AI has moved from promise to everyday practice, and the shift is shaping how organisations design, deploy, and extract value from their video systems. Across the conversations we had with Hanwha Vision, i-PRO, Milestone, Cathexis, Verkada, and Eagle Eye Networks, one theme was clear: AI in 2025 was not about adding more features. It was about making intelligence usable, accurate, and tied directly to operational outcomes.
A major focus this year has been improving the quality and reliability of AI at the point of capture.
At Hanwha Vision, Head of Product and Marketing John Lutz Boorman highlighted how their Wisenet 9 chipset is transforming everyday imaging performance. He explained that Wisenet 9 “delivers clearer, more balanced video in challenging light conditions” and helps operators “identify critical details such as clothing colour and number plates quickly”, particularly in busy or low-light environments. Features such as blocked exit detection, loitering detection, and Dynamic Privacy Masking show how AI is being shaped around the real-world problems operators face rather than theoretical scenarios.
Hanwha’s soon-to-launch AI Rugged PTZ cameras further reflect this shift, designed to maintain full analytics capability even in harsh outdoor or perimeter conditions. The message from Hanwha is consistent: imaging quality remains the foundation of any meaningful AI deployment.
AI as an Investigative Assistant, Not a Search Tool
For Verkada, the biggest strides in 2025 came from reducing the burden of investigations. VP of Product Abraham Alvarez described their new AI Unified Timeline as the update customers referenced “nearly every single time” after the VerkadaOne conference. Rather than relying on multiple motion clips or isolated events, the Unified Timeline brings together facial detection, appearance similarity, and contextual cues to reconstruct a person’s or vehicle’s journey across multiple cameras.
“The amount of time this is saving organisations is truly amazing,” Alvarez said, noting that the feature has significantly reshaped how teams piece together incidents. Verkada also saw surprising demand for their mobile security trailers, driven by customers who want to deploy the same AI-powered search tools in temporary or remote environments.
Operational Insight as a Strategic Deliverable
If 2024 was the year organisations experimented with using video for non-security tasks, 2025 was the year it became normalised.
At i-PRO, this was a consistent message. COO Gerard Figols noted a marked shift in expectations: “End users want more value from their systems by using the data they generate.” People counting, occupancy analysis, and trend dashboards are becoming just as important as object detection. These features are helping teams make decisions about staffing, building usage, energy allocation, and customer experience.
i-PRO’s choice to run AI on the edge addresses both performance and trust concerns. Edge processing reduces cloud dependence and data exposure, while supporting real-time alerts “without the delays that can come with cloud processing.”
Standards also play a central role in building trust around AI adoption. Figols highlighted that adherence to ISO/IEC 42001, the AI management standard, gives customers confidence that the algorithms are trained, managed, and deployed responsibly.
AI for Broader Business Decision-Making
Milestone shared examples of how video intelligence is supporting strategic decisions across sectors. Their team described an increasing number of businesses using insights such as footfall, occupancy, routing patterns, and space utilisation to redesign store layouts, route pedestrian flows, optimise cleaning schedules, and reduce energy waste.
Milestone explained that video intelligence has “evolved far beyond traditional surveillance to add value across the business,” especially as organisations seek tools that can support different departments simultaneously.
This is reinforced by Milestone’s acquisition of BriefCam and Arcules, creating an analytics suite that spans everything from motion detection to advanced video synopsis and appearance similarity. The emphasis is on giving organisations a scalable ladder of capability without locking them into a single analytical approach.
One example of this is Project Hafnia, which redefines how AI models are trained for video analytics. Rather than relying on scraped or unverified data, Project Hafnia provides a curated, ethically sourced, and regulation-ready library of annotated video data. This enables computer vision developers to build and deploy AI solutions up to 30 times faster -while ensuring high accuracy, data integrity, and full compliance with privacy standards.
At Cathexis, the priority is clear: unify data from analytics, sensors, and alarms into a coordinated intelligence layer.
Cathexis engineers told us that “the real advantage comes from centralised coordination,” enabling systems to interpret behaviour in context. Their roadmap includes more predictive alerting, built on recognising patterns that fall outside normal activity rather than relying solely on event triggers. This aligns with the broader industry trend toward moving from reactive oversight to proactive intervention.
AI at Enterprise Scale
Eagle Eye Networks has been investing heavily in both cloud and hybrid AI models. According to Managing Director EMEA Rishi Lodhia, customers increasingly want AI-powered insights for workflow optimisation, compliance, and risk management. Their open cloud platform supports both edge-based analytics from partners such as Axis and Bosch as well as Eagle Eye’s own cloud-trained models like Precision People and Vehicle Detection.
Lodhia notes that this dual approach offers “faster insights, improved accuracy, and greater flexibility,” particularly for large, distributed enterprise or public sector deployments.
A Sector Moving Towards Predictive Intelligence
Taken together, these perspectives point to a market shifting decisively from simple detection to enriched interpretation. AI is becoming a partner in decision-making, whether the goal is speeding up an investigation, improving a store’s layout, supporting compliance, or protecting a remote perimeter.
Organisations are looking for tools that match their operational reality: more accurate imaging, faster search, contextual understanding, transparency in AI development, and a clear return on investment.
As AI becomes deeply embedded in daily operations, expectations around trust, transparency, and responsibility are rising just as quickly. The next article in this series explores how manufacturers are addressing those demands through cybersecurity, ethical AI development, evidential integrity, and environmental responsibility – and why these values now define how technology is judged in the security sector.