Defense agencies are being offered a wide range of AI tools as spending on the technology grows. As much as $30 billion is slated for fiscal 2027 to buy and support next-generation AI hardware, cybersecurity, sensors and related software.
That is according to an article by Joe Kopacz published on October 5, 2026. The article argues that AI agents built into Microsoft Copilot can help defense workers get more done inside secure systems they already use.
Kopacz wrote that defense agencies must also update older technology systems. At the same time, they are enforcing stricter security rules on defense contractors.
He said this has forced officials to make hard choices between competing projects. Those decisions are also coming at a faster pace.
How SNC Uses Copilot Agents
Kopacz pointed to results at SNC, where nearly 5,000 people actively use Microsoft 365 Copilot. Those users have built more than 1,000 Copilot agents for general business and industry tasks.
One cybersecurity team created an agent that cut annual support work by more than 400 hours. According to the article, that freed up $100,000 in value for higher-priority security work.
An engineering team reduced development cycles from months to days. The same team secured previously unknown devices within 24 hours.
A large aircraft program reported improvements in real-time staffing, hangar flow and parts kitting across several sites. Another department turned a fully manual financial workflow into an automated process with better accuracy and faster delivery.
Agents also handle smaller daily tasks. These include guiding password resets, drafting ServiceNow help tickets and building PowerPoint briefings from SharePoint or OneDrive data.
Embedded Agents Versus Outside Chatbots
The article lists three reasons Kopacz believes built-in agents work better than external AI chatbots for defense use. The first is compliance.
Because the AI runs inside software already approved for government use, Controlled Unclassified Information stays protected. Kopacz wrote that this helps agencies avoid software approval processes that can take years.
The second reason is data privacy. According to the article, Copilot prompts, responses and company data fall under Microsoft's enterprise data protection commitments and are not used to train the underlying AI models.
Access is still controlled by each organization's existing identity, permission and data-loss-prevention settings. Kopacz noted that organizations must still use good governance to lower the risk of oversharing and data spills.
The third reason involves workers. Kopacz wrote that fears of job losses often slow down change, but Copilot is designed to act as an assistant while employees stay in charge of their work.
He described the tool as a "force-multiplying engine." The agency's secure file storage provides context and limits for the AI.
Kopacz said putting AI in place is an organizational challenge, not only a technical one. He concluded that federal employees should use AI tools like Copilot that are already built into secure, approved systems as defense budgets set aside up to $30 billion for AI in fiscal 2027.
