Project Thunderforge demonstrates how agentic AI can transform military planning — enabling human planners to evaluate more scenarios faster while maintaining full auditability and oversight.

Outdated Planning at the Speed of Bureaucracy

Military planning processes for the U.S. Department of War rely on decades-old technology, creating a critical mismatch between planning timelines and operational reality. By the time military plans are finalized, conditions on the ground have often already changed, undermining strategic effectiveness. This static approach creates a capacity ceiling where human planners must manually touch every step of complex planning processes, limiting the speed and volume of scenarios that can be evaluated. Major command headquarters like U.S. Indo-Pacific Command and U.S. European Command face the challenge of developing robust, probability-weighted plans while operating within these outdated systems. The Defense Innovation Unit awarded Scale AI a prototype contract for Project Thunderforge in March 2025 to address this fundamental inefficiency, recognizing that government workflows must shift from reactive to proactive to deliver services at the speed of demand rather than the speed of bureaucracy. 

How Thunderforge Works

Project Thunderforge aims to deploy AI-enabled planning capabilities that go beyond automation by autonomously orchestrating complex military planning workflows supervised and iterated on by human experts. The system's agentic nature lies in its ability to enable staffs to ideate on plans and challenge existing assumptions based on changes in the operational environment, creating "self-healing" plans that update and explore options under human supervision. The agents analyze and modify plan variants at machine speed and scale, removing the capacity ceiling of manual processes. Trust is maintained through a comprehensive logging and traceability framework that captures verifiable audit trails of every state change, decision node, and tool invocation. While agents utilize Chain of Thought reasoning to propose options, human planners can inspect not just what the agent claimed it did, but the specific data and tools actually used to generate each plan, ensuring forensic auditability critical for high-consequence government missions.  

Trust, Auditability, and Human Oversight

Though still in early development and deployment phases, Thunderforge demonstrates clear value by automating the manual labor of data synthesis, enabling human planners to evaluate significantly more scenarios than previously possible. This ensures decisions are based on the most robust, probability-weighted analysis available, providing critical decision advantage. The Department of War has recognized this potential by making AI agent development and deployment for both warfighting and enterprise transformation a top priority in its new AI strategy. 

Key Lessons

Key lessons emphasize that reliability must be the focus for high-consequence government work where lives and livelihoods are at stake, requiring deterministic and auditable agents with humans kept in the loop. The core insight applies across the public sector: the future isn't about replacing humans, but moving at real-world speed while reducing cognitive burden, freeing civil servants to focus on uniquely human elements such as empathy, ethical judgment, and accountability.