come out uh of China Manis so this stuff is moving fast it's going to change the internet and I'm so glad that you came by and talked to us a little bit about um how company should be thinking about that how they should prevent some of the downsides and and really make use of of the best parts of this technology so guy thank you so much for coming by thank you so much Alex for having me all right everybody the website is check. cq. a and want to again thank guy for joining us and thanks to you all for watching we'll be back on the feed with more interviews soon