Requirements-Driven Development: The Missing Layer in Agentic Coding

For decades, the bottleneck in software development was coding. We threw everything at it — better languages, frameworks, DevOps, CI/CD, cloud infrastructure — all to make the engineering phase faster and more reliable. And it worked.

Now, with agentic coding tools and AI-assisted development, that bottleneck is dissolving faster than anyone expected. Code generation is becoming near-instantaneous. The engineering phase that once consumed months now takes days or hours. But this acceleration has exposed a deeper problem that was always there, just hidden behind slower delivery cycles: we're still terrible at defining what to build. Ambiguous requirements, lost context between business and engineering, specs that decay the moment they're written. When coding was slow, these gaps got patched along the way. When coding is instant, they become catastrophic. You can now build the wrong thing at unprecedented speed — and do it confidently.

This is the next evolution of the software process. The bottleneck has shifted upstream, from how we build to what we build and why.

In this talk, Xesca Alabart shares how Gluecharm is addressing this new reality with an AI-native requirements management platform built on 150+ specialized micro-agents and knowledge graphs.

She'll walk through why traditional requirements practices can't keep pace with AI-accelerated development, how "Application Maps" preserve business context across the full lifecycle, and what it takes to build the missing layer between business intent and working code.