Ask HN: How are teams productionizing AI agents today?

1 pointsyanitcorel14d ago2 comments

Many AI agent demos work well in prototypes but break when moved to production.

We’ve been exploring how teams are running agentic AI systems reliably on AWS, including orchestration, scaling, monitoring, and guardrails.

We’re hosting a live technical masterclass where we’ll walk through architectures and real-world patterns used in production environments.

If this topic is useful for engineers here, they can join the session.

Event link - https://www.netcomlearning.com/webinar/productionizing-agentic-ai-on-aws-from-poc-to-enterprise-scale?utm_source=newplatform_event&utm_medium=event&utm_content=master_classes&utm_term=hn

Are you curious how others here are running AI agents in production today?

2 comments

chirdeeps12d ago

The biggest thing that surprised us: the constraint shifts from intelligence to reliability the moment agents start modifying shared systems. In a PoC, a wrong action is a failed experiment. In production, it's a corrupted customer record, a duplicated invoice, or a deployment that can't be unwound.The specific properties you need before going live that most frameworks don't give you out of the box: 1. Idempotency — can every agent action be safely retried without duplicating side effects?

2. Rollback semantics — if a multi-step workflow fails, what unwinds?

3. Authority boundaries — what can each agent do without human approval, and what requires sign-off?

4. An authoritative action history — when something goes wrong, can you reconstruct exactly what happened and why, without stitching together logs from five different systems?

Most teams discover these requirements after the first production incident. The teams that define them in an execution layer before going live have a much smoother transition. Keen to check out what you will be sharing in the session.

raffaeleg13d ago

The biggest gap between demo and production isn't the model or the framework. It's three boring things: deterministic fallbacks (what happens when the agent fails or hallucinates), observability (can you trace exactly why an agent took action X), and cost controls (token budgets per task, not per call). Most teams get burned by the same pattern: the demo works beautifully, then in production you realize you need to handle the 15% of cases where the agent confidently does the wrong thing. The teams shipping successfully treat agents like junior employees, not autonomous systems. Guardrails first, autonomy second.

1 more reply

j / k navigate · click thread line to collapse