Progressive Crystallization slashes AI agent costs by 70% in production

AI agents deployed for IT operations are typically permanent cost centers, requiring full LLM inference for every execution—even on solved problems. In a new arXiv paper, researcher Arun Malik introduces progressive crystallization, a lifecycle that treats agent exploration as a discovery mechanism rather than a permanent execution model. The framework defines a three-stage execution taxonomy: fully agent-orchestrated, hybrid, and fully deterministic workflows. An evidence-based promotion mechanism converts repeatedly validated agent behaviors into cheaper, more reproducible deterministic steps, while automatically demoting workflows that regress in performance.

Evaluated on a production cloud networking AIOps system processing tens of thousands of incidents per month, the approach increased deterministic execution from 0% to 45% over eight months. Per-incident agent costs dropped by more than 70% despite the volume of incidents doubling. The method also improved safety through greater reproducibility and auditability, since deterministic workflows are far easier to review than opaque agent reasoning. The paper further details the execution taxonomy, promotion/demotion criteria, trace extraction methodology, economic model, and safety considerations, positioning progressive crystallization as a practical path to making agents viable in cost-sensitive enterprise environments.