State of AI / Enterprise adoption reports (2025)

Multiple 2025 enterprise reports document rapid generative-AI adoption, RAG deployment gaps, and an emerging opportunity for decision-context systems.
Key insights
Generative AI has reached enterprise-scale spending and adoption
Menlo Ventures' 2025 'State of Generative AI in the Enterprise' reports enterprise AI spending rose from $1.7B (2023) to $37B (2025). The report shows strong adoption in programming, department-level copilots and vertical AI; 76% of use cases run purchased solutions rather than self-built. Startups capture most application-layer spend while incumbents hold infrastructure share.
RAG systems face deployment hurdles despite lab success
A systematic review of 77 RAG studies finds common enterprise challenges: hallucination control, privacy, latency/scalability, domain adaptation, and difficult business-impact measurement. Only 13% of research demonstrates enterprise deployment, and many production requirements (sub-100ms retrieval, privacy-preserving search) remain open.
Context Graphs proposed as next enterprise software frontier
A recommendation emerging across briefings argues next-gen enterprise systems must record AI 'why'—decision rationales and reasoning chains—via Context Graphs that link inputs, rules, approvals and outcomes so agents can generalize prior decisions into policy-aware automation.
OpenRouter/a16z data show agentic usage and multi-model patterns
a16z/OpenRouter analysis of large traffic logs reports agentic, multi-step reasoning traffic rose sharply; open-source and 'mid-size' models gained share; and geographic usage patterns shifted toward Asia. The report documents a shift from single-shot generation to tool-using, multi-step flows.