Presented by CACI
Artificial intelligence is affecting much more than intelligence analysis. Its influence now stretches across the entire intelligence cycle, changing how agencies gain access to information, collect and exploit data, produce assessments and disseminate finished intelligence.
Ron Carback, Client Executive at CACI, says agencies throughout the intelligence community are adopting and adapting AI to improve these activities. The goal is not simply to add another technology to an analyst’s desktop. It is to transform workflows and allow intelligence professionals to spend more time on the work that requires uniquely human insight.
Carback emphasizes that AI should generally be considered a tool rather than a complete solution. While algorithms can process information and produce outputs, people must remain responsible for the judgments and consequential decisions that follow.
“AI is not intelligent,” Carback said, recalling an observation from an artificial intelligence expert. “It’s math. It’s data science. It’s computer science. It’s software.”
Those capabilities can be extremely valuable, but they do not replicate the full range of human reasoning. The advantage comes from combining AI’s ability to process large amounts of data with an analyst’s ability to assess context, ambiguity, intent and risk.
Those considerations become more complicated in the intelligence community, where agencies must determine how a technology can be used in both unclassified and highly classified environments. A solution that performs well in one setting may not be appropriate or authorized in another.
The emergence of agentic AI could also fundamentally change the analyst’s daily role. Carback cited a government leader who characterized every analyst as a potential manager or leader of a team of AI agents. Instead of personally completing every step in a workflow, an analyst may direct multiple agents assigned to different research, collection or processing tasks.
That role introduces new responsibilities. Analysts will need to know how to provide effective prompts, direct agents toward the correct data sets, evaluate the resulting outputs and determine whether the agents are answering the intended intelligence questions. In that environment, prompt design and data oversight become extensions of intelligence tradecraft.
Workforce development must evolve accordingly. Carback noted that CACI provides both general and specialized AI training through its internal university. The company also uses communities of practice in which employees share how they are applying AI in their work.
On one contract involving deep discovery analysis, CACI personnel lead regular sessions in a classified environment to demonstrate how they are adopting AI for specific mission activities. This type of peer learning can help organizations move from general awareness to practical, mission-focused application.
The scale of the data challenge is not new. Intelligence professionals have long discussed the volume, velocity and variety of available information. AI expands the community’s ability to triage and analyze that material, helping analysts understand multiple information environments and identify more of the relevant signals within them.
The growing number of AI products creates another challenge, however. Agencies and analysts could become overwhelmed by dozens of tools designed for different parts of the intelligence cycle. Without thoughtful integration, AI could add complexity to workflows instead of reducing it.
Carback argues that organizations must carefully determine which tool is appropriate for each application and intelligence function. Analysts should not have to decide constantly among scores of disconnected AI products. The technology should simplify their work and help consolidate processes, not create another layer of tool management.
Successfully navigating this transformation will also require a closer relationship between government and industry. Carback sees agencies increasingly engaging industry before requirements are finalized and asking companies to help explore problem sets rather than simply respond to predetermined specifications.
That engagement can help government organizations identify their actual requirements, develop clearer roadmaps and reduce program risk. It also allows industry partners to understand the mission before proposing a product or technical approach.
Carback believes the most productive integration happens when government and industry build together from the beginning. That can include joint innovation hubs, exercises and recurring reviews throughout the development process. Instead of waiting until the end of a program to deliver a finished product, both sides can evaluate progress and adjust the solution as mission needs become clearer.
AI adoption in the intelligence community will continue to move at different speeds across agencies and missions. The organizations that make sustainable progress will be those that introduce it methodically, train analysts for new responsibilities and preserve human decision-making. Equally important, they will engage industry early enough to ensure that emerging capabilities are connected to real intelligence requirements.