Intelligent video is rewriting the production rulebook

A single AV manager at a mid-sized university is asked to live-stream graduation to thousands of remote families, record the full event for on-demand, and deliver a polished multi-camera production, all with the same budget as last year.

That brief is no longer unusual. Enterprise teams, universities, healthcare organisations, houses of worship, and live event venues are all being simultaneously asked to serve in-room and remote audiences. The production bar has risen sharply. Staffing and budgets, in most cases, have not.

For smaller AV and production teams, the pressure is structural. One operator may now be expected to manage cameras, switching, audio, graphics, recording, streaming, distribution, and IT support. The workflow has simply become too complex for the number of people running it.

Where automation is genuinely helping

Intelligent video is beginning to make a measurable difference, not by replacing operators, but by handling the mechanical tasks that consume their attention.

Auto-framing can keep a presenter composed. Voice-tracking can help cameras follow discussion in a seminar or classroom. Computer vision can detect when a speaker leaves a podium, flag a missing safety vest on a factory floor, or turn a scoreboard into a production input.

Automation works best when the task is clear: keep this person in frame, tag this moment, flag this anomaly. That clarity also defines its limits. New visual reasoning models are emerging that enable cameras to not only capture what is happening, but understand it and respond in real time, such as adjusting a preset when a speaker moves to the whiteboard. Higher order context-level understanding, like reading the mood of a room, isn’t quite there yet.

The agentic AI moment, and its limits

Smaller production and AV teams are now experimenting with agentic AI: systems that perceive a production environment, maintain context across a session, and take goal-oriented actions within boundaries set by humans such as preparing a layout change, tagging a moment for replay, or alerting an operator to something that needs attention.

Our vision is that AI and people each do what they do best, reaching outcomes neither could deliver alone. Agentic AI is not a producer. It does not understand the emotional arc of an event, when to hold a wide shot, or when the technically correct cut is not the right creative choice. It can also fail when lighting is poor, camera positions are inconsistent, or there is no clear fallback in the workflow.

Advertisement

Automation depends on good production discipline. Standardized camera positions, clean audio, reliable networking, and clear override controls must be in place first. Without that foundation, AI becomes another layer of complexity rather than a source of relief. The organisations getting the most from intelligent video have not automated the most, instead they have identified one specific operational problem and solved it well.

What the vendor community owes these teams

The shift in production expectations also changes what organizations should demand from their technology partners. The test for any tool is simple: Does it demonstrably improve on what the team is already doing?

Systems should connect cameras, audio, control platforms, production software, and management tools without adding integration complexity, and they should fit real rooms with real staffing constraints, rather than the idealized room in the product demo.

Affordability matters for the same reason. Capability that a team cannot staff or budget for is not capability; it’s brochure copy. Browser-based control that runs on any device without dedicated hardware is one direction that brings intelligent tracking and other software-defined features within reach of enterprise and education teams that could never justify traditional broadcast infrastructure. The specific products matter less than the principle: The work should meet the team where it already operates.

Most important, and least common, is honesty. A vendor that details the conditions under which a feature fails, such as poor lighting, inconsistent camera positions, unplanned presets, or no clear fallback, is giving you something more valuable than the one that just shows the demo.

Disclosure is itself a feature. It is what lets an operator design a workflow that holds up on the night instead of one that holds up in the lab. The industry has a habit of leading with capability and staying quiet about preconditions. For the AV manager running that graduation ceremony alone, the gap between the two is the difference between a successful event and a very difficult conversation with the vice-chancellor.

The future is better orchestration

The future of production beyond broadcast will not be defined by whether AI can run a show by itself, but by whether intelligent systems can remove friction from the work humans are already doing.

Enterprise, education, healthcare, manufacturing, and live events are all asking how to deliver more broadcast-quality production with a flat budget. The answer is not one technology. It is better orchestration between cameras, audio, software, control systems, AI, and the production teams at the heart of it all.

As workflows become more adaptive and connected, intelligent video is the next major step in making production systems more aware and easier to manage.

Broadcast taught the industry the language of production quality. The industry’s job now is to make sure that language is spoken fluently far beyond the broadcast studio.

Advertisement

Matthew Davis, PTZOpticsMatthew Davis is the chief technology officer of PTZOptics.