AI monitoring needs are reversing observability tool consolidation

AI observability is adding tools while agent coverage remains incomplete

New Relic reports that one in four AI agents operates without monitoring, leaving technology leaders responsible for systems whose behavior they cannot fully observe. Observability, the ability to monitor and understand how systems behave, has traditionally focused on cloud infrastructure and core business systems. AI workloads bring new requirements, so enterprises can find gaps in coverage despite their existing monitoring investments.

Enterprises are filling those gaps with AI-specific tools, reversing an earlier decline in the number of observability platforms they use. Yet New Relic reports lower outage costs as tool counts rise. So leaders deciding what to buy or consolidate need to assess coverage, platform count, and operating performance separately.

AI-specific demand is reversing the consolidation trend

New Relic’s report shows the shift: organizations reduced their observability tools from six in 2024 to 4.4 in 2025, before the average rose to five in 2026. The company attributes the increase to demand for AI-specific platforms. As an observability vendor, New Relic benefits commercially when enterprises spend more on monitoring, including capabilities for AI.

Demand for AI monitoring can complicate an established consolidation strategy. Conventional observability tools were designed around cloud performance and other essential systems, so they may leave some agent behavior uncovered. Enterprises must then decide whether their current platforms can meet the new requirements or whether they need another product. A specialized tool can close a gap while adding another system for teams to manage.

Buyers face that choice in an already crowded monitoring environment. Research by Enterprise Management Associates, or EMA, found that three in four enterprise IT organizations use up to 12 tools across cloud infrastructure, network, and service environments. The figure is an upper bound rather than an average, and it spans several parts of enterprise IT. Teams consolidating platforms therefore need to preserve coverage across those environments.

Even with those demands, the desire to consolidate remains strong: EMA also found that 62% of organizations consider observability-tool consolidation a top priority. EMA measures tool usage and purchasing priorities, while New Relic tracks changes in tool counts over time. These separate measures reveal a tension for buyers: an immediate AI-monitoring need can justify another purchase even when the broader strategy calls for fewer platforms.

Lower outage costs coexist with frequent high-impact failures

Tool counts alone cannot show whether an additional platform improves operations. New Relic reports that “early high-impact outage costs” fell from $76 million in 2025 to $73.5 million in 2026. It attributes the reduction to faster incident detection and resolution. Costs fell while tool counts rose, but that timing does not establish that adding tools caused the decline.

New Relic’s explanation centers on how teams respond to incidents. Earlier detection lets teams start work sooner, while faster resolution shortens the problem. Both can reduce the financial consequences of failures even when serious incidents remain frequent. So a platform change needs to be assessed for its effect on detection and resolution alongside any management complexity it adds.

Faster responses still leave the question of how often systems fail. New Relic found that more than one-third of organizations experience high-impact outages at least once a week. Frequent failures can keep consuming operational effort even as their reported financial impact falls. Teams therefore need to assess both how often incidents recur and how effectively they respond.

Agent coverage adds a separate measurement to that assessment. The unmonitored-agent figure counts agents, while the weekly outage finding counts organizations, so the two populations cannot be treated as interchangeable. Nor do the outage findings establish that AI agents caused those incidents. Coverage measures what an enterprise can observe; outage frequency and financial impact describe different aspects of what happens when systems fail.

Buying decisions must connect AI coverage with operating results

The need to monitor agents is prompting several kinds of vendor response. AWS, Snowflake, and Cisco each have a commercial interest in enterprises adopting their observability offerings, so buyers should view their launches and acquisitions in that context. Each approach offers a different way to obtain AI-monitoring capabilities, with different implications for the platforms enterprises already operate.

AWS is adding an agent-focused console: on Tuesday, it launched Amazon CloudWatch Omni, an AI-powered observability console for AI agents. Enterprises have another option for monitoring agents, but buyers still need to assess how it fits their environment. The console creates a purchasing option; its effect on tool count depends on how an enterprise adopts it.

Acquisition gives vendors another way to expand their capabilities. Earlier this year, Snowflake bought Observe, an AI-powered observability platform, to expand the insights available to AI-driven enterprises. The purchase brings an existing observability product into Snowflake’s business. For customers, the effect on platform count depends on how they use the resulting offering.

Vendors can also extend a product that customers already use. Cisco has added agent-observability capabilities to Splunk, giving customers AI-related monitoring within an established product. Enterprises using Splunk can therefore evaluate those capabilities as part of their current environment. They still need to determine whether the extension covers their agents and how it affects day-to-day management.

Meeting new requirements within an existing platform also underlies New Relic’s proposal. The company argues that unified platforms should absorb AI capabilities, allowing enterprises to obtain the functionality without adding tools. New Relic benefits if buyers choose its platform to consolidate that spending. The operational benefit depends on whether a platform provides the required AI coverage while limiting the complexity of managing additional products.

Buyers can test that proposition against operating measures. They need to establish which agents the platform can monitor, whether it preserves or improves incident detection and resolution, and whether adoption avoids additional tools and management work. Outage frequency and financial impact need separate checks because one can improve while problems persist in the other. A lower platform count helps when the resulting environment meets those operational requirements.

CIOs remain responsible for making AI work operationally

Those purchasing tests are part of the broader responsibility of chief information officers, or CIOs, to keep deployed technology working. “CIOs in particular have to always deal with actually making this stuff work,” Willy said. After a purchase, teams must run the selected platforms, monitor the agents they introduce, and respond to incidents.

Willy connects that continuing responsibility to earlier technology transitions: “It was very true in the early internet days, it was very true when CIOs were tasked with moving to the cloud and we’re in that period right now [with AI].” His comparison points to the recurring obligation to make new technology work in an enterprise. The systems change, but CIOs remain accountable for turning adoption decisions into working operations.

Key takeaways for decision-makers

• AI coverage complicates consolidation: New Relic reports that one in four AI agents operates without monitoring, while average observability tool counts rose from 4.4 to five in 2026. CIOs evaluating consolidation need to identify agent-monitoring gaps and determine whether existing platforms can close them.
• Outage costs and frequency need separate checks: Reported high-impact outage costs fell, yet more than one-third of organizations still experience high-impact outages at least weekly. IT operations teams need separate measures for incident frequency, detection speed, resolution time, and financial impact.
• Platform purchases need operational tests: Vendors are adding AI monitoring through new consoles, acquisitions, and product extensions. Buyers evaluating these offerings need to verify agent coverage, incident-response performance, and the management work each option adds.
• CIO accountability continues after deployment: CIOs remain responsible for making AI work within enterprise operations. Adoption plans need clear ownership for monitoring agents, running the selected platforms, and responding to incidents.