Models commoditize. Context compounds.

For the last several years, much of the conversation about enterprise AI has centered on the models themselves. Which model is the smartest? Which reasons best? Which is fastest, most efficient, or most capable of handling increasingly complex work?

Those questions still matter, but they may matter less as a source of long-term competitive advantage. Powerful AI models are becoming broadly available, organizations can increasingly choose among multiple models, and capabilities that once distinguished one model from another are spreading remarkably quickly across the market.

As that happens, simply having access to a powerful model becomes less of a differentiator. A more consequential question begins to emerge: What does your AI understand about your business that someone else’s AI doesn’t?

For enterprise operations, that distinction can make an enormous difference. Two AI systems can use exactly the same model and receive exactly the same initial alert, yet reach very different conclusions because one understands the operational environment surrounding that alert and the other does not.

One may see that application latency increased. The other may understand that the increase began shortly after a deployment, recognize which services changed, understand their upstream and downstream dependencies, identify the team responsible for the affected service, connect the behavior to similar incidents in the past, and recognize the potential impact on customers.

The underlying model may be identical. What changes is the depth of understanding available to it. And as enterprise AI matures, that distinction could become one of the most important sources of differentiation.

Models commoditize. Context compounds.

Why context becomes the differentiator

Every enterprise has an operational environment that is uniquely its own. Applications, infrastructure, dependencies, deployment patterns, service ownership, historical incidents, organizational knowledge, business priorities, and years of engineering decisions combine to create an operational reality that no other organization shares.

Experienced engineers understand much of this almost instinctively. When something goes wrong, they rarely make decisions based on a single metric or alert. They combine telemetry with architectural knowledge, recent changes, past experience, conversations with teammates, and an understanding of which systems and customers matter most.

That accumulated knowledge allows an experienced engineer to fill in gaps that may never appear on a dashboard.

AI has no equivalent intuition. It reasons from the information available to it, which means the quality of its conclusions is fundamentally influenced by the quality of the operational context surrounding the problem. A highly capable model operating with fragmented or incomplete information can still reach the wrong conclusion, while a model grounded in a richer understanding of the production environment may produce guidance that is far more useful to an engineer.

This is why the race to build increasingly sophisticated models tells only part of the enterprise AI story. Models can increasingly be acquired, changed, or replaced. An organization’s operational understanding cannot. It develops over time through the accumulated knowledge and experience of the organization itself.

That is what gives context the potential to compound.

The emerging AI Trust Gap

The distinction between intelligence and understanding also helps explain something many enterprises are encountering as they introduce AI into operational workflows.

AI capabilities are advancing quickly. Intelligent systems can already help summarize incidents, correlate information, investigate anomalies, develop hypotheses, recommend next steps, and assist engineers throughout an operational workflow. Yet organizations remain understandably cautious about how much responsibility those systems should assume inside complex production environments.

That caution increasingly creates a gap between what AI is capable of doing and what organizations are comfortable allowing it to do. The question facing technology leaders is therefore shifting from simply “What can AI do?” toward something more consequential: “What should we trust AI to do?”

The distance between those questions is what we think of as the AI Trust Gap.

The gap does not necessarily exist because an AI model lacks intelligence. Often, the problem is that an organization cannot be certain the system understands enough about its environment to make a reliable recommendation.

Consider a significant production incident. An AI system might determine from the available telemetry that rolling back a recent deployment appears reasonable. An experienced engineer, however, may know that the deployment addressed a critical security issue, that another dependent service changed at roughly the same time, or that a similar pattern occurred six months earlier for an entirely different reason.

Without that broader understanding, an answer can appear intelligent while still failing to reflect operational reality.

That distinction changes how organizations should think about building trust in AI. The answer is not simply waiting for a more powerful model. It is giving intelligent systems a sufficiently complete, current, and trustworthy understanding of the environment in which they are being asked to reason.

From observability to operational understanding

Observability created an extraordinary foundation for modern software engineering by giving teams visibility into increasingly complex and distributed environments. Metrics, events, logs, traces, and other telemetry help engineers understand system behavior at a scale that would otherwise be impossible.

That foundation remains essential, but AI changes what organizations need from it.

Historically, observability has largely served people. Engineers interpret telemetry, connect signals, investigate relationships, apply institutional knowledge, and ultimately decide what the information means. As AI participates more deeply in that same investigative process, it needs access not only to telemetry but also to the context engineers have traditionally supplied themselves.

That means connecting what is happening in production with the environment around it: service relationships, topology, changes, ownership, operational history, organizational knowledge, and other relevant context. Instead of treating those as separate pieces of information that an engineer or AI system must reconstruct during every investigation, they can contribute to a continuously evolving understanding of how the production environment actually works.

At New Relic, we describe this as trusted operational intelligence.

The objective is not simply to collect more information or give AI access to more data. It is to transform operational information into a trusted understanding that people and intelligent systems can share. Telemetry helps tell us what happened. Trusted operational intelligence helps establish what it means in the context of the environment.

That shared understanding becomes increasingly important as organizations ask AI to participate more deeply in operations.

Trust grows through understanding

Closing the AI Trust Gap does not require jumping immediately from human-led operations to autonomous remediation. In complex enterprise environments, trust is more likely to develop progressively as organizations gain confidence both in the quality of the context informing AI and in the controls surrounding its participation.

AI might initially help an engineer understand what is happening by bringing together relevant information. As confidence grows, it can assist with investigation, evaluate potential explanations, and recommend appropriate next steps. Organizations can then determine where AI should participate further, which activities require human review, and which decisions should remain explicitly under human control.

This is why governance and human accountability are inseparable from the evolution toward Autonomous Operations. Organizations need to decide where AI participates, what information it can access, which recommendations require validation, and how operational policies are enforced.

The objective is not to replace engineering judgment. It is to give engineers and intelligent systems a common operational foundation so each can contribute what it does best. AI brings speed, scale, consistency, and the ability to analyze enormous amounts of information. Engineers bring experience, accountability, business judgment, and an understanding of risk that extends beyond technical signals.

When both are reasoning from the same trusted understanding of production, the relationship becomes collaborative rather than competitive.

The advantage compounds over time

There is another reason operational context may become strategically important. Unlike access to a model, it can become more valuable as an organization uses it.

Every deployment reveals something about how an environment changes. Every incident contributes to operational history. Every investigation exposes relationships between systems. Every engineering decision adds knowledge, and every outcome provides another opportunity to improve the organization’s understanding of how its production environment behaves.

Much of that knowledge is fragmented today across telemetry, tools, tickets, documentation, conversations, and the memories of experienced engineers. When those pieces remain disconnected, organizations repeatedly reconstruct the same understanding each time something goes wrong.

But when that information contributes to a shared operational understanding, each experience can help make the next investigation more informed. New engineers can begin with context that previously existed primarily as tribal knowledge. Experienced engineers can spend less time rebuilding the same operational history. And AI can begin an investigation with a richer understanding of the environment rather than assembling disconnected fragments of information every time.

This is where the idea that context compounds becomes more than a memorable phrase. It describes a potentially durable enterprise advantage.

A new model can be adopted quickly. Years of accumulated understanding about how an organization’s systems behave cannot.

A different path to Autonomous Operations

This also provides a useful way to think about Autonomous Operations.

Autonomous Operations should not be understood simply as operations without people or as the wholesale automation of operational decisions. It represents an evolution toward an operating model in which people and intelligent systems can investigate, reason, recommend, and participate in operational work from the same trusted understanding of production.

For New Relic, Ground Truth is being built to help establish that trusted operational understanding, while Autopilot enables AI to assist engineers using operational context while keeping human judgment and governance central to the process.

Together, they illustrate a larger change in what observability can become in an AI-driven world. Rather than simply being a system for collecting and presenting telemetry, observability can increasingly provide an operational intelligence foundation that helps both people and AI understand what is happening, determine what matters, and make better-informed decisions about what should happen next.

That is the foundation for the broader vision we describe as Autonomous Operations.

The next competitive advantage in enterprise AI

AI models will continue to improve at an extraordinary pace. They will become faster, more capable, more efficient, and increasingly accessible to almost every organization.

That progress will create enormous opportunities, but it will also make access to powerful AI less differentiating than it is today. The organizations that ultimately create the most value from AI may therefore not be the ones that simply choose the best model. They may be the ones that give intelligent systems the deepest and most trustworthy understanding of the environments in which they operate.

For software operations, much of that understanding already exists. It lives in telemetry, service relationships, deployments, incidents, ownership, business context, documentation, and years of engineering decisions. The opportunity is to connect those fragments and turn them into trusted operational intelligence that both people and AI can use.

That changes the role of observability. Instead of simply helping engineers understand what happened, observability can increasingly help establish the shared operational understanding that allows people and intelligent systems to investigate, reason, recommend, and participate in operational work together.

And that may ultimately be the more durable enterprise AI advantage. Models can be adopted, replaced, and upgraded. An organization’s accumulated understanding of how its systems behave, how its teams operate, and what matters to its business is much harder to replicate.

See how New Relic brings observability, trusted operational intelligence, and AI together to help teams understand their environments and guide their response. https://newrelic.com/platform/autonomous-operations

Models commoditize. Context compounds.