Presented by Carahsoft
The most important conversations at DoDIIS 2026 were not really about individual technologies. Artificial intelligence, zero trust, data platforms, open architectures and cyber resilience were prominent throughout the conference, but government and industry leaders consistently brought the discussion back to a more practical question: How can these capabilities help people make better decisions faster without compromising security or trust?
Across the interviews, several connected themes emerged. Intelligence organizations must reduce the overwhelming volume of information confronting analysts, build architectures capable of absorbing constant technological change, strengthen partnerships with industry and make security part of every system from the beginning. Underlying all of those priorities is a shift away from acquiring technology for its own sake and toward measuring its effect on the mission.
The intelligence community has become extraordinarily effective at collecting data. The harder challenge is determining what matters.
Analysts may begin their day facing tens of thousands of reports, images, signals or other data points. Their value does not come from reviewing every item manually. It comes from recognizing what is significant, establishing context and turning that information into intelligence a decision-maker can use.
AI and machine learning can help filter inactivity, duplication and irrelevant information before it reaches an analyst. In an imagery mission, that might mean removing thousands of images containing only ocean and whitecaps so analysts can concentrate on the relatively few that contain a ship, submarine or other potentially important object.
This creates a more meaningful measurement for AI than the number of models deployed: time to first indication. If a system can bring the 20 most relevant developments to the top instead of asking an analyst to search through 20,000 items, it can shorten the time between collection, recognition and action.
Reducing volume is only the beginning. Information must also be contextualized and presented in a form that supports a decision.
Richard Breakiron, Senior Director of Strategic Initiatives at Commvault, described data as the link connecting every part of the mission.
“Data is the connective tissue between people, process, and technology,” Breakiron said. “Actionable insights come from data in motion.”
Breakiron used a red, amber and green model to illustrate the point. Some cases are clearly safe, while others clearly require intervention. The greatest value comes from identifying the ambiguous cases—the “ambers”—that need human attention. Once a person resolves one of those cases, the result can be captured and used to improve the system’s future recommendations.
That cycle allows automation to handle routine determinations while directing human expertise toward the difficult decisions. It also reflects a broader theme from DoDIIS: AI should augment the workforce, not remove people from consequential decisions.
The ability to produce an answer quickly does not make that answer trustworthy. Analysts and leaders need to know where information came from, how it was processed and what evidence supports the conclusion.
“No analyst will put a machine-derived result in front of a politician or whoever needs to make a decision,” Gould said. “We have to build systems that are going to be able to provide the provenance that they use on a daily basis, along with the result that the machines come to.”
This need for provenance connects AI adoption with longstanding intelligence tradecraft. Models can discover patterns, prioritize information and reduce repetitive work, but subject-matter experts must still ask the right questions, challenge implausible outputs and understand the consequences of acting on a recommendation.
As execution becomes faster and less expensive, critical thinking becomes more valuable. The workforce will need greater AI fluency, but it will also need the confidence to recognize where automation should stop and human judgment must take over.
Speed was another defining theme, but speakers rarely described speed as a goal by itself. The objective is to remove the friction that prevents intelligence from reaching the right person in time to change an outcome.
“It’s not just about can I build a new widget or a better mousetrap and then bring it to market and try to sell it,” Thompson said. “It’s literally how are we aligning to the needs of those mission partners and making sure we’re driving those outcomes.”
For intelligence and defense organizations, friction often appears where information must cross organizational boundaries, classification levels or coalition environments. Necessary accreditation and security processes can also slow deployment when they are not designed for rapidly changing technology.
“Speed and removing that friction in those decision processing systems is more important now than ever,” Thompson said.
The answer is not to eliminate controls. It is to design architectures and processes that allow information to move securely without forcing every new capability through a multiyear integration effort.
No single company or agency can provide every component of a modern intelligence environment. AI models, computing infrastructure, security tools and mission applications are all advancing at different speeds.
Open standards and modular architectures give agencies the ability to introduce new capabilities without replacing entire systems. They also reduce dependence on one provider and make it easier to preserve valuable legacy data while modernizing how that data is accessed and used.
This technical openness reflects a broader evolution in the government-industry relationship. Collaboration remains important, but agencies increasingly need deeper integration with commercial partners. Government contributes mission expertise, operational context and accountability. Industry contributes rapid iteration, emerging technology and experience deploying capabilities at scale.
Greater speed, connectivity and automation also expand the attack surface. DoDIIS speakers consistently emphasized that cybersecurity must be engineered into systems rather than added after cost, schedule and performance decisions have already been made.
Zero trust principles must now apply not only to people and applications but also to AI agents. Each agent needs a defined identity, purpose and level of access. Organizations must be able to observe what agents do and prevent them from moving beyond their authorized roles.
Breakiron argued that programs such as FedRAMP are most useful when viewed through an operational rather than bureaucratic lens.
“If you think of FedRAMP as compliance, nobody wants to do it,” he said. “If you think of it as an operation where you’re going to want to be successful, everybody’s going to want to do it.”
That captures one of the clearest messages from DoDIIS 2026. Governance, cybersecurity and compliance should not exist simply to generate documentation. They should reduce risk, strengthen resilience and give leaders the confidence to use new capabilities in real missions.
The intelligence community’s path forward will depend on combining machine speed with human judgment, commercial innovation with government expertise, and greater access to data with stronger security. The technology will continue to change. The enduring challenge is building the people, partnerships and architectures that can turn that change into decision advantage.