The 6 Pillars of an Agentic Harness for Production — Varun Krovvidi, Resolve AI

AI changed how we write code. But 70% of engineering time goes to running and fixing it. Varun Krovvidi, product marketing lead at Resolve AI, explains why AI that writes code well still struggles to run production. Code is self-documenting, modular and a single domain. Production spans code, infrastructure, telemetry and many teams, and needs one correct answer. He covers where agents crack as you scale: anchoring bias and the model treadmill, too much or too little context, coherent answers that aren't causal, missing guardrails, and no learning across investigations. Then he lays out six pillars of a harness that fixes them: model orchestration, context engineering, causal reasoning, governed actions, learning systems and evals. He finishes with a demo of Resolve AI turning a Grafana alert into a root cause, ruling out a coincidental GCP outage, and holding its ground when pushed. In this talk: • Why coding is easy for AI and production isn't • Five ways production agents break as they scale • Six pillars: orchestration, context, causal reasoning, governed actions, learning, evals • Demo: from Slack alert to root cause with ruled-out theories SPEAKER Varun Krovvidi, Product Marketing Lead, Resolve AI LinkedIn: https://www.linkedin.com/in/varunkrovvidi/ LINKS Resolve AI: https://resolve.ai Resolve AI on X: https://x.com/resolveai CHAPTERS 0:00 Intro 0:48 Who's on call? 1:33 Why AI got good at code 2:33 Running software is the other 70% 2:53 On-call, incidents and daily vitamins 4:13 The end of unlimited AI 5:18 The last mile is the longest 5:28 A multiplayer problem 6:23 Resolve's three kinds of agents 7:02 Why not just point a model at prod? 7:37 Crack 1: anchoring and the model treadmill 8:47 Crack 2: too much or too little context 9:47 Crack 3: coherent isn't causal 10:27 Crack 4: guardrails 11:02 Crack 5: learning loops 11:52 The six pillars: model orchestration 12:42 Context engineering 13:32 A causal chain of evidence 14:12 Governed actions 14:32 Learning systems 15:07 Evals at five levels 15:57 Demo: alert to root cause 18:12 Demo: ruled-out theories 19:02 Demo: pushing back on the agent 20:11 Wrap-up Recorded at the AI Engineer World's Fair 2026 in San Francisco. Subscribe for more talks from the engineers building with AI. AI Engineer: https://ai.engineer YouTube: https://www.youtube.com/@aiDotEngineer X: https://x.com/aiDotEngineer LinkedIn: https://www.linkedin.com/company/aidotengineer/ #AISRE #AIAgents #AIEngineer