AI agents are only as good as the data they're given, and that's a big issue for businesses (www.zdnet.com)

Salesforce’s new State of Data and Analytics report — based on surveys of ~7,600 business and data leaders — reveals a sharp tension: organizations are racing to become “agentic” and AI-powered (93% have at least one AI instance), yet their data foundations lag. Sixty-three percent of business leaders now call their firms very data-driven (up 10% year-over-year), but 63% of technical leaders say their companies still struggle to use data to drive priorities and half of business leaders aren’t confident they can deliver timely insights. Key pain points: 70% of data leaders say the most valuable signals are trapped in unstructured data, only 43% have formal data governance, data volumes are growing ~30% annually, and enterprises run hundreds of disconnected apps and environments — all of which erode trust and slow action. For the AI/ML community this matters practically and strategically. The report underscores that “all AI projects are data projects”: 84% of leaders say AI outputs are only as good as inputs, 91% say technical queries limit analytics at scale, and 92% flag poor data fluency among staff. That drives demand for investments in data infrastructure (CIOs spend ~4x more on data infra than on AI), real‑time access, unstructured-data processing, harmonization, metadata/context enrichment, and governance. Without those fixes, agents risk producing incorrect or untimely recommendations; with them, agentic analytics can democratize insights via natural‑language and action-capable interfaces — but only if data quality, lineage, and contextualization are solved first.