Fares Djenandji, chief growth officer at Ipsotek (an Eviden Business), explains how the next phase of technology will provide contextual operational intelligence

Airports are not short of data. Over the past decade, they have invested heavily in high-spec camera networks, video management systems, and increasingly sophisticated analytics.

Across terminals, airside operations, and security zones, there is little that cannot be seen.

And yet, despite this proliferation of visibility, many airport operations teams continue to grapple with the same challenges: delayed responses, fragmented workflows, alert fatigue, and missed opportunities to intervene before disruption escalates. The issue main issue remains a lack of understanding.

From “What Is Happening?” to “What Does It Mean?”

Traditional Vision AI has focused on identifying events: a queue forming, a vehicle entering a restricted zone, a passenger dwelling too long. These outputs have value, but they are inherently limited. They answer the question: what is happening?

For airport operations, however, that question is only the starting point. What matters more is: what does this mean operationally, and what should happen next? This is where a new class of AI capabilities is emerging.

Natural Language Detection (NLD), Vision Language Models (VLMs), and generative AI are beginning to act as a bridge between raw detection and operational understanding.

Rather than simply flagging events, these systems translate scenes into context, explain why they matter, and connect them to real-world airport processes.

A queue at security is no longer just a queue. It becomes a potential delay risk for specific flights, a pressure point for staffing decisions, and a signal that downstream congestion may follow.

The Limits of Detection-Driven Thinking

In response to operational complexity, many organisations have leaned into more AI models, more alerts, and more data streams. But this approach often compounds the problem rather than solving it.

More detection can create more noise. Without a layer of contextual understanding, scaling AI simply  increases the number of alerts that teams must interpret manually. The result is familiar across the industry: overwhelmed control rooms, competing priorities, and decisions made too late to have meaningful impact.

The real shift, therefore, is not about improving detection accuracy in isolation. It is about introducing intelligence that can interpret, prioritise, and synthesise. Vision AI is a foundational necessity, and it’s time to upgrade that foundation with what’s beyond the traditional approach.

Joining the Dots: From Siloed Systems to Operational Intelligence

Airports today operate across a complex ecosystem of systems: video platforms, Airport Operational Databases (AODB), flight schedules, passenger flow analytics, stand and gate management tools. Each system holds valuable data, but too often they exist in isolation.

This fragmentation is one of the biggest barriers to operational agility. True contextual AI emerges when these data sources are connected. When video insights are fused with flight timings, passenger volumes, and operational milestones, a far richer picture becomes possible; one that reflects how airports actually function as interconnected systems.

Vision Language Models are particularly powerful in this context. By combining visual understanding with language-based reasoning, they enable systems to interpret relationships across multiple inputs. They can link what is seen on camera with what is scheduled, expected, or at risk.

This transforms data into joined-up intelligence. And crucially, it makes that intelligence more accessible. Through natural language interfaces, operations teams can query complex environments in intuitive ways: Which flights are at risk due to current security queues? Where is congestion likely to impact turnaround times? The system does not just retrieve data; it explains it.

Agentic AI: Supporting, Not Replacing, Human Decision-Making

Alongside these developments, the concept of agentic AI is gaining traction. In aviation, however, it must be applied thoughtfully. Airports are safety-critical environments. Decision-making carries regulatory, operational, and human consequences. Full automation is neither realistic nor desirable in many scenarios.

The opportunity lies instead in assisted decision-making. Agentic systems can continuously monitor conditions, identify emerging risks earlier than human operators alone, and surface the most relevant insights at the right time.

They can help prioritise actions, highlight dependencies, and ensure that critical signals are not lost in the noise. But they stop short of removing human judgement. In practice, this means an operations team is not replaced by AI but augmented by it.

Teams are equipped with clearer, earlier, and more contextualised information, allowing them to act faster and with greater confidence.

From Alerts to Orchestrated Workflows

Perhaps the most significant shift on the horizon is the move beyond alerts altogether.

Today, most systems are designed to notify. Something happens, and an alert is triggered. What follows? Interpretation, escalation, coordination; is largely manual.

The next evolution is AI-driven workflow orchestration. In this model, insights do not simply inform; they initiate action. A developing congestion issue can automatically trigger coordination between security, terminal operations, and airline teams.

A potential delay risk can prompt reallocation of resources or adjustments to boarding processes. Generative AI plays a key role here by translating complex, multi-source data into clear, actionable steps embedded within operational workflows.

It becomes the interface between insight and execution. The result is not just faster awareness, but faster response.

Why This Matters for Airport Performance

For airport leaders, the implications are tangible. Improved contextual intelligence directly impacts key operational priorities: reducing delays, improving turnaround efficiency, optimising passenger flow, and enabling better coordination across teams.

These are not abstract benefits. They translate into measurable outcomes: fewer missed connections, smoother peak-hour performance, more resilient operations during disruption.

Importantly, they also shift the role of technology within the airport. AI moves from being a peripheral tool, often confined to security or monitoring, to a central enabler of operational performance.

A New Measure of Maturity

As airports continue to invest in digital transformation, a key question emerges: what defines maturity in Vision AI? Historically, the answer may have been the scale of camera infrastructure or the sophistication of detection models.

Going forward, that will change. The airports that lead will not be those with the most cameras, but those that can turn data into timely decisions. They will be the ones that can understand not just what is happening across their environment, but why it matters, what it affects, and what should happen next.

This requires systems that can interpret context, connect disparate data, and support action; not just generate alerts.

Conclusion

The evolution of Vision AI in aviation is entering a new phase. One defined not only by incremental improvements in detection, but by a fundamental shift towards contextual, operational intelligence.

Natural language interfaces, Vision Language Models, and agentic AI are reshaping how information is understood and used within the airport environment. Together, they offer a path towards more proactive, coordinated, and effective operations.

For an industry where timing is critical and margins for disruption are small, that shift may prove transformative. The question is no longer how much data an airport can collect. It is how effectively it can understand, and act on, what that data is telling it.