Integrators know the drill: a client wants a conference room upgrade, you identify a powerful codec with AI features, and it works well at first. But when the client’s needs change, the system can’t adapt, leading to costly replacements.
The AV industry has spent the last few years seeking “smarter codecs,” packing more AI, better processors and additional features into centralized hardware boxes. Vendors market these upgrades as transformational, and they do deliver real improvements over previous generations. But is the centralized codec architecture itself still the right foundation for modern meeting spaces?
I don’t believe it is. The real transformation in professional AV will come from moving beyond the codec and distributing intelligence across every device in a room.
How the Codec Became the Default Infrastructure
The centralized codec is largely a relic of its time. When video collaboration systems were developed, mobile processors were weak, networking was unreliable, and AI didn’t exist in any practical form. The elegant solution was to build a single powerful brain that handled all compute, logic, and processing. Cameras, microphones, speakers, and displays connected as peripherals through custom cabling, and everything ran through that central unit.
This architecture solved enormously complex problems and pushed the industry forward. I spent years in this world myself while at Cisco. We found that the approach worked because it matched the technology constraints of the time.
But those constraints have disappeared. Since the late 2000s, four massive technology shifts have changed what’s possible:
- Mobile processors now deliver ten times the performance at a fraction of the power consumption.
- Modern software architecture enables distributed services instead of monolithic applications.
- AI has become practical and pervasive.
- IP networking with Power over Ethernet has replaced custom cabling.
Yet many in the AV industry have responded by rebuilding the old architecture with newer components rather than rethinking the foundation.
The Limits of Centralization
A centralized codec — even a “smart” one — contains inherent limitations that no amount of processing power or AI can overcome.
First, capacity is fixed at installation. You choose a box based on room size and anticipated use. If the space changes, if teams grow, if the room gets adapted for different meeting types, the system can’t flex. You’re locked into the capabilities you purchased, and scaling up means replacing expensive core hardware.
Second, installations remain complex. Proprietary connectors, custom cabling runs and detailed programming require specialized knowledge and increased costs. Even when vendors simplify interfaces, the underlying infrastructure stays rigid.
Third, the experience is static. Traditional codec systems rely on presets and programming. Users have to adjust to the system rather than the system adapting to them. Adding AI features helps, but the core limitation remains: all intelligence flows through a single point.
Finally, these systems don’t scale as modern IT infrastructure does. When a company needs more network capacity, they add switches and access points. When they need more compute, they spin up additional servers or cloud resources. But when they need more AV capability in a centralized codec setup, they face a forklift upgrade. Fundamentally, complexity does not scale. Simplicity does.
The Alternative: Distributed Intelligence
Imagine if rooms could scale fluidly by adding or removing components as needs changed. This is possible if AV manufacturers move towards embedding processing, AI, and sensing capabilities into every device in a meeting room, so compute is coordinated across devices rather than centralized in one box.
With a distributed model, instead of peripherals connecting to a “brain,” you deploy intelligent devices that work together over standard IP networks. A camera, rather than a codec, processes its own framing and tracks participants, while a microphone array handles its own beamforming and noise suppression. Each component brings intelligence to the system, and they connect using the network.
This architecture delivers multiple practical advantages for integrators and end users because rooms become modular. Need more microphone coverage? Add another intelligent mic array. Want an additional camera angle? Drop in another camera. The system recognizes new devices and integrates them without reprogramming. Installations are simplified because the entire system operates over the network, with no proprietary cabling diagrams or custom connectors required. Costs are distributed across components rather than concentrated in one expensive box, making incremental upgrades more feasible.
Most importantly, the experience becomes adaptive. Instead of relying on presets, distributed systems can read the room contextually, responding to how many people are present, where they’re sitting, who’s speaking and how the meeting is unfolding.
What This Means for Integrators
For integrators, distributed architecture changes the conversation with clients. Instead of sizing a room to a codec’s fixed capacity, you can design systems that start simple and grow organically. Rather than deal with complex programming and custom cabling, you can deploy network-based systems that leverage standard IT infrastructure. When a space needs a change, you can add or swap components incrementally instead of making expensive forklift upgrades.
This doesn’t mean every room needs distributed architecture immediately. Smaller spaces with stable requirements may continue to work well with simpler, all-in-one devices. But for larger or more dynamic environments — spaces that might be reconfigured, split into breakout rooms, or need to support varying group sizes — distributed intelligence provides the flexibility that centralized boxes can’t match.
The shift similarly aligns AV more closely with IT planning. When meeting room technology runs over the same IP infrastructure as everything else, it becomes easier to manage, monitor, and integrate with larger workplace systems. That convergence has been discussed for years, but distributed architecture makes it practical rather than aspirational. For an industry built on solving complex implementation issues and delivering reliable experiences, that seems like an opportunity worth pursuing.
Tormod Ree is Chief Product and Engineering Officer at Neat.
