Enterprises are rapidly shifting from human-interpreted dashboards to Autonomous Analytical Entities (AAE) that execute decisions directly on production systems. This transition introduces a new failure mode—Agentic Divergence—where decentralized agents act on misaligned, drifted, or out-of-scope data products and metadata, leading to high-impact errors at scale. This paper proposes the Autonomous Analytical Coherence (AAC) framework, centered on an Analytical Control Plane (ACP) that inserts a mandatory, machine-enforced governance layer between AAEs and decentralized data products. The ACP mandates Agentic Data Contracts (ADC) as runtime dependencies and enforces Kullback–Leibler (KL) divergence-based drift checks within Trusted Execution Environments (TEE) to safeguard both analytical coherence and data sovereignty. Simulation-based experiments across finance and logistics workloads indicate that AAC reduces erroneous autonomous transactions by 77% compared with uncoordinated agent deployments, with only a 25 ms median increase in latency. These results demonstrate that treating governance as a runtime dependency is a practical path toward safe, high-stakes autonomous analytics in enterprise data meshes.
Balaram Tripathy (Thu,) studied this question.