Summary
Balázs Horváth, founder of VisualLabs, presents a counterintuitive argument at AI Engineer: as AI takes over code generation, the most valuable skill in software development is no longer writing code — it’s reading the room. Drawing on 13 years as a functional consultant bridging business stakeholders and engineering teams, Horváth contends that requirements elicitation is the last bottleneck AI cannot automate away, because you cannot prompt your way to getting the right people in a room.
He grounds the argument in a concrete internal example: a VisualLabs hackathon produced 21 agent ideas, 17 of which were abandoned for lack of business value, leaving only 4 with meaningful impact. His diagnosis is that AI defaults to the statistical average — producing faster horses rather than cars — unless a human can first articulate the right problem. The Henry Ford analogy runs throughout: customers couldn’t ask for automobiles because they didn’t know automobiles were possible.
Horváth introduces practical frameworks for bridging this gap, including story mapping (breaking user journeys into backbone steps and user stories), the Value-Architecture-Design (VAD) thinking sequence, and a four-question diagnostic covering whose problem is being solved, what winning looks like, what would cause refusal, and whether the solution changes a decision. He also argues that well-structured user stories feed AI systems better context because models were trained on that format. Business analysts, functional consultants, and product managers will find this talk a strong articulation of why their skills are becoming more — not less — critical in an AI-first development environment.
📺 Source: AI Engineer · Published June 29, 2026
🏷️ Format: Opinion Editorial
