I have been building enterprise software for more than 20 years, at BMC Software, CA Technologies, Splunk, now Apica. One thing stays true across all of it: You do not find out your foundation is wrong during the crisis. You find out when the debt comes due. For a lot of organizations, that bill is arriving now.
New research from Omdia, surveying 300+ enterprise IT decision-makers across North America and Western Europe, puts hard numbers on something practitioners have been sensing for a while. The telemetry problem isn't coming. It's already here.
The Volume Numbers Are Not Projections
54% of enterprises saw their telemetry data volume triple in the past 12 months. Not projected. Not modeled. Running right now, in the environments they operate today. AI and machine learning workloads account for roughly 43% of that growth, making AI the single largest driver of telemetry volume in enterprise environments.
And this isn't the peak. Keir Walker, Senior Market Research Analyst at Omdia, said it directly: "This is just the beginning of the growth curve. We are at the beginning of the hockey stick."
Survey respondents expect an average 9.5x increase in telemetry from agentic AI workloads within two years. I'd pay less attention to that average than to the spread: 44% of respondents expect growth somewhere between 6x and 100x. When your most experienced enterprise IT leaders can't put a ceiling on expected data volume, the architecture has to be built for extremes, not averages.
The Cost Surprise Isn't Compute
I have been in technology long enough to know that cost surprises don't come from where you expect them. For agentic AI, the surprise isn't compute. It isn't talent. It's observability.
In 69% of agentic AI projects, observability costs already exceed compute and infrastructure costs combined. The average enterprise spends $3.17M annually on observability. Nearly 20% spend more than $5M. Budgets are growing 28% year over year on average, and for more than a third of enterprises that growth exceeds 52% annually.
Those numbers have started forcing real decisions. 59% of organizations have terminated or delayed at least one agentic AI deployment because monitoring costs were too high. And the agents getting shelved aren't experimental ones: Cybersecurity, legal and compliance, and fraud detection lead the list. The cost of leaving those agents unmonitored isn't measured in IT budget. It's measured in business risk.
Torsten Volk, Principal Analyst at Omdia, put it plainly: "Scalability is the main reason why people can't have agentic projects. They have no way of deploying them without exposing themselves to operational, legal, and security risk."
That's not an IT problem. That's a C-suite problem.
35% Deployment Is Not What It Sounds Like
The research shows 35% of enterprises claiming widespread agentic AI deployment, which is striking given the category has only existed since 2024. The Omdia analysts were direct about what's behind that number: Competitive pressure, not infrastructure readiness. Every CEO is saying the company has to be agentic. The pipelines underneath those agents weren't built for what the agents actually generate.
The gap is measurable. Organizations unfamiliar with agentic AI are 4.5x less likely to be prepared for the data volumes it produces. The organizations most exposed are the ones least likely to know it.
What the Prepared Organizations Did Differently
The research is consistent here. 97% of enterprises have implemented or are actively evaluating a telemetry pipeline solution, routing, filtering, governing, and contextualizing telemetry before it reaches any downstream observability or analytics platform. That's not an emerging concept, that's the market's answer to the problem.
Pipeline adopters are 50% more likely to be prepared for agentic AI data scale within 24 months. They're 80% more likely to have avoided the operational cost problems that are stalling everyone else. That second number matters. Pipeline adoption isn't correlated with readiness by coincidence. The organizations that scaled successfully built the pipeline layer first.
68% of enterprises plan to evaluate changes to their observability solutions within six months. 70% plan to evaluate pipeline solutions in the same window. Most organizations reading this are already in or entering that cycle.
In 20+ years of enterprise infrastructure, the pattern holds. The organizations that come out ahead in a technology shift are the ones that fix the infrastructure problem before it becomes a budget crisis.
The data from this study is not a warning about a future problem. It's a measurement of a present one.