Pentagon Eyes AI Agents to Slash Software Approval Times

The War Department is piloting AI agents to automate parts of its authority to operate (ATO) process, an effort officials say could shorten the time it takes to deploy new software.

“[Generative AI] could actually do a lot of the compliance tasks, versus having humans have to sit down and type out massive pages of documents every time,” Chief Digital and AI Officer Cameron Stanley said Tuesday at AWS Summit Washington, D.C.

Stanley said the current ATO process can take as long as two years, which would mean software and AI models become outdated before they reach users. By using generative AI to perform compliance tasks and generate required documentation, the department hopes to compress those timelines and deliver new capabilities faster.

The effort reflects a broader push by the office to accelerate AI use across the department. Stanley touted the continued growth of GenAI.mil as evidence of that transformation.

“We’ve gone from 60,000 users to over half of the department,” Stanley said. “There are 3.3 million Department of War users that are out there … and over half of them have touched it.”

Last week, the agency launched the Agent Network, an initiative that pairs combatant commands with commercial AI and defense technology companies to deploy agentic AI into military operations. The network helps commanders correlate intelligence from multiple sources and deliver decision-ready information at the speed of conflict.

“In general, we collect [petabytes of] data to make decisions … Human cognition is just not going to be able to keep up on the modern battlefield,” Stanley said.

Earlier this year, CDAO reorganized its Advana enterprise analytics environment into three separate programs, including the new War Data Platform, which provides standardized data access for AI applications.

Stanley said the War Data Platform has already demonstrated value. During Operation Epic Fury, the platform integrated dozens of new data feeds in real time, allowing commanders to receive operational information faster while supporting AI-enabled decision-making.

“The War Data Platform has actually already been quite successful,” Stanley said. “We were able to incorporate dozens of new feeds in real time … and get data at the speed of conflict.”

He said CDAO’s strategy is to rapidly field commercial technology, iterate alongside users and continuously improve capabilities rather than waiting for fully mature systems.

“Our goal is to get to 80%, then put it through its paces, find out what’s broken, fix it, iterate again and keep going until we get to a solution that’s actually world-beating,” Stanley said.