events without disclosing any private information. This approach could advance AI across the industry while still respecting competitive boundaries. VIAVI: Based on the above reasoning, past experience is very valuable and if this can be pooled and federated into a common set of experiences / learnings it would be beneficial for the overall industry, it is though not clear what party would be the right one for such a project as it entails both sharing of internal data and validating that the local learnings are generally applicable.
Will SaaS become the dominant model or are‘ in-network’ monitors still vital? Accedo: Hybrid models will dominate, with in-network monitors remaining critical for specific scenarios. SaaS is growing rapidly due to faster deployment, lower upfront costs, continuous updates, and reduced maintenance burden. Cloud-based AI models can be updated continuously, provide scalable compute for training, and cross-customer learning offers value with appropriate privacy controls. SaaS platforms excel at integrating diverse data sources, and cloudnative architectures enable rapid feature development. However, we observe that a significant portion of deployments still require hybrid or on-premise models, driven primarily by latency requirements, compliance needs, or scale considerations. Aprecomm: While SaaS is rapidly emerging as the preferred model for AI in Telecom and Media, in-network monitoring remains a vital component of effective service management. SaaS platforms offer scalability, agility, and cost efficiency, enabling operators to deploy advanced AI capabilities quickly and benefit from continuous innovation. They also support centralised learning, where insights from multiple networks can be pooled to enhance predictive performance and customer experience management. However, not everything can move to the cloud. In-network monitors continue to play a crucial role by providing real-time visibility, precision, and control necessary for critical operations. Bitmovin: SaaS has already become the preferred model for many testing and monitoring deployments because it delivers scalability, faster updates, and easier integration across complex streaming ecosystems. Cloud-based tools make it possible to deploy new AI models rapidly, centralise observability, and lower operational overhead, a trend we’ ve experienced firsthand across large-scale streaming deployments. That said, in-network monitors remain essential, especially in latency-sensitive or securityconscious environments where operators need immediate visibility within their own infrastructure.
10 EUROMEDIA
Bridge: Context is key. For many operations, dedicated appliances or embedded probes remain indispensable, especially when monitoring the full broadcast chain at the network level. But flexibility is essential too. That’ s why we’ ve introduced containerised services that can be accessed remotely, giving customers the option to scale up capacity or functionality when needed- for instance, during large-scale global sports productions. SaaS will continue to grow in prevalence, but it’ s not yet at the point of dominance. The future will be hybrid: appliance, embedded, software and cloud-based probes coexisting, each serving its purpose where it fits best. Interra Systems: From a broadcast and OTT operations perspective, the choice between SaaS-based solutions and in-network systems depends largely on what is being monitored and the specific monitoring objectives. SaaS is particularly effective when remote data collection is just as reliable as local collection. In the case of OTT monitoring, where the entire workflow resides in the cloud, Saas offers faster deployment, increased scalability, and lower upfront costs— making it an attractive option for monitoring applications distributed across multiple locations. However, for security-sensitive environments and specialised signal monitoring— such as RF signals or transport streams with protocols like RTP, SDI, or ST-2110— the monitoring device usually needs to be connected within the network to access the original signal accurately. Leader: We see both models coexisting. Cloud and SaaS-based monitoring offer scalability, particularly for distributed or remote production, and simplify the deployment of analytics and updates. However,‘ in-network or edge-based instruments remain critical where timing precision, signal fidelity, and immediate operator feedback are required- for example, in live production trucks or control rooms. Leader continues to design products that can operate either locally or integrate seamlessly with cloud-based monitoring and orchestration platforms. Telestream: Most broadcasters and service providers are predominantly producing and delivering video services using fixed infrastructure with dedicated hardware executing single functions. The transition to private or public cloud infrastructure is not a six-month project, which requires multiple deployment models to accommodate the time horizon. Torque: I am now seeing a slow move away from the cloud back to on-prem systems. Cloud offers freedom of management and the ability to quickly deploy or retire services, but I believe that operators have figured out that cloud is very expensive in the long run. Further, since operating profits continue to be squeezed, services are moving back from the cloud. Therefore, if part or all of the service is operating on-prem, there are limitations to what a cloud-based SaaS T & M tool can provide. The only solution is to deploy onprem. Do note, however, that those in-network monitors could be software with a monthly subscription, or could bespoke hardware devices. Both are equally likely. VeEX: SaaS is growing rapidly, as it offers scalability, flexibility, centralised management, and secure access from virtually anywhere. However, there are still cases where innetwork monitoring remains essential, such as compliance, security, or real-time diagnostics that don’ t rely on latency or other issues from the SaaS connection. So, it really depends on the use case, and in a lot of systems, the approach is hybrid, combining both SaaS and in-network monitoring. VIAVI: This is very much a question of the security model of the network and the regulatory framework. Certain segments will be almost 100 % SaaS and others close to 0 % due to their regulatory and security models as well as the overall organisational and service design principles.
Do AI applications reduce the real cost of providing T & M( customers will presume that it does)? Accedo: Yes, but cost reduction comes primarily through efficiency gains rather than eliminating T & M entirely.
This is one of the most important questions customers ask. AI does reduce test & measurement costs, but perhaps not in the way customers initially assume. In operational efficiency, manual intervention has decreased significantly, mean time to resolution has dropped substantially, and false positive alerts have been reduced dramatically. Self-healing capabilities eliminate many support tickets that would have required manual handling. In infrastructure optimisation, intelligent resource allocation helps customers rightsize capacity more accurately, and automated scaling reduces waste from over-provisioning.
More importantly, there’ s business impact prevention. Churn prevention systems can identify at-risk users and intervene proactively. By preventing revenue loss from service degradation and reducing negative user experiences, business value becomes even more significant.
But costs don’ t disappear, they shift. AI infrastructure itself has costs, initial integration requires investment, and ongoing model maintenance, data quality assurance, and performance monitoring also require resources. The key is managing customer expectations: AI doesn’ t eliminate test & measurement, it makes it dramatically more efficient and valuable. Cost structure shifts