Production software keeps breaking and it will only get worse — Anish Agarwal, Traversal.ai
Source: YouTube · AI Engineer · published Jul 10, 2025 · 18:10
AI-powered autonomous troubleshooting is essential to address the growing complexity of software systems, as current methods like AI ops and LLMs fail to deliver actionable insights in production incidents. The core goal is to enable engineers to focus on creative, high-impact work like system design, while AI handles repetitive, error-prone troubleshooting tasks.
Key Takeaways:
• AI tools like Cursor or GitHub Copilot are narrowing development work, but leave system design and troubleshooting as critical, unautomated challenges 1:02.
• Troubleshooting in production is currently inefficient, involving "dashboard dumpster diving" and manual root cause analysis, leading to long on-call times and team burnout [2:54–3:05].
• Traditional AI ops and LLMs fail due to false positives, lack of numerical understanding, and overwhelming data volume [4:23–5:11].
• A novel solution combines causal machine learning (for identifying root causes), semantic reasoning (for understanding log and metric context), and a "swarm of agents" for efficient, parallel data exploration [7:24–8:52].
• Traversal’s AI system reduced Digital Ocean’s mean time to resolution by 40% by autonomously identifying root causes from petabytes of observability data [13:41–14:48].
This approach transforms incident response from a chaotic, team-driven effort into a fast, confident, and automated process—offering a scalable solution for enterprise observability and beyond.
Sources:
- 1:02 Discussed the shift in software engineering workflow due to AI tools.
- 2:54 Describes the inefficiencies in current troubleshooting workflows.
- 4:23 Highlights limitations of existing AI ops and LLM-based troubleshooting.
- [7:24](https://www.youtube.com/watch?v=L6_NiGI
Generate CPE Credits
Generate a professional CPE document from this video's transcript.
Estimated credit: 0.5 CPE hours
Estimate uses the video runtime (1 hour ≈ 1 CPE, rounded to the nearest 0.5, minimum 0.5, maximum 2.0). The final amount can be lower after review, never higher.
Topic: Cybersecurity. Commonly maps to: Security and Risk Management, Security Operations. Exact CISSP domains are assigned during generation.
Free account. One generation at a time, with a daily limit.
CPEBuddy is independent and not affiliated with or endorsed by ISC2, ISACA, or any certification body. Exports are formatted for common CPE submissions; acceptance is at your certification body's discretion.
Transcript Preview
First 800 characters of the transcript
[Music] hi everyone. Thank you for uh coming to our talk. Uh so he was kind enough to already introduce us. Um so I'm the CEO. Matthew was the first person who joined us. Uh if any difficult questions, please direct it towards Matt. Um now if you think about the three major categories of of software engineering at least as we see it there's three things that show up to me the system design where you think about you know how do you actually architect a system a lot of the the talks we saw in this uh track have been about that in some sense at a high level second is actually developing software right putting in the business logic of your particular company and and all of the devops that comes with it and then when you when your software actually hits production invariably it's going to break…