How AI Agents Use Shared Context to Resolve IT Issues With Less Manual Work

Current IT stacks reflect how organizations scaled one layer at a time:

• Observability platforms detect and correlate issues across environments
• ITSM tools manage incidents, workflows and governance
• Automation tools execute predefined actions and remediation steps

Each system performs an important function, but incidents rarely stay within a single domain. A performance issue might trigger an alert, generate a ticket, require investigation across logs and change records, and ultimately result in a remediation workflow. Resolution often depends on someone manually interpreting available information and deciding what should happen next.

Scripts, runbooks and automation improved this, and they remain valuable. But they are also designed around known scenarios. When something unexpected happens, people step in to assess the situation, evaluate risk and determine whether established procedures still apply. 

That expertise is valuable but creates a scaling challenge. As environments grow more complex, human intervention cannot remain the primary mechanism for adapting operational workflows. The lag between spotting an issue and resolving it is what customers feel: poor service, failed transactions and the same incidents resurfacing while teams decide what to do.

Even in environments with strong predictive signals, organizations still face delays when determining the right action to take and ensuring it is applied consistently.

The challenge has shifted from detection to response, and the goal now is extending how long operations can run effectively before requiring human involvement and enabling systems to respond effectively when situations don’t fit neatly into predefined workflows.

People don’t solve problems by following scripts alone. They use context.

They understand which systems and services are connected. They know when a recent deployment, configuration change or patch may be contributing to an issue. They recognize recurring problems and the actions that resolved them previously. And they can see which customers, business processes or transactions may be impacted.

When operational data, service relationships, change records and historical knowledge are connected, teams can see not only that something is wrong, but also what caused it, what it affects and what action should happen next.

That context becomes increasingly important as organizations look to introduce more autonomous, agentic ways of working. 

Without context, automation remains limited to predefined actions. With context, operational workflows can adapt more effectively to changing conditions and make better use of the information already available across the environment. 

Consider a multinational jewelry brand whose digital platform had to stay fast and stable through holiday and sale-driven traffic spikes. Dependencies between systems made each issue hard to isolate: a delay in one service could slow checkout, and a failure upstream could ripple across customer journeys with no clear view of where the breakdown began. Connecting signals across those systems, so teams could see how a problem in one affected the others, cut major incidents by 82 percent and aging tickets by 80 percent while holding uptime at 99.99 percent.

The same pattern appeared at a global beauty brand running more than 50 sites across 28 interconnected platforms. Each team could see its own part of the environment but not how an issue moved through the full customer journey, so a single checkout failure might pass between teams before reaching the right owner, and recurring problems were fixed again and again instead of at the source. Bringing performance, incident and business-impact data into one shared view improved mean time to resolution by 50 percent, cut repeat issues by roughly a third and lowered the cost of running operations by 35 percent.

Automation follows predefined paths, but agentic operations introduce a different model. Rather than simply executing a predefined workflow, agents can reference available context, determine which actions are appropriate, use existing tools and workflows to complete tasks and escalate to people when oversight or approval is required.

Automation vs. agentic AI: what’s the difference?

Traditional automation executes a path defined in advance. Agentic operations use current context to choose among approved actions and involve people when judgment or authorization is needed.

This creates a more flexible operating model. Agents can help organizations respond to a wider range of situations without requiring every possible scenario to be explicitly programmed in advance.

Equally important, it allows organizations to keep humans in the loop where they matter most—governance, compliance, risk management and complex decision-making—while reducing the manual effort required for routine operational work. 

This is the model Nissan moved toward. Across its websites and dealer platforms, failures in the forms that capture sales leads once surfaced as a flood of alerts that engineers triaged and traced by hand, work that could stretch resolution to roughly half a day. With shared operational context connecting monitoring data, service relationships and change records, the platform now correlates those signals on its own, generates a root-cause analysis and assembles what an engineer needs to act. 

People still make the call on the fix; the agent removes the manual stitching that used to come before it. That shift helped Nissan move 80 percent of its operations from reactive to proactive and cut the cost of running them by 40 percent, without replacing the tools already in place.

The goal is an operating model where agents, automation and people each contribute where they are most effective. Most enterprises already have the tools required to run complex environments. The next step is helping those investments work together through shared enterprise context.

The organizations making the most progress are building operating models where agents  can handle more routine decisions and actions, while humans remain responsible for oversight, judgment and governance. Success depends on maintaining operational context, applying knowledge consistently and adapting without constant manual intervention.

Over time, that creates an operating model that becomes more effective with experience—one that can respond to change with greater speed, consistency and resilience. 

Learn more about how we are helping customers do this with our AI platform for autonomous IT operations, Sapient Sustain.