Production AI failure modes — companion videos

The article treats production AI quality as failure-mode management: context, evals, tool correctness, prompt injection, stale data, weak fallbacks and observability. This companion pairs it with a real production case study: a team that ran agents across many companies and hit those failure modes for real.

Primary pick

1:08:40

The Agent Landscape - Lessons Learned Putting Agents Into Production

MLOps.community

Prosus's VP of AI and an AI engineer report what actually broke when they deployed agents across the group's portfolio companies: prompt-injection pen-testing before launch, an unsafe write when a Jira agent choked on human shorthand, stale context handled by making agents surface their assumptions, and fallback design that merged or killed agents once they added cognitive load. It reads like the article's failure-mode register replayed as a live postmortem.

What you should get from this: Recognize the failure modes your own agents will hit first and pre-plan the tests, fallbacks and kill criteria for each.

Watch or know first: Familiarity with LLM agents, tool calls and basic production monitoring; the article's failure-mode register as a map.

AI Expert note: Recorded in early 2025, so the tool landscape has moved on, but every failure class discussed is still being rediscovered by new teams. Watch for the incidents and the operating decisions, not the framework names.

Open video page