Taking artificial intelligence from experiments and pilot projects to real business operation is one of the greatest challenges facing companies today. In an environment where generative models and intelligent agents promise to transform customer experience, the difference between a scalable implementation and a failed effort still lies in the foundations: data quality, orchestration, and real-time speed.
In this new episode of Digital Talks by t2ó ONE, Esther Checa speaks with Óscar López Cuesta, Senior Solution Consultant at Tealium with over 15 years of experience in data, technology, and digital marketing projects for major corporations such as BBVA, Orange, AXA, and Mutua Madrileña. Throughout the conversation, they analyze the infrastructure companies need to process high-quality data in real time and how new trends, such as local processing (AI on the edge), are changing the rules of the game.
Most companies have historically accumulated large volumes of information without a clear structure, creating a «digital hoarding» habit. For an AI model to correctly interpret a user’s context, feeding traditional analytics based on codes or proprietary dimensions from isolated tools is no longer enough; it requires a homogeneous, destination-agnostic foundational layer.
In this episode, you will discover:
- the main bottlenecks preventing companies from moving from AI pilots to scalable systems;
- how data validation strategies evolve when feeding generative models compared to traditional analytics;
- the technical and latency challenges posed by integrating real-time data;
- why the move toward AI on the edge reduces costs and protects privacy;
- how to orchestrate modular architectures (best of breed) to avoid vendor lock-in;
- and how the future of measurement is shaping up with the arrival of AI agents and new protocols.
One of the most critical aspects today is latency. When AI systems interact with users in present time—such as in fraud detection or real-time offer hyper-personalization—data processed a day late no longer adds value. Maintaining architectures capable of streaming events and delivering responses in under 100 milliseconds is essential to activate the exact context at the precise moment of interaction.
At the same time, local processing or AI on the edge is gaining ground driven by two main factors: protecting consumer privacy and the high computational cost per token in the cloud. Running smaller models directly on mobile device chips allows systems to infer sentiment or intent without personal data ever leaving the handset, providing a more secure and efficient governance framework.
Finally, the shift toward an ecosystem dominated by conversational agents represents a paradigm shift in web analytics. Since AI agents do not execute JavaScript code in traditional browsers, a company’s ability to capture server-side events and make its business rules available in real time will determine its relevance in this new digital era.
🎧 Want to discover how to structure your data strategy to power your AI systems? Listen to the full episode of Digital Talks with Esther Checa and Óscar López on your favorite podcast platform or watch it on our YouTube channel.
