Okay, I am joined now by Agus Huerta, the SVP of digital innovation at Globant to discuss kind of all sorts of things, but specifically I know a recent agreement that you signed with Anthropic, but Agus, first of all, thank you for joining me. >> Thank you, Victoria, for inviting me, too. >> Absolutely. So, first, why don't we start kind of with the basics? If you could, tell me a little bit more about yourself and about the company. >> Sure. Well, Globant is is an Argentinian company.
I mean, it was born in Argentina in 2003, so 23-years-old company. I have spent here 17 years already. So, I could say I have been uh pretty much on most of the growing phase of the company. Um I have had many different roles. Right now, I'm uh the SVP of digital innovation, as you mentioned, and as such, as part of my role is, you know, understanding what are the trends that are coming in the technology world and how that would affect and transform the way we deliver services.
In particular, one of the areas of focus right now is, of course, AI and the GenAI adoption. Uh so, in that sense, I am very close to many of our partnerships with AWS, with Nvidia, OpenAI, and Anthropic, uh which is the topic of today's session. Um but I'm also working with the teams uh doing some of stuff on some other technologies like robotics and quantum computing, too. >> So, you know, you've got your plate full, I think it's probably safe to say. >> Yeah.
I don't sleep too much. I read a lot. >> So, let's maybe dive in on the AI work, and like I said, we'll talk specifically about Anthropic in a moment, but I know Globant has built a an AI pods model. How does it kind of describe it to support your customers as they work through AI adoption? Could you tell me a little more about kind of the structure of that model and how generally speaking you work to enable your customers on their AI journeys. >> So, our original bots model, which were like interdisciplinary teams confirmed by humans, were born almost 15 years back now.
Uh 8 years ago, when I was leaving our AI studio, we started our first initiatives toward, you know, starting to apply AI in different ways to accelerate the way we deliver services. And from that, it came many of our initial products like Augur, uh Magnify, that were focuses on understanding code bases in the case of Augur, in the case of Magnify, improving the quality in the deliveries that we do by automation part of that uh through the usage of AI.
Now, the evolution of all that came, of course, with the availability of these super capable models that are the LLMs. And as we're starting to build a foundation to enable the whole company and also enabling future ideas, products, and services to be built upon the power of the LLMs with Global Enterprise AI, which is the product that we started developing early 2023 as soon as the LLMs were available as an API. And then it has turned into like the basis of what uh right now enables our AI bots to deliver on these modeling, which we are having agents running under pretty fine-tuned workflows that we have, leveraging skills that we have pre-built already for taking some of the most challenging task, and having human supervise the output that we get from the agents.
Yes. Uh our uh first version of the AI bots framework uh operated on top of Coda, which was an agent for development, testing, and it covered mostly every phase of the software development life cycle, but was sitting on top of global enterprise AI. Now, the newer version of the framework that we launched earlier this year allow us to start leveraging not only our coding agent, but also other coding agents available out there. And that's where our partnership with Anthropic, for example, comes into play in there.
Yes, we can now use Claude Code as part of the different nations that we have available for running the services. And so we are capable of, you know, leveraging state-of-the-art agents and state-of-the-art LLMs that are available in in the market. And also that help us a lot more adapt to the text stack, to the technologies that our customers have already chosen. If you If a customer is already customer of Anthropic, we don't need to go and, you know, force them to accept the usage of Claude by our teams to deliver, but rather we can keep on working on on their Claude Code instances, for example. >> So maybe now is is a good opportunity to dive into Anthropic a bit more specifically.
I know you've recently kind of signed a new agreement with the company. You're, I believe, the largest kind of Latin American-born company to be named a global preferred partner in Anthropic's system, which is no small feat, I'm sure, to be able to say. So, let's dive in a little bit to kind of how you've collected that relationship, how you've built that relationship with Anthropic, and what excites you about now being able to move forward as an Anthropic partner. >> Sure.
So, as a company that always strive to be on the edge of technology, we have been working with technologies that Anthropic has been provided since they were made available. Yes, different type of licensing, maybe at the first using some individual licenses to understand the technology, to try it out, to test how it will, you know, cooperate within our AI pods framework. Then maturing that into uh you know, enterprise-wide concept and having enterprise-wide licensing for uh the tooling.
And once we were ready on those conversations, we started exploring the fact that, "Hey guys, we can help you with your customers driving adoption. We are customer zero. We are already doing it internally at Globant. We can teach others how to do it. We have all the practices are needing there. We have very strong knowledge on everything related to AI first, but also software engineering. We have practices around uh more like consultancy services, what we call our business hacking team that understands our customer business and tries to hack it and find new ways of creating revenue streams or optimizations.
We have our cultural hacking team that is about transforming the culture of a customer. And in this sense, when we are talking about AI, the adoption of AI implies heavy changes on the culture. Driving adoption implies doing change management, understanding where the pain points are, driving uh some users to become power users and in in some way internal evangelists of adoption of the platforms. So, we have like already the full tool set to help drive the adoption.
From Tropi again, again, we are customer zero. We can, you know, show and tell about that because we are already leveraging. And that's how the conversation started, and both parties were uh pretty much aligned on the interest of working with each other. They fall in love with the concept of the AI pods and how that will transform the way of delivering, you know, uh software. Um and well, basically, now we are already partner. >> When you think about the way that kind of becoming a partner, but to your point leveraging Anthropic kind of before this formalization, if you will, of the relationship, how does access to Anthropic's models and now kind of some of the other resources you've unlocked in the partner network impact that AI pods model and how you're ultimately bringing success to customers and and what do you think you can continue to do now maybe that you couldn't have done prior to the relationship? >> So, um, for example, I mentioned before that we have global enterprise AI.
Global enterprise AI became the factor standard for the company to access LLMs. And through that platform we were able to connect to all the state-of-the-art LLMs that were available there from many different providers and even models that are available in more than one way we have them available across. So, our own go to agent was already leveraging Anthropic models before Claude code became a reality and and widely available for everyone. Yes. So, we already knew about the capabilities of the models and what they were able to do.
Uh, we had some limitations first. Our AI pods model is based on the fact that we charge the customer by token supervised by our people. So, we need to have a very good control over the tokens that are being handled. And that's why at first we started with our uh, agent because that give us the full visibility that we were looking for and interacting directly with the LLMs. Now, Claude code, for example, has evolved as an agent and now it gives you a lot of information about what's going on with the tokens.
So, since we right now are able to also capture all the interactions that that agent doing with the LLMs and the token consumption that is going on. We are able to, you know, enable it for our model and allow us to build by token supervised. So, once we see those capabilities were available, we started testing them and trying them out. And once we were sure that that would fit with our new uh AI platform subscription model, it when we you know, made the final decision of, "Hey, let's go ahead." What this brings into the table for us?
Well, first of all, it's not only us now saying, "Hey, Globant has these capabilities." is Anthropic also validating that. Yes. Some of the things that you may have noticed on on our announcement is the strong commitments towards having people certified on Anthropic's technology. And yeah, being completely honest, uh in terms of the knowledge that Globant has, I am fully confident that we are smart enough and that we're knowledgeable enough to deliver.
But the certification being provided by a partner is a way of credentializing you in front of a customer by a third-party more objective authority that says, "Hey, this team is ready to deliver." Yes. It's the same approach we have followed with many of our other partners like AWS, GCP, Azure, you know, all the hyperscalers have certification programs, uh Salesforce, SAP. All of our most important partners have these kind of programs. And again, when you go and you are uh creating a new relationship with a customer that you may have not worked before or even with an area of a customer that you have never worked on before, coming with the credentials that a third-party recognize of your capabilities is a strong play.
And and that's part of what we are also getting through uh this relationship seems the uh in particular the cloud certified architect foundation certification from Entropik can only be accessed by partners. Yes. So, not a single individual out there could get it just by, you know, doing the exam. Uh. >> I'm curious, too, August, within kind of the explanation of Globant's and and everything that you bring to customers, I'm sure you spend quite a bit of time talking to customers about what they want to get out of AI, what they expect to get out of AI.
How do you see kind of conversations on the understanding of what AI can deliver evolving over time? And and how do you think that maturity cycle will continue through the rest of this year? >> Well, uh, putting it in actually the same concepts that Entropik push forwards. Yes, they talk about, you know, like three different pillars. I will just focus on the first two because most of the companies are like struggling to get through those. Uh. Pillar number one, or P1 as they call it, is people.
Yes, it's getting the people get in touch with the technology, get it enabled on leverage the technology on a daily basis, uh, get familiarized with how interacting with an agent look like, uh, how interacting with an NLM look like, how do I can create by myself some automations on my daily work. Then you go into P2 or pillar number two, which is like the processes. You start deploying agents that are operating behind the scenes that may be triggered by, I don't know, an invoice getting into your company or by a customer reaching out with a, I don't know, an issue.
Yes, or a vendor reaching out for some information on their payment order or stuff like that. And then an agent, you know, taking that task, interacting with a human whenever possible and needed, and then getting back and and managing that loop. So, where are many of the companies right now? Well, many are still on the P1 and some getting the people enabled and and that's pretty much uh of the stage of many of of the organizations and it's it's really very valuable.
Yes. Uh because something that is different and this is like part of the conversation that I am getting with customers is like when you thought about adopting I don't know, ChatGPT Enterprise or a Claude Enterprise a year and a half back, you were talking about a solution that will allow you to interact with an LLM. Yes, and maybe you will have some, you know, capabilities for creating some assistance that will be pre-prompted and and so you can get some information outline faster and people are waiting to, you know, repeat a prompt time and time again.
But now, actually, all those platforms are great excuses to start deploying agents into your company. Just by deploying Claude Enterprise, actually, you are deploying the capability of interacting with an LLM uh in different ways like through a direct chat or creating a project that can be shared with many people, adding knowledge to that project, interacting with that information, but also you get at least, you could say, two uh agents already deployed.
Claude Co-work and Claude Co. And the line them is the same, I would say, for the lack of a better term, multi-purpose agent that is in the end Claude Co. Uh but you have these two suites that are intended for different purposes and different roles within the organization. There are agents that will be executing for you. Yes, and if we are talking about Co-work in particular, you're going to be getting an an agent deployed in your organization that can interact with your files in your computer, that can interact with your, I don't know, Google or Office Suite and interact with the documents that you have been creating in there.
You can, you know, put it as a cron, as something that runs on a specific dates and times and creates information for you, summaries, briefings, uh you know, it prepares you for the upcoming meetings that you're going to be having, briefing through, for example, I don't know, the previous emails that you have had or previous conversations that you have had with that individual. So, right now those um platforms are great excuses to get the first agents towards your organization.
And once you see the agents working, once you see the value that agents can bring to the organization, it's easier to jump into the conversation of, "Hey, why don't we do this with uh your actual processes?" And help the people, you know, go through their routine work faster with a higher percentage of confidence. Uh I I I was talking just before jumping into this meeting about an internal process that we also have. So, hey, how how do we deploy an agent in here that can help us capture how the processes actually happening, yes, give us information about how we can optimize that so we can, you know, go and do some work on the initial steps of the process and make everything a lot more streamlined for our people. >> I'll ask you maybe one final question here, Liz, as a way to wrap the conversation.
With all of this in mind, with the way that AI continues to innovate and kind of roll out over time to customers, with the way that your customers are coming to you for projects, what excites you the most about just kind of looking forward to where the AI market is and how Globant will continue to grow? >> Well, what really excites me the most from a Globant perspective is we have had a long journey on AI. It's going to be more than 12 years already that we have jumped into this space.
I have been heavily involved in the past eight. Um of course it's really to see how many of the things that we were discussing maybe between and just a small group of us in our offices about the future and how it could look like. Now it has it has turned into a reality. Um the interesting part is that I'm having a lot more conversations with people that I have never imagined before that I would have an AI related conversation and that gets really exciting.
Um but at the same time I believe that we still have a long way ahead of actually getting the most out of this technology. We know that there are capacity restrictions. Yes, that are still impact how much availability we have from these models, you know. Um so there is a lot of still to make the capacity be available for the processes, the agents that the companies needs to run to be really available. It's really interesting the pace at which those things are developing and us having the flexibility of you know, creating these awesome partnerships with companies like Anthropic can help us understand and be closer to the edge of technology, the state of the art of the technology and always think about how we can implement that, how we can bring that into our customers and into ourselves to drive the best out of the technology.
And I'm sure that as we keep on evolving our AI pods framework, the methodology itself, how it works, the business model around it more and more organizations will see the benefits of and and the importance of having this kind of services in which you are supervising the tokens and where you get the most value out of every single dollar that you spend actually on a on a token and how you get the most out of the capacity that is available out there, too. >> Well, if anybody found this video and wants to learn more about Globant and about your services, where can they go to find that information out? >> Well, they can search for myself in LinkedIn and feel free to reach out.
They can also go into our corporate website globant.com. There you will find a lot of information about the AI pods, our partnership with Anthropic, all the other partnerships that we already have, our full service offering and they can reach out through there also to, you know, have a conversation and discuss on how we can help them. >> Great. Well, thank you again for joining me, taking some time today. I appreciate it. If you want to learn more about Anthropic's partner ecosystem or where the AI market is headed, head to channelinsider.com. But, thanks again. >> Thank you.
This transcript was generated automatically from the
video's captions and may contain errors.
