AIOps: What we learned in 2025
AI Summary
Thoughtworks shares insights from delivering 20 Proofs of Concept across 16 clients in 2025, noting that while AIOps has moved past pure concept, it has not yet industrialized. Key findings show that the highest value lies in knowledge retrieval and incident analysis rather than autonomous action, agent communication protocols remain immature, and operations are shifting toward hybrid deterministic-probabilistic systems.
Key Takeaways
- The POC-to-production gap is driven by missing AI governance, unstructured operational knowledge, and the need for continuous tuning.
- High-value use cases focus on knowledge tasks like detecting duplicate incidents, retrieving operational knowledge, and assisting with root-cause analysis, reducing L1/L2 ticket volume by 35-40%.
- Agent communication protocols like MCP remain immature for mission-critical systems due to context growth and high orchestration entropy.
Why it matters
Understanding the real-world limitations and architectural bottlenecks of AIOps helps engineering teams avoid costly pitfalls, manage hybrid deterministic-probabilistic systems effectively, and focus AI investments on cognitive augmentation rather than risky autonomous remediation.
Conversation 0
Be the first to share your perspective.