Enterprises Spending $2.5 Million A Year On Log Tools While Discarding 86 Per Cent Of Their Data

Enterprises are spending an average of nearly US$2.5 million annually on logging solutions while excluding 86 per cent of their log data just to keep costs and systems under control, according to new research from Dynatrace that argues AI workloads are breaking traditional approaches to log management.

The State of Log Management 2026 report, based on a global survey of 450 senior technology leaders at enterprises with annual revenues of $750 million or more, found that AI workloads have driven a 93 per cent increase in log and telemetry volume over the past 12 months.

Despite that surge in data, organisations are being forced to throw most of it away. Nearly half of organisations report discarding or not collecting logs, excluding an average of 86 per cent of log data from ingestion, storage or analysis to manage cost and system limitations.

The result, according to Dynatrace, is that teams are flying blind at the point when AI systems demand more visibility to be explainable and trustworthy.

Seven Tools And Manual Correlation

The research found that organisations rely on an average of seven different tools to manage logs and telemetry, forcing teams to manually correlate data across systems. That manual correlation consumes an average of 58 per cent of the time teams spend analysing cloud and AI telemetry.

The consequences are tangible. Eighty per cent of respondents reported that the time and effort needed to turn telemetry into actionable insights is negatively impacting customer experience and delaying AI initiatives from moving beyond pilot to production.

Mala Pillutla, Vice President of Log Management at Dynatrace, argued that existing logging systems were never designed for the demands of AI-driven environments.

“AI is accelerating enterprise innovation, but most logging systems were never built for the scale, speed, or complexity of AI-driven environments,” Pillutla stated.

“As AI agents operate probabilistically, treating logs, metrics, traces and events as separate signals is no longer viable. To make AI systems reliable and trustworthy, organisations need a unified, intelligent approach that brings all telemetry together in real time, enriched with deep context to drive confident decisions.”

The Economics Of Traditional Logging Under Pressure

The report paints a picture of an industry caught between rising data volumes and the economics of traditional log management.

Respondents estimated they spend an average of nearly US$2.5 million annually on logging solutions, covering ingestion, management, storage, indexing, rehydration and querying. Existing log tools now consume nearly half of observability and monitoring budgets, according to the survey.

Three-quarters of organisations reported higher log management costs over the past year, with 35 per cent seeing a significant increase. A further 87 per cent expected AI to increase their log management costs in the current year.

To manage those costs, organisations are resorting to blunt measures. The most common approach was limiting storage duration, used by 57 per cent of respondents, followed by discarding or not collecting some logs at 50 per cent, sampling a limited set of commonly used logs at 43 per cent, limiting queries at 38 per cent and limiting investigations to avoid the cost of rehydrating logs in cold storage at 34 per cent.

In 67 per cent of organisations, the cost of the existing log management approach now outweighs its value, the survey found. Eighty per cent reported that traditional approaches and legacy tiered architectures cannot keep pace with AI workloads, while 74 per cent indicated that indexing and rehydration costs and delays make it difficult to extract value from the logs they do ingest.

Fragmented Telemetry Stalling AI Transformation

The challenge extends beyond cost. The report found that fragmented telemetry is directly stalling AI transformation efforts.

Seventy-one per cent of respondents reported difficulty collecting and correlating AI health metrics from multiple sources. Eighty-two per cent indicated that AI workload data has increased the difficulty of stitching together fragmented telemetry. And 76 per cent acknowledged that traditional log management makes analysis difficult, with AI workloads making the problem worse.

Eighty-five per cent of organisations found it difficult to ingest logs at the scale driven by AI workloads, while 79 per cent of technical leaders worried that their current ingest and storage approaches will not meet future needs.

The report noted that 80 per cent of organisations believe they must update their instrumentation practices to support new AI-specific telemetry, pointing to standards such as OpenTelemetry Generative AI Semantic Conventions as an example of the direction the industry is heading.

Logs Remain Critical But Insufficient Alone

Despite the challenges, logs remain central to how organisations monitor their systems. Eighty-seven per cent of organisations rely on logs for insight into their AI applications, while 22 per cent depend on logs exclusively for collecting and correlating AI health metrics.

However, 73 per cent of respondents acknowledged that logs only reveal part of what is happening with their AI workloads. Sixty-five per cent reported relying heavily on logs because they lack clear visibility into other telemetry signals.

The survey found that organisations rank traces as the top telemetry source for evaluating AI performance and behaviour at 70 per cent, followed by logs at 61 per cent and metrics at 54 per cent. The report argued that all three together, in context with business KPIs, AI model metrics and user feedback, are needed for end-to-end observability of AI applications.

The most common use case for logs remains monitoring conventional applications that do not involve AI, cited by 58 per cent of respondents, followed by security investigations at 54 per cent and monitoring AI workloads at 46 per cent.

Platform Approach Demanded

The report found strong demand for a shift from standalone logging tools to platform-based approaches. Seventy-two per cent of respondents indicated that AI workloads demand a platform approach to log management that combines all types of telemetry in one place.

Eighty-one per cent believe log ingestion and processing must be open and automated for real-time analysis without rigid schemas, indexing overhead or rehydration delays.

When asked which capabilities would have the greatest impact on accelerating AI workload management, 62 per cent of respondents pointed to automated correlation of logs, traces and metrics. Real-time analytics across all telemetry in full context was cited by 57 per cent, dynamic storage that scales to any volume without indexing by 48 per cent and explainability of AI decisions by 35 per cent.

The explainability question is particularly pressing. Eighty-three per cent of respondents identified explainability as essential to managing AI workloads, while 84 per cent indicated that customer trust in AI depends on the ability to use log analytics to predict and prevent problems. The top challenge in instrumenting cloud and AI workloads was limited visibility into “black box” AI models, cited by 56 per cent of respondents.

Regional And Industry Differences

The report included regional breakdowns across the United States, Europe and Asia Pacific. US respondents reported the largest increase in logs and telemetry from AI workloads at 99 per cent, compared with 90 per cent in Europe and 89 per cent in APAC.

APAC organisations have the largest annual budgets for log management solutions, spending US$2.61 million on average, compared with US$2.47 million in the US and US$2.34 million in Europe. However, companies in APAC are the most likely at 74 per cent to feel that the cost of their current log management approach outweighs its benefits, compared with 69 per cent in Europe and 57 per cent in the US.

APAC respondents were also the most likely to rely heavily on logs because they lack visibility into other telemetry signals, at 74 per cent compared with 61 per cent in the US and 58 per cent in Europe. Organisations in APAC were the most likely at 67 per cent to limit storage duration to control costs.

Sixty per cent of organisations globally are moving away from traditional log management tools because they do not meet the needs of AI and cloud-native workloads, with the rate higher in APAC at 66 per cent than in Europe at 60 per cent or the US at 54 per cent.

By industry, healthcare organisations have the highest annual budget for log management solutions at US$3.9 million, followed by software and digital at US$3.47 million and financial services at US$2.99 million. Software and digital companies have seen the largest increase in log and telemetry volume from AI at 127 per cent, followed by telecommunications at 125 per cent and energy and utilities at 123 per cent.

Redundant Spending Compounds The Problem

The report also identified waste within existing observability spending. Roughly a third of organisations are paying for redundant or underutilised observability features, and more than a quarter are consuming engineering cycles keeping multiple tools running across environments.

Dynatrace argued that this capacity should be directed toward making AI workloads production-ready rather than maintaining fragmented tooling. The report characterised the real cost of observability fragmentation not as the infrastructure bill itself, but as the opportunity cost of AI initiatives that stall between pilot and production because teams cannot trust their telemetry.

The research was conducted by Coleman Parkes on behalf of Dynatrace in January and February 2026. Respondents included CIOs, CTOs, Heads of IT, Senior IT Managers, Platform Engineering Managers and Site Reliability Engineering Managers across the US, Europe and APAC, including respondents in Australia, New Zealand, Singapore, India, Thailand and Malaysia.

Last Updated on June 18, 2026 by Nick Ross