CoreWeave and Humatica: Scaling Proprietary Enterprise AI
CoreWeave EVP Chen Goldberg explains how bringing managed PostgreSQL into Kubernetes clusters reduces latency and helps firms own proprietary AI workloads

TL;DR
- Businesses are adopting AI faster than their organizational structures can support it, creating vulnerabilities.
- 66% of surveyed organizations depend on AI for operations that would fail without it, yet only 35% have a formal AI operating model.
- Organizational structure and readiness are greater barriers to AI progress than technological limitations.
- Companies prioritize proprietary IP (73%) over cost savings when moving critical workloads to self-controlled infrastructure.
- Bringing managed PostgreSQL into Kubernetes clusters reduces latency and allows firms to retain control over proprietary AI workloads.
- AI creation has decentralized, with 42% of production use cases originating in business units and 24% launched by employees.
- Five practices for lasting AI advantage include treating AI as an operating model, enabling decentralized creation, transforming knowledge into assets, enforcing strict security, and maintaining continuous ownership and observability.