5 Challenges DevOps Must Solve to Prepare for AIOps

AIOps is the wave of the future, but getting there will involve clearing some remaining hurdles. Our expert offers some guidelines.

Written by Derek Ashmore
Published on Jul. 16, 2025
A man in a suit holds symbols saying AI, DevOps, and a symbol
Image: Shutterstock / Built In
Brand Studio Logo
Summary: DevOps teams are adopting AIOps to automate incident response and boost observability, but challenges like data silos, poor data quality, cultural resistance, risky automation and lack of explainability must be addressed to enable safe, scalable and intelligent IT operations.

As DevOps practices evolve, teams increasingly turn to AIOps (AI applied to IT operations) to drive faster incident detection, smarter analytics and self-healing automation. Though AIOps can dramatically augment DevOps capabilities, integrating intelligent systems requires overcoming complex technical, procedural and cultural hurdles. This article explores five pressing challenges and offers practical strategies to address them.

5 Challenges DevOps Must Overcome to Prepare for AIOps

  • Data fragmentation and integration barriers.
  • Maintaining data quality and model reliability.
  • Cultural resistance and changes to roles.
  • Balancing automation boundaries.
  • Building trust through explainability and governance.

More From Derek Ashmore3 Times You Should Skip Infrastructure-as-Code in DevOps

 

1. Data Fragmentation and Integration Barriers

Challenge

AIOps thrives on rich, unified telemetry (logs, metrics, events and traces) that together form the digital pulse of modern infrastructure. In many organizations, however, these data streams remain fragmented. Monitoring tools, CI/CD pipelines, cloud platforms and ticketing systems often operate in isolation, producing siloed outputs that are difficult to correlate. This disjointed data environment diminishes the effectiveness of AI, leading to false alarms, missed anomalies and an incomplete view of system health.

Solution

The solution begins with centralizing observability. By implementing an observability platform or constructing a centralized ingestion pipeline, teams can bring disparate telemetry sources together into a unified data fabric. This consolidation should be supported by schema enforcement to ensure data follows consistent formats and structures. 

Preprocessing steps such as normalization, deduplication and timestamp correction enhance the reliability of AI insights. Moreover, introducing metadata catalogs and scalable storage systems allows for efficient indexing and retrieval, enabling the AI layer to draw meaningful correlations. These efforts collectively give AIOps models the holistic view required for accurate detection, root cause analysis and system-wide awareness.

 

2. Maintaining Data Quality and Model Reliability

Challenge

Even after consolidating data sources, the quality of telemetry remains a challenge. Incomplete log entries, inconsistent formats, missing values or noisy data can undermine the predictive capabilities of AIOps. These inconsistencies not only generate false positives but also erode trust in the system, making engineers less likely to rely on AI-generated outputs.

Solution

Maintaining data integrity requires a proactive approach. Real-time validation mechanisms should be embedded within telemetry pipelines to detect schema violations or anomalies in the data stream as they occur. Flagging issues early ensures that poor-quality data doesn’t contaminate downstream processes. Anomaly detection tools can be integrated upstream to monitor the data source health itself, providing early warnings when telemetry sources begin to behave abnormally. 

In addition, AI models must be regularly retrained on recent data to reflect system changes and reduce drift. This retraining, paired with systematic model evaluation, keeps outputs relevant and trustworthy. Layered tracing and logging strategies, such as enriching logs with contextual metadata, also bolster data fidelity, ensuring AI has a consistent, high-quality foundation to work with.

 

3. Cultural Resistance and Changes to Roles

Challenge

AIOps can introduce significant shifts in how DevOps teams operate. As automation assumes tasks traditionally performed by engineers, concerns about job security, loss of control and role ambiguity often arise. Engineers may view AI as a threat rather than a tool, leading to resistance that stalls adoption and undermines collaboration.

Solution

Navigating this cultural transition requires positioning AIOps not as a replacement, but as a force multiplier. Leaders should clearly communicate that AIOps is designed to reduce toil, freeing engineers from repetitive diagnostics and reactive firefighting, and empowering them to focus on innovation, architecture, and performance optimization. Upskilling initiatives are essential, offering training programs, AI literacy workshops and hands-on experimentation that show how AI enhances daily workflows. 

Highlighting internal success stories, where AIOps caught a hidden issue or accelerated incident resolution, helps reinforce its value. Encouraging cross-functional collaboration among developers, SREs and data specialists ensures that AIOps evolves as a shared initiative, rather than an imposed technology. With the right framing and involvement, teams begin to see AIOps as an ally, not a competitor.

 

4. Balancing Automation Boundaries

Challenge

AIOps enables powerful automation capabilities, from triaging incidents to initiating remediation. But full autonomy brings risks. Without proper guardrails, AI might misdiagnose root causes or apply incorrect fixes, potentially compounding system instability or masking deeper issues. Determining which tasks should be automated, and to what extent, remains a nuanced challenge.

Solution

The key to safe automation is progressive adoption. Rather than aiming for full autonomy immediately, organizations should take a phased approach. Initial efforts can focus on low-risk automation such as alert deduplication or incident enrichment, gradually expanding into AI-generated diagnostic suggestions and eventually to guided remediation workflows.

Importantly, all automation should include human-in-the-loop mechanisms. For high-severity or unfamiliar issues, AI can propose actions while engineers retain the final decision-making authority. Audit trails and comprehensive logging of AI-driven actions add traceability, ensuring that any unintended consequences can be traced and corrected. Regular post-incident reviews should assess how AI systems performed, allowing teams to refine thresholds, update rules and recalibrate confidence scores. This measured strategy ensures that automation scales safely, with human judgment preserved where it matters most.

 

5. Building Trust Through Explainability and Governance

Challenge

Trust is the foundation of effective AI adoption. DevOps teams must understand not just what the AI is doing, but why. Without transparency, AIOps can seem like a black box, leading teams to ignore, override or disable its outputs. Unexplained recommendations or opaque anomaly scores create hesitation, especially in high-stakes environments.

Solution

Building trust begins with explainability. AIOps tools should offer both alerts and the reasoning behind them, highlighting contributing factors, anomaly scores, correlating signals and supporting telemetry. Engineers should be able to trace recommendations back to logs, traces and metrics, enabling rapid validation.

Alongside this technical transparency, governance structures must be implemented. Approval gates, escalation policies and automated remediation limits introduce oversight, ensuring that AI remains accountable and controllable. Teams should also adopt a model performance review process, tracking metrics such as false positives, false negatives, and remediation accuracy over time. These practices build a culture of confidence, allowing AIOps to become a reliable partner rather than a mysterious agent.

More on the Future of DevOpsThe Rise of Crypto-Native DevOps: New Tools for Building a Blockchain-First Future

 

Accelerating Impact With AIOps

To put this framework into action, DevOps teams can follow a staged AIOps maturity model. In the foundation phase, organizations focus on telemetry centralization, establishing common data schemas and building preprocessing pipelines to clean and normalize inputs. The alert streamlining stage uses AI for anomaly detection and alert deduplication, reducing noise and focusing engineering attention. 

Next, in the root cause analysis phase, AI correlates signals across systems to deliver diagnostic suggestions, helping teams pinpoint underlying issues faster. The guided remediation phase introduces contextual recommendations, often paired with human approvals. In the pilot autonomy stage, organizations test controlled automation in production with strict governance, rollback mechanisms and audit logs. Ultimately, the goal is a self-driving ecosystem, where AIOps supports predictive capacity planning, adaptive response to system behavior and continuous model improvement.

Aligned with integrated tooling, iterative feedback loops, and transparent governance, this roadmap enables a smooth and scalable transition from reactive DevOps to intelligent, AI-enhanced operations. AIOps offers a compelling vision of automated monitoring, intelligent diagnostics and self-healing systems that empower DevOps teams to innovate instead of firefighting.

By investing in unified observability, ensuring data integrity, supporting cultural transformation, adopting automation with care and embedding explainable AI and governance, teams can realize the full promise of AIOps. The result is a DevOps environment that is more resilient, responsive and forward-looking. The future of IT operations isn’t just automated, it’s intelligent, collaborative and continuously improving. Those who navigate today’s challenges with rigor will lead that future.

Explore Job Matches.