Gil Cohen, Chief Product Officer at Cognyte – Interview Series

Gil Cohen is a seasoned software executive with proven track record leading large global organizations, over two decades of experience and expertise in enterprise software, big data, telco and artificial intelligence.

Gil leads Cognyte’s product organization, the company's global R&D centers, and the go-to-market strategy for the full portfolio.

Prior to joining the Cognyte business in 2021, Gil was the GM of NICE, a voice recording platform, and CEO of Telefonica Israel.

Cognyte is an Israeli-based global provider of investigative and security analytics software designed to deliver “Actionable Intelligence for a Safer World™.” Founded in 2021 as a spin‑off from Verint Systems, the company serves hundreds of government and enterprise clients across roughly 100 countries, enabling security, intelligence, and law enforcement agencies to merge, analyze, and visualize large, fragmented datasets for timely threat detection, investigation, and response .

What originally motivated a career path focused on intelligence, analytics, and national security? Were there early experiences or formative events that helped shape this direction?

It was a pretty natural fit. Cognyte’s mission – to help make the world safer – lines up perfectly with what I care about, both personally and professionally. I’ve spent years working in enterprise software, big data, telecom and AI, and I also have a military background, so stepping into a role that allowed me to contribute directly to that mission just made sense.

Over the years, I’ve led global organizations and R&D teams to deliver cutting-edge enterprise solutions and go-to-market strategies, giving me the experience to help make sure Cognyte stays ahead with tech that really empowers our customers. It’s incredibly rewarding to see how law enforcement, national security, and national and military intelligence agencies use our analytics solutions to spot threats early and keep people safe. What really keeps me motivated is knowing we’re building tools that help these agencies stay one step ahead of criminal and terror activity and adapt quickly to new challenges. Seeing our tech in action – making a real impact – is something I’m genuinely proud of.

Over the years, how has the investigative technology landscape evolved, and how has Cognyte’s mission adapted to meet the changing needs of law enforcement and intelligence agencies?

The investigative technology landscape has undergone significant evolution, primarily driven by the rapidly growing volume and diversity of data, the emergence of increasingly sophisticated adversaries, and the constant evolution of technology. A core challenge shaping this environment is the inherent imbalance between bad actors and those working to stop them. Bad actors (whether criminals or terrorists) operate with agility and persistence, able to exploit a single weakness with devastating impact. Law enforcement and security organizations, by contrast, must identify, understand and respond to a wide array of evolving threats, often in real time and with absolute precision. This asymmetry places immense pressure on investigative teams to act faster, smarter and more accurately than ever before. Analysts frequently grapple with growing volumes of data, fragmented tools and pressing operational demands, making it challenging to translate their investigative insights into effective interactions with advanced solutions.

In response, Cognyte has continuously adapted to fulfill our mission of helping customers reveal insights and eliminate the unknown through actionable intelligence. Known for delivering breakthrough technologies, Cognyte leverages AI across our platform, at scale, to fuse, analyze and visualize complex data – enabling faster, more precise decisions and actions. The intelligence co-pilot was built with this asymmetry in mind, designed to level the field by enabling defenders to work at the speed of the threat. The intelligence co-pilot is a generative AI assistant that simplifies investigative workflows through natural language interaction and explainable outputs, purpose-built for the needs of law enforcement, national security and intelligence agencies. Now even with sophisticated systems, analysts can quickly gain value without needing technical know-how to operate the system or define complex queries.

The launch of the intelligence co-pilot introduces a generative AI capability specifically built for investigative workflows. What makes this capability fundamentally different from existing AI tools currently in use within the security sector?

Cognyte’s intelligence co-pilot is fundamentally different from other AI tools because it’s purpose-built for real world investigations, not just generic tasks. Unlike typical GenAI solutions, it’s embedded directly into the investigative workspace and process, understands the terminology analysts use, and delivers transparent, explainable results. It is engineered to analyze sensitive and classified data sources which government agencies deal with, operates securely in any deployment environment, and is designed to accelerate – not replace – human expertise. This enables investigators and analysts to reap the full benefits of human-machine collaboration with a solution that is uniquely designed for the high-stakes demands of security and intelligence agencies.

Many investigators face a disconnect between their real-world instincts and the technical demands of legacy platforms. How does this new co-pilot address that challenge and bridge the gap between human reasoning and machine output?

Investigators and analysts are trained to conduct investigations and intelligence analysis, not to engineer complex queries or translate their thinking into rigid system syntax. Cognyte’s intelligence co-pilot is built with that reality in mind by letting users work in their own words. It takes natural-language input and converts it into structured, explainable logic, eliminating the need for technical queries or deep system expertise. By aligning with how investigators actually think and work, the co-pilot removes friction, accelerates insight generation and frees analysts to focus on what they do best: following leads, examining evidence and analyzing threats.

Investigative environments often involve overwhelming data volumes and highly fragmented tool ecosystems. In what ways does the co-pilot enhance data synthesis, accelerate analysis, and improve the overall pace of threat resolution?

Cognyte’s intelligence co-pilot transforms the investigative workflow by eliminating the manual bottlenecks caused by siloed tools and overwhelming data volumes. Instead of requiring investigators to translate complex questions into system-specific queries, the co-pilot allows them to use intuitive natural language to explore diverse, multi-source data. It automatically synthesizes structured and unstructured data, surfaces relevant connections, and presents results with clear, step-by-step logic. For example, an investigator can simply ask the co-pilot to “show me all connections between suspect A and suspect B in the last 6 months.”

The co-pilot will do the heavy lifting across multiple data sources and formats, like CCTV, financial records and handwritten reports. It will quickly sort, filter and categorize vast data sets to uncover insights and visualize direct and indirect suspect connections, removing the need for the investigator to sift through raw data. By unifying search, correlation and insight generation within a single GenAI-driven assistant, the co-pilot accelerates decision-making, enhances precision and significantly reduces the time needed to act on threats.

Security, explainability, and transparency are critical for mission-driven operations. What safeguards and architectural principles were prioritized to ensure that the AI co-pilot meets the stringent demands of intelligence organizations?

With over 30 years of experience supporting mission-driven operations, security, explainability and transparency were foundational to the design of Cognyte’s intelligence co-pilot. Security is built in at the architecture level: the co-pilot supports both cloud and on-premise deployments to meet stringent requirements around data sensitivity, privacy and sovereignty. It inherits the native user-permission model of the host system, ensuring it cannot expose data a user wouldn’t normally access via standard workflows.

For explainability and transparency, the co-pilot provides a clear view of its end-to-end reasoning, down to the exact logic, conditions and behind-the-scenes steps it took to generate results. Users can review, validate and even challenge the system’s interpretation using their own judgment. All reasoning is fully auditable, enabling traceability and accountability in high-stakes environments.

The platform offers visual logic flows and transparent reasoning in response to natural language queries. How do these features influence the way investigators make decisions under pressure, especially in high-stakes or time-sensitive cases?

Cognyte’s intelligence co-pilot enhances investigative decision-making by combining natural language input with step-by-step visual logic flows based on proven domain expertise. Results are visualized within an intelligence analysis workspace for context and further exploration. These visualizations reveal how data points are connected and how insights are generated, making the system’s reasoning fully transparent. In high-pressure environments, this allows investigators to rapidly validate relevance and move from question to confidence without relying blindly on black-box outputs. The result is faster, smarter decision-making grounded in explainable AI, fully aligned with real-world investigative workflows.

What types of machine learning models form the foundation of the co-pilot’s generative capabilities, and how were they trained to reflect the language and logic specific to intelligence workflows rather than general conversational AI?

The models at the base of our copilot are some of the industry’s latest and most advanced LLM models. We fine-tune the LLMs to each customer’s domain and jargon using our proprietary data models, that represent the system’s data structure – including entities, relations and parameters. This enables us to train the copilot on a representative or ‘digital twin’ of our customers’ domain, rather than on the actual data which is sensitive and not available as a training set. We also include best-practice user flows derived from decades of internal, industry-specific methodologies, our customer success teams, and with the help of our community of customer champions. Using this method, we created a scalable approach to take off-the-shelf LLMs and turn them into domain models, fine-tuned to security needs.

Explainability and structured reasoning are emphasized in the system's outputs. What machine learning techniques or model architectures were used to ensure transparent, traceable decision-making, especially when analysts need to validate results under scrutiny? 

The co-pilot is trained on the data model and is fine-tuned for intelligence and investigative best practices, reflecting a training approach that goes beyond general conversational AI. This specialized training is built upon more than three decades of domain expertise in investigative analytics. This deep integration of real-world operational knowledge ensures that the co-pilot inherently understands and processes the specific terminology, logic and nuances critical to intelligence workflows. It takes the user’s natural language question and understands how to translate it within the Cognyte solution as an API call, querying for the requested data. The query is at all points available to the analyst, so it's transparent which query was sent. The data is then returned from the solutions databases, removing any chance of hallucinations, with structured, explainable logic. No separate traceability is needed for the data itself to validate that the summary and insights make sense.

This product emerges at a time when AI is transitioning from pilot projects to full-scale operational deployment. How does this reflect broader trends in how national security organizations are approaching AI adoption in 2025?

 The launch comes as generative AI adoption accelerates across enterprise and government, with over $2 trillion in investment expected around AI in the sector over the next three years. 2025 marks a clear inflection point, as organizations move from experimentation to real-world deployment. Agencies see the potential value of leveraging LLM-powered co-pilots, and in a recent Cognyte survey, 47% of agencies surveyed cited GenAI-powered data analysis and exploration as the top technology capability that can accelerate their investigations. For law enforcement and national security agencies, this transformation is setting a new standard for speed, scale and precision in decision-making. With lives and national interests at stake, the demand for trusted, operational AI is no longer a choice – it’s a mission-critical imperative, and one that the Cognyte intelligence co-pilot was built to meet.

Looking five years ahead, how is the role of generative AI expected to evolve within military intelligence, law enforcement, and national security? What impact will this have on investigative strategy, organizational structure, and global threat response?

In the coming years, generative AI will play a pivotal role in helping close the operational gap between offense and defense. Bad actors will continue to innovate rapidly, often unconstrained by rules or risk. Public safety organizations, by contrast, will need to maintain consistent, adaptive coverage across an expanding range of threat vectors. In this environment, AI won’t just be a force multiplier; it will become an operational necessity to keep pace with adversaries who are fast, unpredictable and increasingly automated.

Thank you for the great interview, readers who wish to learn more should visit Cognyte.