Graph‑agentic retrieval‑augmented generation (RAG) couples structured evidence with adaptive controllers that can plan retrieval, traverse relations, verify intermediate claims, delegate subtasks and invoke tools. This synergy is crucial when answers depend on cross‑document, entity, temporal or institutional relations, yet it creates coupled failure paths: a defect in graph construction may become retrieved evidence, steer later control decisions and amplify toward the final outcome. In social‑good settings we must treat freshness, authorization, traceability, oversight and recourse as equally important to answer quality. We organize prior work by graph substrate, lifecycle, agent function, coordination pattern and authority boundary, and we separate graph‑based retrieval from observation‑driven graph control. Reported risks are synthesized into an evidence‑to‑action failure chain, and we propose an assurance‑by‑construction blueprint consisting of five interface contracts for evidence, retrieval, reasoning, capability & delegation, and outcome. These contracts make provenance, temporal validity, authorization, uncertainty and recoverability explicit at system boundaries. An illustrative public‑benefit information design shows how the framework constrains graph structure, permissions, abstention and operating authority. Finally, we outline an evaluation agenda that spans graph assertions, trajectories, claims, coordination and outcomes.
Review