India leads in AI agent monitoring but pays the world's highest outage costs, New Relic finds

A large sample with a split picture

The report was produced with Enterprise Technology Research (ETR). It draws on 500 IT and engineering leaders from India, the largest single-country sample in this year's global survey of 2,575 respondents. New Relic describes the result as a paradox: India amplifies the main patterns seen in the global group, in both the positive and the negative direction.

According to New Relic, the country is a world leader in AI adoption, OpenTelemetry standardisation and return on observability, while also battling the heaviest operational strain and the steepest outage costs of any market surveyed.

AI agents and AI-generated code

On average, 56% of Indian enterprises actively monitor AI agents in production, compared with 35% globally. Just 17% of deployments are unmonitored. Indian organisations also lead globally in AIOps capabilities at 54% and in machine learning model monitoring at 57%.

AI-generated code is pushing observability up the agenda. In the survey, 80% of Indian leaders agree or strongly agree that the rise of AI-generated code makes robust observability more critical. 47% strongly agree, the highest figure in the global study.

AI coding tool use is higher in India than anywhere else, according to the report. GitHub Copilot is used by 54% of respondents, Google Gemini Code Assist by 51% and OpenAI Codex by 45%. Only 1% report using no AI tool. The effect on complexity is mixed: 56% say AI integrations have simplified stack management, while 39% say AI has added complexity.

Outages, OpenTelemetry and tool sprawl

The cost of downtime is where India stands out most negatively. High-impact outages occur multiple times per day for 31% of Indian organisations, a tenfold jump from 3% in 2025. The average revenue cost per hour of downtime has risen to $2.43 million, which puts the estimated cost of a single average high-impact outage at $3 million. The top three causes of unplanned outages were third-party or cloud providers, network failure and deploying software changes.

On OpenTelemetry, 42% of Indian organisations have fully standardised on it as their primary form of telemetry collection, nearly double the 22% global baseline. New Relic says this advantage is weakened by fragmentation. The average Indian enterprise now uses six observability tools, up from five last year, and 12% run ten or more.

The report links this to AI performance: standardising collection without consolidating tools can stop AI models from correlating insights. Only 20% of Indian leaders say AI-assisted observability has significantly improved event correlation, below the 27% global average.

What the findings suggest for 2026 and beyond

According to the survey, organisations in India that combine aggressive AI adoption and OpenTelemetry standardisation with unified platform visibility are significantly more likely to capture financial returns of 3x to 5x. Those that deploy AI agents across fragmented, multi-tool environments report a growing operational burden, spending half their engineering capacity on firefighting and facing the region's steepest outage costs.

For engineering teams, the message from the report is that early monitoring of AI agents and standard telemetry are not enough on their own. Consolidating tools appears to be the factor that separates strong returns from heavy firefighting. As the findings come from New Relic's own survey, they reflect the company's framing of the data. The full report is available for download from New Relic.