Building on our previous exploration of agentic marketing campaign workflows, this session deep-dives into the "intelligence engine" that adds a predictive layer to those workflows: Silicon Personas. We move beyond simple "chat with your data" to show how harness engineering in practice creates a robust simulation system. By systematically combining real-world data and behavioral instructions we can replicate the decision-making "black box" of thousands of individuals. This deep dive focuses on another approach where we leverage the power of these AI harnesses to drastically simplify traditional research while reducing biases and "AI hallucinations". Participants will learn how to integrate these high-fidelity synthetic outputs directly into their existing marketing automation to validate business assumptions in minutes and drive measurable growth in Customer Lifetime Value (CLV).

Most organizations are no longer experimenting with AI - they are trying to scale it. And this is where things start to break. Despite strong models and successful pilots, many AI initiatives fail to deliver real value. Not because the technology doesn’t work, but because success is not well defined, measured and managed in real-world conditions. In this interactive session, participants will step into a simulated AI transformation scenario, making decisions under real constraints such as data limitations, cost pressure, user behavior and governance requirements. Through this exercise, we will uncover why AI systems that “work” in demos often fail in reality and what it actually takes to build systems that are measurable, reliable and worth scaling. Target audience: product managers, data/AI professionals, business leaders, curious minds and anyone involved in implementing or scaling AI systems in practice.

Join us for an interactive, engineering-driven workshop exploring the design of LLM agent systems. Together, we will examine how GenAI applications evolve from simple single-prompt interactions into more structured, agent-based architectures built for reliability, scalability, and more complex workflows. We will unpack key design questions, including when agents truly add value, which types of problems they are best suited to solve, and what trade-offs they introduce in terms of complexity, control, latency, and maintainability. Along the way, we will cover important concepts such as reasoning patterns, tool integration, memory, planning, and orchestration strategies. In a hands-on session, we will also take a deeper look at frameworks such as LangGraph and CrewAI, giving participants the opportunity to explore different implementation approaches in practice and better understand how to design effective agent systems for real-world use cases.