Gartner estimates 30% of generative AI projects get abandoned after proof of concept. With agentic AI, the failure rate risks running higher because agents don't just give bad answers, they take bad actions. Research before deployment typically costs 5-8% of the implementation budget. Skipping it costs 3-5x the original implementation to fix. The math favors doing the research first.

The article discusses the “Impact Problem” faced by qualitative UX researchers in Big Tech, highlighting the tension between achieving measurable business outcomes and maintaining the integrity of user experience research. It outlines the challenges of communicating the value of qualitative insights in a business environment that prioritizes quantitative metrics and ROI, often leading to a disconnect between UX professionals and business stakeholders. The author calls for a paradigm shift towards better integration and acknowledgment of qualitative research within business strategies, advocating for mixed methods approaches and greater collaboration across disciplines to truly measure and understand impact.