The Evolution of IT Operations

Three horizons of IT operations maturity: Reactive, Proactive, and Agentic

Understanding where agentic AI fits in the history of IT operations is not an academic exercise. It is the foundation for making honest assessments of where your organization is today, what it takes to advance, and what the cost of staying still actually is.

Three Horizons

The evolution of IT operations can be understood through three distinct eras, each defined by a different primary question, a different organizing metric, and a different relationship between humans and automation.

Horizon 1: Reactive Operations (1990s through 2010s)

The defining question of reactive operations was: How quickly can we restore service after something breaks?

The primary measure was Mean Time to Resolution (MTTR). Every process improvement, staffing decision, and tool investment was evaluated against its ability to reduce the time between incident detection and service restoration. Detection itself was largely passive, relying on users to report problems through a service desk.

The dominant strategies were cost-focused. Offshoring provided labor arbitrage. Scaling provided volume throughput. Lean and Six Sigma introduced process discipline. Input-driven commercial models aligned managed service costs to ticket volume and headcount. These strategies reduced cost per resolution, but they did not fundamentally change the human-at-every-step operating model.

Automation in the reactive era was rule-based and narrow. Scripts handled specific, known remediation steps. Robotic process automation addressed repetitive UI interactions. Neither category could handle the variability of real-world incidents without significant human oversight.

The reactive era produced organizations that were good at executing defined processes. It also produced organizations with chronic alert fatigue, knowledge that lived in individuals rather than systems, overnight staffing gaps, and an inability to act before users were affected.

Horizon 2: Proactive Operations (2010s through early 2020s)

The defining question of proactive operations was: How do we reduce the frequency of failures, not just the time to resolve them?

The primary measure shifted to Mean Time Between Failures (MTBF). Organizations invested in monitoring, predictive analytics, and AIOps platforms to identify conditions likely to cause incidents before they occurred. Shift-left practices brought support teams into development workflows. Preferred vendor relationships replaced commodity outsourcing. Capacity and availability management became strategic functions rather than reactive ones.

The technology capabilities that characterized this era included continuous monitoring with exception-based alerting, common scripting and automation platforms that worked across multiple applications, observability frameworks that provided dashboarding and SLA-based metrics, and the first generation of AI-assisted anomaly detection.

Proactive operations reduced incident frequency in organizations that invested seriously in it. But it preserved the human-centric operating model. Automation handled specific, well-defined tasks. Humans still performed triage, root cause analysis, cross-team coordination, and resolution for anything outside the automation envelope. The MTBF focus addressed how often things broke; it did not address the efficiency of resolution when things did break.

The proactive era produced organizations with better data, better tooling, and better process discipline than the reactive era. It also produced organizations where the gap between monitoring capability and operational response capability became increasingly visible. Telemetry quality improved; the ability to act on telemetry at scale did not keep pace.

Horizon 3: Agentic Operations (mid-2020s to present)

The defining question of agentic operations is: How do we design systems where agents handle resolution, so humans can focus on the work that requires judgment?

The primary measure is the autonomy ratio: the proportion of ITSM work that is executed by agents without human intervention at each step, rather than time to resolution or frequency of failure. This measure captures the fundamental shift from the previous two eras: the operating model, not just the outcome, has changed.

Agentic AI systems differ from both rule-based automation and predictive AI in a critical way: they are goal-oriented and context-aware, pursuing an outcome rather than following a script, and continuously sensing the environment rather than waiting for a trigger before acting. Dumas et al. (2026) describe this as a shift from design-driven process automation to data-driven, autonomous process management, where agents sense process states, reason about improvement opportunities, and act to maintain performance without explicit instruction.

In practical terms, this means an incident management agent that does not wait for a ticket to be filed. It detects the anomaly, creates the incident record, correlates it with CMDB data and current change activity, selects and executes a remediation runbook, validates the outcome, and closes the record, with no human involvement for categories where resolution is well-understood. Early enterprise deployments are demonstrating 60 to 90 percent reductions in Tier 1 ticket volume and MTTR compression from hours to single-digit minutes (Kolagani, 2024; Kore.ai, 2026).

This is a different operating model, not merely a better version of reactive or proactive operations, and it renders many of their embedded assumptions obsolete.

The Cost of Staying in Horizon 2

Organizations that remain in Horizon 2 while their industry and peer organizations advance to Horizon 3 face compounding disadvantages that are worth naming directly.

Cost structure diverges. Agentic organizations decouple ticket volume from headcount. Organizations still in proactive operations maintain a roughly linear relationship between volume and cost. As ticket volume grows, the cost gap between the two models widens continuously.

Talent allocation misaligns. The highest-value IT professionals should be working on architecture, security posture, application reliability, and capability development. In organizations without agentic coverage, those professionals spend significant time on Tier 1 and Tier 2 resolution work that agents can handle. That is a strategic cost, not just an operational one.

Data systems fall behind. Agentic operations produce better data as a byproduct of operation. Every resolved ticket becomes a structured knowledge artifact. Every CMDB update is automatic. Organizations in Horizon 2 accumulate data debt that will make Horizon 3 adoption progressively more difficult and expensive the longer they wait.

Vendor ecosystems accelerate. Gartner predicts that by 2030, 80 percent of ITSM core workflows will run autonomously (Gartner, 2025). By 2028, more than 30 percent of ITSM platform costs will go to AI capabilities. Vendors are building for Horizon 3. Organizations that stay in Horizon 2 will find that the platforms they depend on are increasingly optimized for a model they are not using.

What the Transition Actually Requires

The transition from Horizon 2 to Horizon 3 is not primarily a technology decision. Organizations do not advance to agentic operations by purchasing new tools. They advance by changing three things simultaneously:

Data foundations. Agents require accurate CMDB data, consistently categorized incident history, a knowledge base with high coverage and currency, and clean SLA configurations. None of these are given in most organizations. They must be built deliberately before agents are deployed, not concurrently.

Operating model design. Roles, escalation paths, performance metrics, and team structures must be redesigned around the assumption that agents will handle a defined category of work. This means redefining what human analysts do, not just adding agent capability to what already exists.

Governance architecture. Agents operating autonomously at production scale require identity management, audit trails, action boundaries, override mechanisms, and performance review processes that most organizations have never needed to build. See Agent Governance for the framework’s full governance model.

Organizations that invest in all three simultaneously will advance to Horizon 3 in a managed, value-generating way. Organizations that invest in the technology without the operating model and governance foundations will produce capable agents operating in an environment that cannot support them, and they will pay for that mismatch.

References

Dumas, M., Milani, F., & Chapela-Campa, D. (2026). Agentic Business Process Management Systems. arXiv preprint arXiv.18833.

Gartner. (2025). Gartner Predicts Agentic AI Will Autonomously Resolve 80 Percent of Common Customer Service Issues Without Human Intervention by 2029. Gartner Newsroom.

Kore.ai. (2026). Agentic AI in ITSM: Benefits, Use Cases, and Challenges. Kore.ai Blog.

Kolagani, S. H. D. (2024). Agentic Automation and Work Flow Orchestration in Enterprise SaaS: Effects on Ticket Resolution Time and Employee Productivity in IT Service Management. International Journal of Science and Advanced Technology (IJSAT), 15(4).

Maes, S. H. (2026). Agentic Smart ITIL, And The Disruption Of The Market Of Conventional Enterprise Applications. Stephane H. Maes’ Blog on WordPress / Multi-Agent Research Notes.