What If AI Could Decide Who Gets Hired? Inside RChilli's AI Revolution with Sneh Lata

this artificial intelligence it it's it's not a new concept it it was there right from a century ago so I would say like this like just look at a calculator right a simple calculator it's [music] it's also an AI machine and AI is a is a big umbrella you have machine learning you have deep [music] learning then you have generative AI within it and this AI agents they basically sit inside these generative AI so so we have [music] seen this proion and in simply if I if I'll say AI are the machine that possess [music] the human intellectual equivalence. So, it's a bigger umbrella. It's not a new concept. >> Hi, my name is Dmitri Bonichi and I'm [music] a content creator, agency owner, and AI enthusiast. You're listening to the AI agents podcast [music] brought to you by Jot Form and featuring our very

own CEO and founder, Idkin Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show. Hello and welcome back to another episode of the AI agents podcast. In this episode I have the director of product at our chili.a. How you doing today snake? >> I'm doing very well. Thank you so much for having me. >> Yeah, absolutely. Really excited to chat with you. So just to kind of kick things off, I'd love to know a little bit about um your background and kind of how you got into the world of AI. I work at the intersection of AI agents and human capital management. I mostly work with the clients and enterprise platforms like Oracle cloud, CM, SAP, Salesforce and HR techch platforms and my focus is to build AI agents that actually work in

production inside these workflows. Um as far as my journey is concerned I would like to share I did not start as an AI expert even back in my high school we didn't even attending conferences and requirements and then translating those needs into systems I remember back in 2009 the first conference I attended was at Chicago at HR technology expose and I I I think I had the quest for knowledge and that kept me going and that kept me like in a technology how it is moving where it is going and then later on I moved to United States and work with enterprises like uh GAPC Bank of the West and I spent my first decade on a on as a business analyst leading transformation projects in retail and banking. So to give you more insight of uh how we started it at uh at GAP

in San Francisco. I worked on very high visibility project uh it was a flagship transformation project called product end to end ET [snorts] which was to modernize how gap planned sourced and moved their merchandise from vendors to more than their 3,000 and all customers. So my role was to translate a business vision into detailed workflows and systems and they had this uh old legacy platform called ACT assorted costing pool which was running for a very long time and then uh Pete was designed to replace it and connect to all other fragmented system um which talks to this legacy system but still keeping it alive for the downstream application and reporting. So it was it was like a uh complex project and um it was rolled out for first year gap outlet then old navy and then banana republic all these gap brands. So uh work

working in retail and then I moved into the financial services with banking with bank of the west. Uh there I worked on auto lending products and we applicated it to RV and Marines and I designed the extension of their current single auto lending products to multiple offerings that enhanced more business for the bank and for leadership. It was called plus program as the preferred lender in US. So because bank of the west was under this BNP group it was a French entity and they were trying to build more business get more market in US. Um so today if you see took over the bank of the west u so working in the in the bank I also worked on cloud-based lending transformation progress uh focused on non-prime lending. So if you see like it's very easy for the prime scorers to get lending products right

but uh for the non-prime things get little tough so so we build a fee based model to accommodate the non-prime segments so this way bank can cover risk and lending products accessible to the non-prime scorers as well so it was mainly to build a full spectrum from the bank perspective and then uh later on I I worked at Shackling. It's an online wellness company. I joined there as part of the e-commerce modernization program and my focus was to redesign their digital workflows, integrate several backend systems, accommodate new motions, um help the teams adopt new agile ways of that work improves site stability and performance uh of their e-commerce presence. uh but more importantly it was uh basically making possible for marketing technology operations all to move in sync when they launched their new customer experiences across regions and I became the bridge between their uh

business development and because I could keep everyone aligned while they're being very ethical. So all these roles in United States they taught me the same lesson whether it's a retail brand link or an an online company transformation can only work when you connect strategy data and execution very tightly and across all these roles I one pattern the main problem was was really the algorithm but it's the fragmented distance for data unclear ownership of those when I came to RG I brought my this is exactly what I do at RG I I sit on the product side leading AI and cloud innovations that that power the systems for enterprises these platforms and job boards. So I take everything I learned in those early US projects and use it to design these AI agents that are not just exciting in demo but they that actually run in

production inside these large complex ecosystem like Oracle, Salesforce and SAP. So my focus become if you want AI agents to succeed in HR we must fix data uh the workflows and the integration with these four platforms. So I moved from being a business analyst into becoming a product leader for AI and cloud innovations in HR. >> Interesting. Okay. Yeah, that that makes a lot of sense. What do you um uh personally uh I guess my followup question would be um you know how long have you actually been in in this specific role again? >> Oh, so I joined like uh uh specifically in Archel I joined like last year and then uh for a decade like I was working in US. Yeah. Yeah. currently operating from Canada. >> Very cool. Um, what I would say is that it seems like Arch Shell is doing a

lot of different interesting things to, you know, uh, I guess the the taglines that I'm seeing, right? that you're um you're reinventing things like ERP recruiting and I just kind of want to understand from your guys' standpoint, >> how do you go about yeah >> taking a process as complicated as as some >> attempting to work reverse engineer the process so that AI can essentially uh function as that recruitment. >> Okay. So, Archi is basically u it's an AI powered HR tech company right which that structures and enriches the talent profiles uh in very simple words we turn and profiles into usable data inside HR systems. So if you see the core problem which enterprising face they have they have basically three issues the data is very unstructured it's often stale and it can carry bias and unconssistency and when I took on my current

role in arch has around 15 fragmented offerings my responsibility was to unify them into a modular AIdriven suit that's easier to use and easier to scale. So I compiled them in four major modules and that was a uh that really helped. Now we have like four main products from RJ that's recruitment AI, data hygiene, bias and recruitment hub. So uh if I'll go in a little bit uh detail of what these products do. So one is the recruitment AI that's our industry-leading parser which passes and matches talent data for your keywords and brings it in front of your dashboard in no time. Just imagine like you have heaps of talent data like in your system or you have so many res. So based on what you're looking for based on this particular uh uh university or based on your search keywords from that piece of

data recruitment AI engine agent that will search the right candidate the right profile for you in no time that that's basically the power then on top of it like we have data hygiene which removes any noise cluttering from the data and leaves you with very clean structured normalized data. So how it works? So so think you're looking for a uh for a Java person but in resume is resumeuma it's written like JS or react right. So our tool will not filter those resumeums out for you. It will roll up into the right skill families and present it to you. So how it works? We have built five taxonomies. We have uh sales job hires degrees universities and postcards across many industry and lang languages. And we use our onologydriven mapping which which means you do not just match keywords but map every skill job title

and technology from a resumeum your job description into a into a structured knowledge graph. Right? So this lets the ais platforms search match and build their agents on top of this clean and standardized tail and data of the messy text. >> Right. >> Interesting. And then the third I say that's the unbiased hiring. So uh with like in hiring you have seen unconscious bias that can easily slip into decisions. Sometime it comes from the way the resumeum is written or or on the details that stand out. So to avoid any biasness in the hiring this tool really works. So what we have built we built a reduction tool in the hiring workflow and it uses around 55 DEI related parameters diversity equincclusion or the list of values you can say to identify information. So there will be biasness in terms of gender or age lues

or certain demographics or any person really it must those data and then the hiring panel sees a more neutral profile that focuses purely on skills and experience. So we are not replacing human judgment. We are supporting it by removing any noise and bias signals from the data and how it benefits enterprises. It builds trust in the fitness of hiring process. It maps those signals before the hiring panel reviews the candidates and it's it keeps the focus purely on skills, experience and capabilities. So basically supporting the diversity goals and reduces any legal and reputational risk and it improves your and coming to the last one that's uh that's recruit recruitment hub. So imagine like you are looking to hire a candidate and then you are getting results from different channels from indeed from uh career builder from linkading and then someone sending in the emails or

you have your tailor data inside sitting in your system as well. So what this recruitment hub does we provide a browser extension and then u you will get all these uh decentralized data into a consolidated location uh within your system. So from all scattered data through this extension you will able to have a a good view of this complete uh data which is which was fragmented and now it's in a centralized location for you. So it helps you to make a better decision. So that's basically are the products which archely is offering for these tech platforms. >> And what do you think is um some of the harder types of I guess uh ways or not ways what are the harder types of things that you're having to discuss in order for people to potentially be convinced that this is like a a feasible thing?

Not saying that I don't think it is. I'm just trying to articulate that, you know, there's optimism and skepticism in the uh market right now with like the capabilities of AI. So, how what how do you go about articulating what you do in a way that makes companies feel comfortable that they should be able to take the human element off their plate and be able to kind of replace it with AI uh for this process? >> Yeah. So, uh that's a great question. So I I really when people hear I I want you to think of us as the reliable intelligence layer that powers the AI agents for tailing data. You just pull the clean data in and better decisions out and then Archley has proven record for uh making the candidate experience up by 85%. 72%. Your profile accuracy goes up by 68%. So

this reduces the deployment time from months to less than an hour which is 98% reduction in hiring process. So, so when these numbers speak like that they they shows the trust like and then u uh we have like these uh names with us which means they are there trusting archely Oracle Salesforce and currently we operate in like in 36 industry domains healthcare media HR CRM insurance investment real estate we operate in 39 languages in more than 52 countries uh have more than 88 80 1 plus customers and and pass 4 plus billion resumes annually. So these these this number speaks and then um uh our proven track record that that makes things easier and and companies do trust us and uh they have seen the results and this is this is just about the products which which are our chill is offering this is not

even the AI agents I I haven't even touched that point so these are the basic products and then if uh if I talk about the about the AI agent so we delivered uh this talent data refresh agent to so we shipped it and it's working in production currently um within the Oracle >> Okay. All right. Well, then let's touch a little bit more on that specifically then. Talk talk more to um the Asian component then. >> Yeah. So, um yeah. So, I I'll just say like this artificial intelligence it it's it's not a new concept. It it was there right from a century ago. So I would say like this like just look at a calculator right a simple calculator it's it's also an AI machine and AI is a is a big umbrella you have machine learning you have deep learning then you generative

AI so so we have seen this progression and in simply if I if I'll say AI are the machine that possess the human intellectual equivalence so it's a bigger umbrella it's not a new concept Um now talking about this u AI agents which belong to this generative part of this machine they can they can execute the tasks and goals on our behalf across browsers across applications and across data links. So in other words it's it's a series of a workflow that works right. So in in simple language AI agent if you if you think it it's it's like a driver. it has been given a car or or tools and a destination and which is defined in the context of the prompt. Now now the driver will have to get to it and has the choice to drive it in his best way possible based

on the available workflow that's been provided to it and the talent data refresh AI agent which are actually uh has shipped to Oracle. We are the proud partner of Oracle fusion cloud application a a agent marketplace. Uracle accepts only validated partner built agents that has a rigorous 21point security and functionality checklist. um applying the same standard which Oracle uses for its own internally developed agents and arch uh we are one of the flagship agents uh in terms of talent data refresh agent and and this what this agent does actually so uh this automatically scans existing talent records that's the first thing then it enriches them with [clears throat] fresh structured validated data and then it updates the skills experience and attributes uh depending on what the need is [snorts] and it sets that updated data back into the Oracle SEM in a very compliant way.

So so uh just think like it's an it's a next generation AI powered profile enhancement system which is built inside Oracle AI agent studio and it's integrated within Oracle Fusion Cloud SC. So it autonomously identifies missing or outdated information in candidate profiles and enriches them with verified realtime data from external sources and platforms. To make it uh a more simplify I would just uh divide it into steps like so first what it does it has AIdriven gap detection right what it means it identifies any missing employment history education or skills then resumes or talent profiles people change their uh skill they they keep upgrading like like never before so it it keeps on changing and then second is the parsing integration so this talent data refresh agent it leverages Archile's industryleading parsing API to extract structured data and standardize the candidate in then it uh

connects with the external sources for validation and real-time updates in terms of data enrichment and then we do the data normalization on top of it. So I'm I I see I'm I'm getting maybe too technical but these are the things which this agent is capable of. Yeah, this this ensures this consistent consistency across millions of candidate profiles across uh these standardized taxonomies I mentioned earlier like across job titles or JS or JavaScript or so it just roll them up and then uh remove any clutter or noise but the last one is the decision intelligence which which means the decision is still in the hands of the recruiter. It's it's a it's a human-based humanentric u tool which improves the recruiter decision making through precise bias-free up-to-date candidate insights. So um you can say recruiters can search on what is true today not what was true

3 years ago. So it doesn't work on the on the stale data. uh this agent can give you can enrich the profiles with the the fresh data of the candidate better search results um more trust in the data. So this is one example of an AI agent that that quietly improves the quality of the tail end data in the background. >> Interesting. Okay. Yeah. By the way, I like that really u I like that beginning part where you you mentioned that calculators are AI. I think to be to be honest that that hadn't quite uh came to mind for me. Um it's not quite um what I uh would imagine, but I I can see your point. I think to a certain extent that's true. I think that it's kind of a framing thing. For a long time, AI has been kind of a a

term that has been now synonymous with like large language models, but that doesn't mean that it's the only thing AI is, I guess. >> Yeah. Yeah. Those were the part of the like the machine learning and now as I said like we are going into the deep learning images and like last week you know the Gemini rolled out its new uh tool Gemini 3.0 pro with the nano banana pro. So we are just we are just progressing these uh these AI and these agents they they are they augmenting us they are elevating the enterprises especially like look at the small and medium businesses they they are becoming like undistinguishable they they are looking like large companies they are becoming larger and larger with the with the a with their AI agenting teams and I think the question is what will you do if you have

like unlimited number of resources and you are no longer limited by the number of employees where would you go If you have like unlimited traditional human resources, right? This is this is this is your moment. What do you want to do? That's the question with the help of these AI agents. >> Yeah. No, that's um that's yeah, that's a good point. And where where would you say is kind of some of the um biggest misconceptions that you kind of have to overcome when explaining uh what you guys do to um whether it be uh potential clients or in marketing materials? It's like speaking of you know it being like even calculators can be AI. What's kind of some of the misconceptions you have to overcome when trying to explain what you're doing and what AI can do in general for potential clients? >> So I

would say like u uh AI agents and humans they both have to work together. This is the the the bus has left the station like they are already moving right. AI agents are there in the space right now. So what I would like to start with like think simple things simple uh stay close to uh what you want to achieve like start from the workflow who is the user which system are they in what decision are they trying to make at that moment so uh just work on the fundamentals and then second you fix your data foundations so in terms of like uh human capital management uh where I work so work on your job titles, skills, locations, work on your data foundation, your tax families, get them uh consistent and then your AI agent will work on them so they can automate um better

results not like they are automating confusions and then in our we invest huge in taxonomies and multilingual normalization before anything else. So if your customers live in Oracle, if your customers live in any other platform, your agents what you are building, they should feel like first class citizen there. So I really would suggest to work on the fundamentals and then and then um I strongly believe like working on the on the governness from day one. So that's a very important part and when I talk about the responsible AI, I think of like eight pillars of the uh of this responsible AI. So the first one which comes is the is the compliance part. So whenever you're building anything just make sure it's legally compliant. In arch what we see we are like fed ready which is which is federal risk and authorization management program and

it's a mandatory security certification for any cloud service used by the federal government and this achievement itself opens our chill pathway to federal contracts so supporting US government hiring operation and then we have like uh four major AI regulatory frameworks in place we have New York City's AI hiring law CCPA CP PR, EUAI act, HIPPA, GDPR, the uh general data protection regulation. So what I'm trying to say is like if you're trying to build AI agents or responsible AI, first thing is the legally compliant. That's that's the first and the foremost thing before you even think of the technical part. [snorts] So think of the governness part, think of the compliance part. Keep that in place. And then second pillar would be the fear and the inclusiveness. Think of that part. And then the third p third third part comes up technically robust and building

secure mechanisms. There is where your technology play a role, where your workflows, how your uh AI agents are uh are working, what is their u uh outcome, how accurate they are, how strong they are and then work on transparency part, accountability part, explainable, you should be able to explain how this workflow is working, uh how it is uh you know what is the if anything comes in the scrutinies, you should be able to explain and Um and then comes the humanentric part. So um you know the decision would always be and still be in the hands of the on the humans right. So the humanentric part is very important. There should always be left a loop for the human to come in between and make the judgment which will be driven by empathy. And I strongly believe uh technology should be led to humanity. That

that's what the a that's what the aim is. [snorts] And the last is the environmentally sustainable. So I I strongly whenever I build products I I work on these eight pillars of the responsible AI uh making energy efficient models and um I think this is the this is the thing which we should not overlook. That's my advice to people who are building agents or who are who are moving in this direction. And if you follow these steps, AI agents will become less like science experiments and more like reliable team members in your HR stack. Because see this AI agents, they are they they have this accuracy of 92 or 93%. They are good but they they are not better. So the humans are the one who are still controlling them. They are the one who are guiding them. They are the one who are leading

them and building them for humanity. That's what I believe. >> Interesting. So you know I appreciate the comments about the human in the loop. the humanity aspects. What are you guys doing at your own company to make sure that you kind of like stick to your guns on that uh sort of stance and make sure that there is human in the loop capability for the AI agents and um that you know the empathy can be expressed there as well by individuals. >> So as I said like the talent data refresh AI agent which we have uh shipped to Oracle. So in the end like we have set all these workflows mechanism in place. It just uh it has the this uh AIdriven gap identification. Then it does the parsing on its own. Then it does the data normalization and then it uh uh stores the

validated data from uh from trusted external sources and then it present everything in front in your dashboard in front of the recruiter and then the recruiter has the power either to accept it or to override it or or you know whatever uh he decides to do at that particular moment. So, so that's the human loop uh in the place and and I would strongly advise whenever we build AI agents. Uh we need to keep this loop in place and this is what we have done. So in and the recruiter has the yay or the no say whether he wants to move ahead the final pressing button would be in his hand. Right? So, so that's that's like you know the you know repetitive task or the or the low IQ things we can get them uh we can get them solved by these AI agents

and then the final decision the empa empathetic decision would be given in the hands of the human in our uh industry it's in the hand of the recruiter if it's in another industry you have to see right but in the And the decisive factor will always always be the hands of human. >> And that's a I well I think that's fair. Um because I I usually have some follow-up questions, but it probably is pretty obvious where you stand on this. Um, I have some questions more regarding like the hot button issues of AI, but you seem to probably be in the stance that there quote what quote sorry there won't quite be a sort of overtake of AI kind of completely doing the majority of work um for people um or do you maybe take the stance that a lot of the work will be

done but maybe there will always be the human component to make sure it's in check because everyone has a different opinion on like how jobs will be impacted and whatnot. what's your kind of take on on where this will play out in the next 5 10 years? >> So see as I said like this AI this is augmenting us it's not to replace human but it is basically to uh keep the repetitive task the low IQ task uh this going to be done by the AI a much much much better way. So in a world where you could automate everything, it's it's a question that that you're going to answer at some point soon like on a business and on a personal basis. What do you want to automate and what do you want to keep on your own? What one person can do today?

It's it's about 90% more than what one person could do in 1995. It's almost 400% more than what one person could do in 1960. and it's about a million% more than what one person could do in 1,800 on a GDP basis. So on a global domestic uh production basis if you see we are ripping every nation or or every uh every country and coming on another prospective every enterprise we are delivering more and more the production has been increased. We are so impressive in terms of our ability to create value and thanks in large part technology and and that's very exciting because it means we are going to create a lot more and everyone who are listening to this podcast is is believing is is going to have this opportunity to do more than what your great great grandparents could have fathom. So the direction

with where we are moving it's it's from it's from uh uh you know they are elevating us they are augmenting us we we could more we could do more and we moving from single agents even to the ecosystems now so right now you see very specialized agents a sourcing agent matching agent a reduction agent a refresh agent a pre-screening agent we coming up with more agents as well now over time these agents will start talking to each other. We are moving towards agentic enterprises. Agents will prepare data and insight and human will focus on judgment and empathy. This AI is augmenting us as an enterprise. It's we just going up and up and up like like Billy Wong and touching the ceiling. >> The question is is what do you want to do when I when you are no more limited by the traditional number

of human resources? Seriously, think about it. Like our human brain, it it uses just 20 watt power and it runs 86 billion neurons. It's like 20 watt meat computer. And uh when I see Oracle, it's building a data center consuming 1.2 billion watt power, which is enough to power like 1 million American homes. So, so AI brings like massive computational power and the human brain remains remarkably efficient. So AI plus humans together. It's a team now. We cannot work without them. It's it's a team. We we have to work together. And then we are more powerful uh like never before. As I said like AI agents are good. They are 92 to 93% accurate but but they are not better. I would say AI is the brush but human potential is the masterpiece. So it's it it's it will make us better better recruiters in

my uh industry. It make us much better scientists, much better engineers, much better teachers, much better surgeons because they have better eye hand coordination movement and they are microscopic. You can easily see and distinguish where cancerous cells are ending and where healthy cells are starting and it can scoop out those parts. So we have never built a tool like um anything like this. Seriously, AI is the most powerful tool ever created, but it's still a tool. still a tool that enhances human capability. [snorts] So I think I think we will move from a world where you must learn how a machine works in order to use it to a world where a machine is designed to know how you work. So if you look at these new um uh charged to know how how I work, how you work, right? This is this is where

we are going. See and then and then at our we are building this technology that adapts to how HR professionals work not the other way around. So we aim to be the intelligence and data layer behind these agents. So if an enterprise is uses Oracle, Salesforce and multiple ATS tools, our goal is to make their tail end data their AI behavior should feel consistent everywhere. So, personally, I'm I'm excited about making it a boring, reliable, something people trust and use every day, not something they are afraid of. What do you think is your favorite kind of uh personal tool that you use on a daily basis that helps save, you know, the nonsense of your time um from being a thing? You know, like everybody everybody's got uh something in their job that they used to do prior to AI that now uh practically speaking,

they don't really do much besides tweak or they they automated it completely. What's your favorite um thing that got AI out of your own work? Yeah, I I use AI tools a lot and then seriously it it augmented me. I am able to deliver a lot more than I what used to be like um a year or two ago. I use use chat gt I use their paid version then I use claude as well their uh that's very fantastic when I have to work with the large documents so I use claw at that time and then uh charge I use a lot I use Gemini for for images they this uh new nano banana pro that's fantastic I I haven't seen such revolutionized product earlier seriously so I use all these uh tools and u just a tip like for listeners as well what I

used to do in my daily work like I there's one toggle it's there in charge GBT it's in broad as well so in their settings and data controls you see they by default it's it's it's on improve the model for everyone and in the privacy allow the use of your chats and coding sessions to train and improve anthropic uh models so I personally when I do my personal work I use that toggle off so because uh but when I'm doing my um my my office work or my enterprise work, I definitely let them in to that chats and then see what I'm doing because what I'm doing is they have access to my data, right? if I'm keeping that toggle on. So I can do that when I do that my um my office work but when I'm doing something uh uh you know my

some very uh personal work or my anything which I don't want to let go out so I use that toggle off. This is one thing I would uh want to let listeners know that it's by default open when we are using these AI tools and it's your personal preference and your choice if you want to uh let that model improve on your uh on your um banking communication or your personal communication. So it's it's a personal choice but this is something which is there and I want you to know that if you want you have the ability to uh make it on or off depending on uh the way you want to uh let your information out to those systems. >> Yeah. No absolutely that makes that makes sense to me. So, um, just to kind of, uh, get a couple last questions before we

do close the episode out, out of curiosity, when, uh, was it that you were kind of working on AI and or with AI that you kind of realized personally through an interaction, wow, I can I can really do way more work now. I know you mentioned like some examples now, but I'm saying in the timeline of AI history, right? All of us had that moment where we first used AI where we were really blown away. For me, I thought chatbt 3.5 was cool. I wasn't like exceptionally blown away with it because it was pretty rudimentary. What was that like moment for you when you were using AI and you said, "Wow, this is this is incredible." >> Yeah. I I very vividly remember it like it was when open AI came. So earlier we were using like as I said in broader term we were

using AI like when we were using these calculators or this face recognition on our iPhones but uh when I see this yeah when I when I see this open AI came with their chat model that was kind of like mindboggling yeah so because Google was doing just the search right it's it's just work on the rag system it's [clears throat] just retrieving the documents and then just it's just getting the information to you but when I see this uh I think it was in uh uh way back around 6 years ago and uh that was the moment when I say like wow this is this is something which is which is a game changer and when then I start using it and slowly slowly and now we see like it's it's it's taking everything it's it's everywhere now it's even I don't know if it

has enter uh yet but uh apart from that it's just everywhere people are using it like like anything and and uh I think we need to be very um very responsible when saying whole thing that we have to lead them with with lot of judgment with lot of uh with responsibility because they are capable of uh of doing a lot. So the question is the the real question is not what we can do or what we can build but the real question is what we should be doing with this and what we should be building with with these kind of uh tools. So this is I think a personal uh responsibility which I would say every enterprise or or every user should take into account when they when they work with these tools and when they build their uh agents or the capabilities to keep

in mind like you know it should eventually be for humanity because because they can take over all these um tasks and they can they can uh do it in a much better way but but we need to steer them. We need to direct them and this is something which everyone should consider when they work with these kind of forces. >> Yeah. No, that that makes a lot of sense. Um I think there is always a lot of things to consider and you kind of covered off on a on a good amount of them. So I really I really do appreciate that and I I liked a couple of the call outs that you made about new models. I really like Nana Banana as well. I think Gemini 3 was a incredible leap forward um in capabilities across all the boards and then funny enough Opus

4.5 comes out a week later. So the [laughter] like we we just continuously are seeing these massive improvements. Um one final question I did have about your product that kind of is catered towards more this model type of conversation [snorts] um with your AI agents themselves, right? Obviously reasoning comes into agents uh you know that kind of seem to be a correlative um expansion of capabilities. Once reasoning got better with uh models like with 03 and chipt earlier this year I felt like a lot of different um agent capabilities became possible for you. what really stood out as kind of the moment improvement wise with AI models that where you really started to see some large level reasoning and just overall capabilities improvements in this last year. Was that a model specifically or a time of year specifically where you were like wow I feel

like we're just getting much better outputs than we were previously? Oh yeah definitely definitely the way they have been uh designed or was giving output earlier it's it's in in a much better way that we see today and these large language models they are they are really kind of like improved a lot but again like these models have been uh have been designed and they have been cultured on the public data. We are still not there where uh they have been uh uh worked on the private data. So like as I said like I keep my toggle off when I work on some uh one my uh personal uh data but uh this is what the this is what the ask is right now from these LLM models because so far they have been trained on all public data but these uh models they are

looking to work on the your private data as well because that's when they can actually make decisions for uh um you know when when it is really really required. ired. So we need to be very mindful uh what we are exposing to them, what we are not because they also want to uh be ahead with what they are doing and what they are capable of uh giving to the giving to giving back to us. So, so definitely they have improved a lot but they need to work more now on the private data as well and we will see that in the near future when uh because as you know the Alexa keeps on hearing like every time to us our our phones are hearing us. So in a way we are surrounded by these gadgets and they are they are kind of like seeing us

what we are doing. They are and and then they work on this data and then they get back with the results for our benefit right. So so this is the next next phase of their improvement. So earlier we saw Google all these uh you know Wikipedia and all these documentation part now we are onto these chat models where they are working uh or they are uh you know designing the systems the way we want to work not the other way around so and the next level is when they're going to go a layer deep in right so you see my point so so that that that's the next phase we are uh we are looking forward to >> no absolutely I I think That's um that's a really good call out and you'll you're seeing it a lot with like knowledge bases and uh rules

and background that you can give agents already with the context of your own data. You know, I work in content, so a lot of the writing stuff continues to be improved the more information you give it. Um I have like a persona of how I sound when I write. Sounds pretty good. Sounds pretty close to like my story and my vibe and all that kind of stuff. And I think a lot more data personally given to the LLMs is going to do is going to do wonders um for for the outputs. Uh granted that I hope it's willing. You know, people obviously are sensitive about their data um still at this point, but I do to to some extent I kind of have an opinion that I think the cat's been out of the bag for a while. you know, like um there's a reason

Gemini is continuing to improve and improve and improve at such a fast rate when they weren't maybe the most popular out of the apps uh or out of the LLMs and I do think a decent percentage of that is uh they are Google so they do kind of have a lot of info on everybody and uh Open Eye is getting their support from Microsoft and um it's all very interesting but um you know with that being said I I really just want to let you take the floor for the last minute here and just uh plug whatever you would like to plug so that everyone can find what you guys are doing over there at our chili. [snorts] Yeah, I would just say last thing like whatever you guys are building just stay close to your end users. Uh stay close to your customers. uh

talk to them, see what their pain points are and then build accordingly because that's that's where the big ideas arises when you when you talk to them and listen to them and you want to help them because if you augment them uh make things which are which are good for them, make their work easier, that's that's where the growth lies and and that's where when everybody grows. Yeah, >> absolutely. Well, everyone, make sure to go check out everything that Snay is doing over there at RC Chile. That's rchilly.com. That's rl i.com. Thank you so much everyone for listening/watching to this episode. And if you did like it, make sure to leave a like and also subscribe to the YouTube channel. And if you're listening to it on Apple or Spotify podcast, make sure to leave us a review. We'd really appreciate it. So, with that

being said, thank you so much for listening slashwatching and we'll see you in the next one. >> Byebye.