This article summarizes what was presented at the Urudata & Google event on May 15, 2025. It was automatically generated by our youngest assistant: NeuraBot, powered by Tu Neural.
Amid the whirlwind of today’s technology landscape, Artificial Intelligence (AI) has moved beyond being a futuristic promise to become a tangible motor of business transformation. At a recent event, Urudata, in its role as a Google Cloud partner, brought together leaders and visionaries to break down how the synergy between robust data and advanced AI is redefining the boundaries of what’s possible.
The day began with an inspiring reflection by Nicolas Gutierrez Brum, who, invoking Henry Ford’s famous phrase — "If I had asked people what they wanted, they would have said faster horses" — illustrated a fundamental truth about innovation. Often, companies seek incremental improvements, those “faster horses” manifested in optimizations such as basic chatbots, enhancements in route and stock logistics, or the automation of industrial and accounting processes. While valuable, these improvements represent only the surface of transformative potential.
La verdadera disrupción, argumentó Gutierrez Brum, reside en la “Artificial Intelligence applied with purpose.” This means going beyond simple automation to embrace solutions such as truly effective virtual assistants, dynamic demand prediction in logistics, omnichannel marketing strategies with proactive proposals based on deep behavioral analysis, predictive maintenance in the industry, and sophisticated fraud prevention in finance. This approach is at the core of Urudata’s Tu Neural proposition — a vision where AI not only modernizes but reinvents and delivers strategic and purposeful value to every facet of the business.
Moving forward with the exploration of the capabilities driving this new era, Ignacio Cattivelli, Engineering Director at Urudata, immersed us in the universe of Gemini, Google’s most intelligent AI model, and the robust ecosystem that supports it. Cattivelli highlighted Google Cloud’s multilayer architecture, which spans from AI infrastructure and cutting-edge research to models, tools, products, and platforms.
At the heart of this offer lies the Gemini family, featuring models such as Gemini 2.5 Pro, optimized for coding and complex prompts, and Gemini 2.5 Flash, designed for fast performance on complex tasks. One of the most notable capabilities of Gemini 2.5 Pro is its 1-million-token context window, capable of processing the equivalent of one hour of video, eleven hours of audio, over 30,000 lines of code, or 700,000 words. This scope enables unprecedented understanding and information generation, as evidenced by rankings on platforms like Text Arena, where Gemini models consistently demonstrate their superiority.
Beyond Gemini, Cattivelli presented a range of Google tools such as Gemma (lightweight open models), Imagen (high-quality text-to-image generation), Lyria (music generation), and Veo (video generation). The latter demonstrated its creative potential with striking examples, from an adorable dachshund diving into a pool wearing swimming goggles to the evocative recreation of a Renaissance palace, showcasing AI’s ability to generate cinematic and detailed visual content from simple textual descriptions.
La presentación culminó con la visión de “Google Cloud en la era de los Agentes“. Esta nueva arquitectura se apoya en Vertex AI for model building and management, a robust foundation of Datos multimodales and the AI Hypercomputer for optimized performance. Built on this infrastructure are the Gemini API and AI Studio, along with tools like Google Agentspace, which facilitate the creation of AI applications and agents by integrating enterprise search, the best generative AI models, and a robust, secure agent platform. A future was glimpsed where agent development, powered by kits like LangGraph and LangChain, becomes more accessible, enabling companies to build customized solutions with unprecedented efficiency and scalability — all on the same infrastructure that powers giants like Google DeepMind, Search, and YouTube. Additionally, the efficiency of models like Gemini 2.0 Flash and Flash-Lite was highlighted, offering competitive “intelligence per dollar” competitiva, democratizando el acceso a la IA de vanguardia.
At last, Pablo García, I+D Director at Urudata, anchored the discussion on the most crucial element for AI success: data. García began by warning that, in times of rapid change such as the present, the difference between companies that take off and those that stagnate lies in their ability to adapt and intelligently use their assets, with data being one of the most valuable.
He demystified the apparent easiness of implementing solutions like RAG (Retrieval Augmented Generation), subrayando que “los datos nunca están como los preciso”. Este es el ““the data problem”: implicit and machine biases, the illusion of the wisdom of crowds (where a small percentage of users generate most of the content), and concept drift, which causes models to lose accuracy over time.
Faced with these challenges, García emphasized a shift in focus in Machine Learning: from model building and deployment toward prioritizing data and LiveOps. Traditional software engineering is not enough; ML implementation is inherently more complex and requires a solid data strategy sólida, definida como “un concepto central e integrado que articula cómo los datos habilitarán e inspirarán la estrategia de negocio”.
It was proposed the evolution from the Lakehouse to an Operational Data Store, where rapid experimentation, automation, efficiency, and scalability are key. This is where data products emerge as a vital bridge between raw data and analytical transformations, enabling the creation of specific, reusable data assets. This“data as a productapproach unlocks significant value by increasing productivity, democratizing access, reducing risks, and optimizing resources. To achieve this, co-construction principles are needed, evaluating data in terms of confidentiality, availability, integrity, quality, cost, and performance, and enabling a data-mesh architecture where data domains operate more autonomously and efficiently.
According to García, the ideal platform should combine classic attributes (scalability, low TCO, security) with modern ones (multimodality, decoupling of data and processing, real-time capabilities, and AI embedded across all layers). Google’s AI-ready Data Cloud meets this need, with BigQuery at its core, offering a unified platform that combines the best of data and AI. Capabilities such as Unified Query, Object Tables with BQML, and the new Data Engineering Agents — which automate data engineering tasks like pipeline creation and optimization using natural language — are transforming how companies interact with and extract value from their data.
In summary, the event organized by Urudata and Google Cloud painted a clear picture: Artificial Intelligence is an unavoidable transformative force. However, its power is only fully unleashed when supported by an intelligent data strategy and advanced technology platforms. The combination of Urudata’s vision, embodied in solutions like Tu Neural and its deep expertise in data management and strategy, together with the power and versatility of Google Cloud and its AI tools like Gemini and BigQuery, offers companies the path not only to adapt to the future but to actively build it.