Five Practitioners Discuss Balancing Speed and Safety in Production AI Systems

When AI agents can retrieve internal documents, generate code, or execute actions autonomously, the stakes shift dramatically. Engineering teams face an immediate tension: platform and engineering teams want agents to move fast and do more, while privacy and security experts push back with tighter controls. A production workflow that succeeded in testing can behave unpredictably when it encounters real users, operational constraints, and actual incidents.

On October 14, InfoQ is hosting a 60-minute live panel bringing together five senior practitioners who build, secure, and operate AI systems at scale. Attendees can register for free to participate live or watch the recording later.

Key Topics Under Discussion#

The panel will examine how organizations navigate several critical production challenges:

• Agent autonomy — determining which actions agents execute independently versus which require human sign-off
• Verification and change management — approaches to validating AI-generated code before it enters CI/CD pipelines
• Data exposure and threat modeling — controlling access to sensitive or regulated information and identifying risks upfront
• Retrieval-augmented generation at scale — architectural and operational patterns that survive beyond controlled demonstrations
• Internal advocacy — strategies for building the case for stronger governance before failures force the conversation

Rather than offering a single playbook, the discussion will surface the reasoning and tradeoffs behind different approaches, helping attendees evaluate options against their own systems and constraints.

The Panel#

Hien Luu, who chairs QCon AI New York 2026, brings experience leading machine-learning platforms at DoorDash and Zoox, plus authorship of MLOps with Ray. He will address reliability, RAG systems, and production AI architecture.

Katharine Jarmul, a privacy and security specialist and author of Practical Data Privacy, will examine how teams control data access and assess risk in AI systems.

Zichuan Xiong and Premanand Chandrasekaran from Thoughtworks will explore the engineering practices around coding agents, including verification workflows and governance decisions across the delivery pipeline.