Adam Terlson is the chief technology officer of BlueFolders, an AI-native venture studio, and the developer of Agentic Chartering, an architecture that separates AI-generated planning from tightly controlled execution. His work applies lessons from large-scale retail infrastructure to a central problem in autonomous software: allowing models to devise sophisticated workflows without surrendering operational oversight.
As a lead engineer at InterNations in Munich, Terlson rebuilt the expatriate network’s mobile experience with React Native, helping raise its application rating from 2.3 to 4.6 stars. His incremental mobile-migration architecture applied the strangler pattern to replacing legacy applications without disrupting existing users. He also argued for reusable React components that remain independent of application-specific assumptions.
At Best Buy, he led mobile self-checkout across 900 retail stores, using state machines to coordinate mobile behavior and backend services, and supported cloud-native development and DevOps on Amazon Web Services. After leaving the company, he adapted those techniques for autonomous systems and released AgenticStateMachines, a pattern library for tool use, feedback, collaboration, human approval, orchestration, and generated workflows. He now develops AI-native products and platforms at BlueFolders.
- Statecharts as enforceable guardrails. Explicit states, transitions, and guards constrain agent behavior while supporting audit trails, event replay, recovery, and model-based testing. Terlson also acknowledges practical difficulties including state proliferation, workflow versioning, and concurrency.
- Actor-model coordination for autonomous systems. Independent actors communicate through messages while maintaining bounded internal state. Azure Service Bus sessions and Amazon SQS FIFO message groups preserve ordering across distributed workflows; authenticated approvals and transition guards prevent unauthorized actions. His AI Engineer Summit presentation develops these patterns through specialized planning, retrieval, critique, and procurement agents.
- Agentic Chartering and zero-trust execution. A model composes a declarative workflow from approved capabilities, while governed infrastructure handles retries, permissions, human approvals, and recovery. Terlson’s account of chartering in production describes a flexible, model-authored process that never requires unrestricted execution of generated code.
