Businesses are betting big on AIOps to cut downtime and costs, but the real returns remain hard to measure.
getty
A failure at Amazon Web Services’s US-EAST-1 data center knocked out large swathes of the internet last month. Platforms from Snapchat and Reddit to Fortnite and major financial apps went offline or limped through degraded service. The disruption, lasting several hours, was a sharp reminder of how fragile even the most advanced systems can be.
The incident also reignited a broader discussion among enterprises: How can organizations build infrastructure resilient enough to recover faster next time? For many, the answer points toward AI-driven operations, or AIOps — the idea of systems that can spot, explain and resolve failures before they spiral.
The AIOps Promise
Across the enterprise world, AIOps has become the latest shorthand for reliability — a way to cut through alert noise, fix problems faster and keep revenue flowing. Yet few companies can clearly explain what kind of return they’re actually getting from it.
Gaurav Toshniwal, CEO of Sherlocks.ai, explains that the value of AIOps often comes down to two things: Cutting out noise and speeding up fixes. “Lesser false alerts means lower alert fatigue, fewer missed incidents, and less time wasted chasing false alarms — giving teams more time to focus on what matters: Shipping features to customers,” he noted. Faster remediation, he explained further, shows up in the numbers too: “Higher retention rates, lower churn, lower loss on ad spend and a better customer satisfaction score — all of which are fairly calculable.”
That logic tracks with what many companies are already finding. A 2025 New Relic report put the median cost of a major outage at nearly $2 million per hour — a figure that makes faster detection and recovery into a financial imperative rather than an engineering goal.
Measuring What Really Changes
For many teams, the hardest part of proving ROI isn’t adoption, it’s attribution. Performance often improves after rolling out AI tools, but it’s often unclear what’s actually driving the improvement. Toshniwal says his company focuses on making that connection clearer. “Sherlocks targets the debugging and triaging layers, helping engineers pinpoint issues faster and even suggesting fixes,” he said.
“We isolate our impact by benchmarking mean time to detect and mean time to resolve before and after deployment and by tracking the percentage of issues automatically triaged or resolved through Sherlocks’ recommendations.”
But that kind of benchmarking is rare across the industry. Many enterprises still can’t easily separate the gains from cleaner data or new workflows from what AIOps alone contributes. As Riverbed found in a global survey, while 87% of organisations say their AIOps investments have met or exceeded expectations, only 12% have achieved full enterprise-wide deployment. The report further highlights persistent barriers including data quality, infrastructure complexity and integration challenges — the same factors slowing broader AI adoption across IT operations.
Different Roads To Payback On AIOps Tools
ROI looks completely different depending on the size and maturity of the company. Startups, which deploy rapidly and face frequent incidents, tend to see value faster. According to Toshniwal, that’s because automation allows smaller teams to stay reliable without adding heavy operational layers. “For startups, frequent deployments mean more incidents. Sherlocks helps them maintain reliability without adding heavy operational layers, resulting in time savings and fewer disruptions.”
However, larger enterprises face a different reality. Legacy systems, overlapping vendor tools and dependence on a few experienced engineers make it harder to measure reliability and far more expensive when it fails. Toshniwal noted that the loss of critical context occurs when those individuals become unavailable or depart.
“The ROI for enterprises lies in turning implicit knowledge into explicit, reusable intelligence.” That shift reframes AIOps as more than just automation, portraying it as a way to preserve hard-earned expertise and make it accessible to everyone who keeps the system running.
Investment in AIOps continues to rise, but as buyers begin to ask for real proofs, it risks being dismissed as another vendor. And without clear results, that just might happen. That’s why Toshniwal thinks a clearer framework to measure whether AIOps tools actually work is overdue.
He suggests a kind of “reliability scorecard” that tracks how quickly teams detect and fix problems, how often new updates cause failures and how much system downtime is avoided. In his view, having consistent benchmarks like these would make results easier to compare and bring more transparency to the AIOps market.
That push for accountability is rapidly gaining traction across the industry. After the AWS outage, even major financial institutions began rethinking how they track performance and risk. Christer Holloman recently noted in a Forbes article that even major financial institutions are exploring multi-cloud strategies to limit risk after the AWS outage. The message across sectors is the same. With downtime costs climbing, executives are asking for clear evidence that their tech investments actually add business value.
What’s Ahead for AIOps
AIOps has reached a turning point. The rush of investment and experimentation is giving way to a more disciplined phase where proof matters more than promise.
As budgets tighten and scrutiny grows, the conversation is shifting from whether AI can run operations to if it can make them smarter, faster and more accountable. And as Alois Reitbauer, chief technology strategist at Dynatrace, noted, “observability is shifting from reporting telemetry about application health to informing the decisions that run the business.”
The next frontier for AIOps lies in exactly that — helping enterprises move from reacting to incidents to predicting and preventing them entirely.
If the last decade was about seeing systems more clearly, the next one will be about understanding them deeply enough to act in real time. Reliability will sit at the center of business strategy as the clearest sign that data, not guesswork, runs the show.