Catching AV failures in 3 minutes across thousands of rooms — Innomate on Elastic Observability

Australian workplace-technology company Innomate uses Elasticsearch and Elastic Observability to monitor thousands of audiovisual (AV) spaces across universities and enterprises. Its Innomesh platform flags faults within three minutes and fixes many of them automatically, often before a class or meeting even starts.

Before Innomesh, managing AV across a large organization was essentially a waiting game. A lecturer walked into a room and, if the projector didn't work, they reported it and waited for a support technician to drive over and fix things on-site. By then, the class would have already started, if not finished. Plus, pulling meaningful data out of AV systems was an afterthought.

Innomate built Innomesh to change that. It's a SaaS platform that lets universities, government bodies, hospitals, and enterprise teams manage every screen, camera, and mic across thousands of rooms the same way IT manages servers — centrally and proactively, and with real data. Elastic is the backbone that makes it possible: Telemetry from every device flows into Elasticsearch, surfacing faults within minutes. Dashboards give both technicians and executives a single view of what's failing and what's being fixed — as well as what's sitting unused.

The latest Innomesh innovation is meshIQ: an AI layer built on top of the data that lets teams ask plain-language questions and understand root causes and trends. Now, instead of triaging a wall of notifications, a technician can ask "Why did six devices in that room drop last night?" and get a real answer. The shift is from reactive break-fix to infrastructure that resolves many common faults on its own, and it's all running on Elastic under the hood. Here's how.

Picture a lecturer walking into a theatre at 9:00 a.m. The screen is black, there's no sound — and in the audience, 40 students are watching and waiting for their class to start. By the time a technician has been called, driven over, and diagnosed the issue, the class is already running 20 minutes behind.

This was just how AV worked. And for a long time, it was because there was no other option.

The consequences rippled across the whole organization. Technicians were dispatched on-site to diagnose issues that could have been caught hours earlier, and service owners had no real visibility into whether their spaces were ready. Plus, executives who were tasked with making decisions about where to invest in technology were working from instinct rather than data.

Innomate, an Australian company building cloud-native AV and Unified Communications (UC) management software, decided to tackle things head on.

The company built Innomesh to do for AV what IT teams had been doing for servers and networks for years: Monitor it all centrally, catch problems before they surfaced, and fix them before anyone noticed. Innomate's customers are largely universities and government bodies, as well as healthcare institutions, including four of Australia's Group of Eight. To give you an example of the scale of what was being done manually: A single university could run 1,500 spaces, and every part of incident management for those spaces was being managed on-site, as they happened.

Innomesh was born from a desire to see a data platform that could do two things at once: hold data across an entire estate, and pinpoint a single failing device in a single room. The team evaluated several platforms but eventually chose to migrate to Elastic for its advanced capabilities. Elastic has been the backbone of the platform ever since.

The deciding factor early on was the economics of scale. Most alternatives Innomate modeled were priced on data volume — inbound and outbound — and came out too expensive to make the business work.

While this might seem on the surface a cost issue, the real payoff is fidelity. When a business truly understands its cost structure, it’s able to keep full telemetry flowing from every space rather than sampling data down to control spend. Full telemetry is what makes proactive detection and, later, AI-assisted root-cause analysis possible at all.

Innomate's other founding principle is vendor agnosticism. Rather than limiting customers to using the manufacturer's proprietary control system, the team writes drivers to communicate with new equipment so a mixed estate can be unified. It's why a single Elastic-backed back end (rather than a closed manufacturer tool) sits at the center of the architecture.

Innomesh runs as a multitenant SaaS on Elastic Cloud, with each customer's estate onboarded as its own tenancy. The end-to-end flow is short and deliberate.

Devices and sensors across every managed space emit telemetry on state, usage, and configurations. That telemetry is ingested into Elasticsearch, which serves as the single back end for the platform. On top of that, Elastic Observability drives live alerting and the dashboards that technicians and executives work from.

Remediation policies can then act on the fault directly, such as rebooting devices, without anyone needing to be on-site. Separately, meshIQ, Innomate's AI-assisted technology, draws on the same Elastic data to answer natural-language questions and surface root causes.

Consolidating alerting and visualization alongside the AI layer onto one Elasticsearch back end is the structural choice that makes everything work. There's no separate metrics store, no second system to reconcile, and no glue code fanning queries across tools and merging results in application logic. In fact, Innomate uses the phrase "single pane of glass" to describe what customers see.

For its teams, correlation is possible because everything lives in one store and is queried one way.

Technical highlights

• Data backbone: Elasticsearch as the single store for all AV and IoT telemetry
• Alerting: Live alerting on Elastic Observability with a three-minute fault-to-alert time across the estate
• Multitenant SaaS: Each customer estate onboarded as its own Elastic tenancy on Elastic Cloud
• Data sources: Control systems, displays, audio, lighting, and environmental sensors per space
• AI layer: meshIQ on AWS Bedrock, querying Elastic data for insights

3-minute alerting across the estate

The problem live alerting solves is latency of awareness. Across large estates of over 1,500 spaces and 14,500 devices, a fault that goes unnoticed until a class starts is a fault that has already cost a lecture. Innomesh runs its alerting on Elastic so that, from the moment a fault condition arises, an alert lands on the operations dashboard within about three minutes. That window is what converts "we find out when someone complains" into "we know before they walk in," and it is the number the whole operating model is built around.

Proactive remediation through policies

Detection is only half of it. Innomesh pairs alerting with remediation policies that act on a detected fault automatically, so that issues are resolved before they affect the room. The same control that lets the platform remediate a fault also drives standardization gains — because Innomesh controls the systems, it can power down lighting and idle AV equipment out of hours, reducing energy usage.

Executive insight dashboards

Beyond keeping rooms working, Innomesh turns telemetry into decisions. Elastic-backed dashboards show executives how spaces are actually used through the day and how much energy the estate consumes, so technology and facilities budgets can follow real utilization rather than assumption. So, for many institutions, AV data has become an IT and business asset rather than a maintenance byproduct.

meshIQ: Natural-language questions on live data

meshIQ is Innomesh’s AI layer. It runs on AWS Bedrock but draws entirely on the telemetry Innomesh holds in Elastic, so a user can ask a plain-language question and get an answer grounded in the estate's own live data. In practice a technician can ask what the likely root causes of a fault are, what could be done now to fix it, and what would prevent it recurring, and IQ reasons over the collected data to respond. Users can also generate their own dashboards and draw their own insights without waiting on a build. Because the AI enters every question with the full picture already in Elastic, it starts from context rather than from a blank prompt.

The clearest way to see what changed is a normal morning. A sensor in a teaching space reports a display or control fault overnight. Within about three minutes the alert is on the service owner's dashboard, and in many cases a remediation policy has already acted on it. 

The on-site technician's first look at the estate is a board that says what needs a human and what has already handled itself, rather than a phone ringing at 8:55 a.m. with a lecturer standing in a dark room. When something does need judgment, the technician can ask Innomesh IQ what the data suggests the cause is before deciding what to do. Of course, a person still makes the call, but the platform makes sure they make it early, with the estate's own data in front of them.

According to the Innomate team, this proactive model is only the beginning of what they plan to do. The team is migrating its Elastic usage to Serverless, a consumption-based model that suits how the platform now scales better than the fixed footprint that served the startup years. It keeps the cost discipline that drew Innomate to Elastic in the first place: Spend now tracks customer growth, so each new estate adds a proportionate, predictable cost rather than a step change in capacity.

meshIQ is set to grow from answering questions toward acting on them, letting users instruct the system to carry out changes rather than just surface them. And the operating model that Innomate proved across four of Australia's Group of Eight universities is now going global: The company is active in the UK and US, targeting Russell Group and Ivy League institutions, and opening a corporate vertical where the same telemetry drives the same proactive operations.

If your team is still running on reactive support tickets and gut feel, Innomate's story is a useful benchmark for what's possible when full telemetry meets a proactive operating model. Learn how Elastic powers it.

See how Elastic Observability turns full-fidelity telemetry into proactive, automated operations, or start now with a free trial.