Pentagon’s AI Push Faces a People Problem, Not a Technology One
The U.S. Army’s biggest obstacle to deploying artificial intelligence is not a shortage of software, Benzinga reported Saturday. It is the soldiers and civilian staff expected to use it.
Outgoing Army Chief Information Officer Leonel Garciga made the assessment as he prepared to leave the role, arguing that military AI adoption has moved faster than the workforce can absorb. Demand for basic training and clearer guidance, he said, now ranks among the loudest signals his office receives from the field.
Troops Left Behind by a Fast-Moving Rollout
Garciga said personnel across the force are asking two recurring questions. The first is how to get trained on tools they do not yet understand. The second is how policy is supposed to keep up with a technology that evolves faster than procurement rules were designed to handle.
His diagnosis points to a structural tension inside the Pentagon. Acquisition timelines that once spanned years are struggling to match an industry where major AI capabilities can shift within months. Garciga has pushed for a different posture entirely, one that prioritises wide deployment first and refinement afterward.
A ‘Break Some Glass’ Approach to Procurement
Rather than wait for multi-year approval cycles, Garciga advocated pushing tools broadly into the force and iterating quickly, telling Business Insider the approach should be to make capabilities “ubiquitously available” and watch what follows. He also pushed decision authority downward, giving battlefield commanders more direct control over technology access without requiring drawn-out internal processes.
That philosophy broadly aligns with acquisition reforms pursued across both the Biden and Trump administrations, which sought to compress the gap between commercial innovation and battlefield deployment.
Broader Pentagon AI Drive Gains Momentum
The Army’s internal challenge sits inside a much larger military AI expansion. In April, the Pentagon’s AI chief confirmed that Alphabet’s Gemini model is being extended into classified defense environments, a sign that large commercial language models are moving beyond unclassified use cases.
Contractors including Booz Allen Hamilton and Palantir Technologies have positioned themselves as primary integrators for that push, building infrastructure to connect commercial AI to sensitive government workflows.
The pace of that expansion makes Garciga’s warning more pointed. Deploying capable systems quickly is one challenge. Ensuring the humans operating them understand what they are seeing, and what to do about it, is proving to be another matter entirely.
Read Next: OpenAI Eyes $3 Trillion Valuation as SoftBank Deal Reshapes AI Race
