Automation Delivered on Its Promise. So Why Does IT Still Feel Underwater?
Five years ago, automation promised enterprise IT something it had been chasing for decades: time.
Self-service portals reduced password resets. Software catalogs eliminated manual provisioning. Workflow engines standardized onboarding and offboarding. Response times improved, ticket volumes dropped, and the service desk got faster.
On paper, automation worked.
Yet ask most CIOs whether their teams have more capacity today than they did before those investments, and many will answer the same way:
Not really.
The reason is simple. The work never disappeared. It moved. The repetitive tasks that once dominated service desks have largely been absorbed by automation, leaving behind a long tail of unpredictable work that refuses to sit still inside a workflow. Instead of resetting passwords, experienced engineers now spend their days untangling access exceptions, supporting aging business applications, coordinating unusual onboarding requests, and troubleshooting endpoint issues that require judgment rather than scripts. Individually, these activities appear manageable. Collectively, they consume thousands of hours every year and prevent IT organizations from focusing on transformation.
“The work never disappeared. It moved.”
This is the hidden labor tax of modern IT. Unlike infrastructure costs or software licensing, it doesn’t appear as a budget line item. It shows up as delayed modernization initiatives, burned-out specialists, slower project delivery, and scarce technical talent consumed by operational exceptions rather than strategic priorities.
“This is the hidden labor tax of modern IT. Unlike infrastructure costs or software licensing, it doesn’t appear as a budget line item.”
IDC predicts that by 2026, more than 90% of organizations worldwide will be affected by the IT skills crisis, resulting in an estimated $5.5 trillion in delays, quality issues, and lost revenue. The problem isn’t only that companies need more IT professionals. It’s that the people they already have are tied up resolving operational exceptions. Hiring alone won’t fix that.
Where the Hidden Labor Tax Lives
For years, automation strategies targeted the highest-volume work first, and that was the right call. Password resets, software deployment, standard provisioning, and routine service requests were predictable and highly repeatable. They represented the obvious starting point for automation.
“But every successful automation program hits an inflection point. The remaining work becomes increasingly unique, contextual, and difficult to standardize.“
That inflection point arrives at a different time in every organization, but the pattern is remarkably consistent. The highest volume work is gone. What remains is a collection of lower-frequency, higher-complexity work that demands experience and judgment. Together they become one of the largest consumers of skilled capacity you have.
Access Entitlement Exceptions
Consider access governance. Modern identity platforms provision users quickly and consistently, but compliance-driven access reviews still demand careful human oversight. Role changes, temporary project assignments, acquisitions, and regulatory requirements frequently create situations that fall outside standard policies. In highly regulated industries, one overlooked entitlement review can introduce audit findings or security exposure. Automation handles the routine. Skilled professionals resolve the exceptions.
Legacy Applications and Institutional Knowledge
Legacy applications create a similar challenge. Many organizations continue to rely on business-critical systems built long before APIs, telemetry, and cloud-native architectures. When those systems fail, automated diagnostics often reach a dead end. Resolution depends on a handful of experienced specialists whose institutional knowledge has accumulated over the years rather than through documentation. As those experts retire or move on, the operational risk grows.
Employee Lifecycle Edge Cases
Employee lifecycle management illustrates another common blind spot. Standard onboarding workflows function well for full-time employees who follow established processes. Real organizations, however, depend on contractors, consultants, international hires, seasonal workers, acquired teams, and people who change roles twice a year. These scenarios introduce countless variations that frequently fall outside automated workflows. Someone eventually intervenes manually, often after delays and sometimes after an access risk has opened up.
Software Provisioning and Endpoint Decisions
The same pattern appears across software provisioning and endpoint management. An application that sits outside the approved catalog requires manual licensing, approvals, and configuration across multiple teams. A remote employee’s laptop may present symptoms that automated tools cannot confidently classify as hardware failure, software corruption, or user error. Every exception requires investigation. Every investigation consumes valuable expertise.
Why Traditional Automation Reaches Its Limits
Traditional automation was built to execute, not to learn.
That distinction matters. Workflow automation excels at following predefined rules, standardizing repeatable processes, and delivering consistent outcomes. But exceptions exist precisely because they break those rules. They require context, judgment, and the ability to adapt when conditions change.
So, each new automation project gets harder than the last, because the work still on the table isn’t repetitive anymore.
“What looks like the ceiling of automation is usually the ceiling of tools that were never designed to learn from the work they couldn’t handle.“
Agentic AI Changes the Economics of IT Operations
Agentic systems don’t just run predefined workflows. They learn from patterns across incidents, endpoint telemetry, identity requests, approvals, and past resolutions. Every interaction adds to what the system knows about how your environment actually behaves. Recurring exceptions start to look like recognizable patterns. Yesterday’s escalation becomes tomorrow’s recommendation, then eventually an automated outcome nobody has to think about.
The goal shifts. You’re no longer trying to automate one more task. You’re trying to shrink the exception pool itself.
That suggests a better question for evaluating AI investments. Not how many tickets got automated. Instead: are fewer requests requiring human intervention this quarter than last?
What this means for choosing a partner
Most managed service providers can automate routine requests. Far fewer have built a delivery model that learns from operational complexity over time. That difference decides whether your efficiency gains flatten out after the first wave or keep compounding.
AI shouldn’t just strip cost out of work that was already predictable. It should capture what your organization knows, spot emerging patterns, and take on progressively harder work so experienced engineers can spend their time on modernization, resilience, security, and the things that actually move the business.
“Every exception your team resolves should make the delivery model smarter instead of vanishing into one engineer’s memory.“
From Exception Handlers to Enterprise Architects
“The organizations that will lead over the next decade are unlikely to be those that automated the most routine tasks. They’ll be the ones that got progressively better at eliminating complexity itself.“
CIOs should rethink what they measure, too. Ticket deflection, mean time to resolution, and cost per incident still matter, but they only show part of the picture. A stronger scorecard tracks how fast new exceptions get folded into future automation, how much institutional knowledge gets captured instead of staying siloed, and whether scarce technical talent is drifting toward transformation work or away from it. Those numbers tell you whether you’re building adaptive capability or just static efficiency.
The first generation of automation proved that repetitive work could be standardized. The next one will decide whether organizations can change how knowledge itself is captured, applied, and scaled. That is the real promise here. Not faster service desks. Skilled people getting their time back.
Is Your Automation Strategy Eliminating Exceptions or Just Automating the Routine?
If your service desk metrics look good but your IT team still feels stretched, the problem probably isn’t your people. It’s the volume of exceptions traditional automation was never built to handle.
Pomeroy applies AI to the complex, high-value work that consumes skilled IT resources. The result is an operating model that keeps learning, reduces manual effort, and gives IT teams room to focus on innovation instead of firefighting.
See how Pomeroy’s AI-powered SmartServices Suite helps organizations reduce operational complexity and free their teams for the work that matters.
