Prompt Practices TLV

Autonomous AI agents are moving fast ! from cool demos to real systems powering workflows, customer support, research, engineering, and internal ops. But the big question remains: how do we actually run them in production?

Batel Zohar & Mey Beisaron

## Shipping AI Agents: What Really Happens in Production

Shipping AI agents is exciting, but once they hit production, you quickly discover what the demos didn’t prepare you for.
In this talk, we’ll share the real best practices that emerged only after breaking things ourselves what failed, what surprised us, and what we wish we knew earlier. You’ll hear about the common pitfalls teams run into, the patterns that actually improve stability, how to manage unpredictable behavior, and the guardrails and evaluations that make agents safer and easier to operate.
The goal is simple: help you avoid the mistakes we’ve already made and give you a realistic, practical approach to running AI agents in production without the chaos.

Omri Nardi Niri

The AI Parliament: Our Multi-Agent RAG Setup for Smarter Decisions at Work
At Semperis, we've got a tool simulating a personal squad of AI consultants for our team. It helps our team tackle industry-specific challenges. This approach is inspired by research from both academia and the industry.
We built the AI Parliament around a multi-agent system, where each AI "consultant" has its own specialized role - like a Senior Engineer, a Data Whiz, each with its own personality to keep it realistic. These personas get created on the fly for the specific purpose of fulfilling the user's immediate goal. When they operate, they're pulling from our company's knowledge base (RAG), so their advice feels tailored and draws on our actual history.
From Theory to Practice: How Teams Build and Operate Agents Today