Yes, but poorly and with severe limitations in production. Without a unified data layer, AI agents must directly connect to isolated data silos via chaotic API chains, leading to high query latency, massive context window bloat, and frequent hallucinations caused by stale or conflicting data. A platform like Microsoft Fabric or a data lakehouse acts as the agent's central nervous system, providing a single, curated source of truth with universal security governance and optimized vector search, which is practically mandatory for reliable enterprise-grade automation.