AI agents are entering their rebuild era as enterprises confront the reliability problem

2026-05-29 · Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, short

Summary

Enterprise AI agents are facing significant reliability challenges as they move into production, according to Preeti Somal, Senior VP Engineering at Temporal Technologies. Organizations are discovering that large language model (LLM) performance alone does not guarantee success for long-running AI workflows, which must manage state, recover from crashes, coordinate across APIs, and control inference costs. Many teams are now rebuilding first-generation agent architectures, focusing on workflow orchestration, observability, governance, and robust recovery mechanisms. Somal highlights the need for a "deterministic spine" where orchestration software ensures consistent execution around probabilistic LLMs, enabling recovery from failure points and providing visibility into token consumption. This shift reflects a broader realization that production AI systems require durable execution and sophisticated systems engineering, moving beyond initial rapid deployments.

Key takeaway

For MLOps Engineers deploying or scaling AI agents, prioritize robust systems engineering over solely optimizing LLM performance. You should redesign architectures to incorporate durable workflow orchestration, state management, and recovery mechanisms to prevent costly restarts and ensure reliability. Implement observability for token spend and establish internal governance frameworks. This approach will significantly improve agent stability, reduce operational costs, and provide the necessary controls for enterprise-grade AI.

Key insights

Enterprise AI agent reliability in production hinges on robust workflow orchestration, state management, and recovery, not just LLM performance.

Principles

• LLM performance alone does not ensure agent production success.
• Long-running AI agents require durable execution and state management.
• Orchestration provides a "deterministic spine" for probabilistic LLMs.

Method

Redesign agent architectures for workflow orchestration, observability, governance, and recovery. Implement a "deterministic spine" to ensure reliable execution and state management around probabilistic LLMs, enabling recovery from failure points.

In practice

• Monitor token spend through observable, step-by-step workflows.
• Resume failed agent workflows from interruption points to save costs.
• Establish internal "paved paths" for agent governance and controls.

Topics

• AI Agents
• Workflow Orchestration
• Enterprise AI
• Reliability Engineering
• State Management
• LLM Operations
• Cost Management

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Engineer, MLOps Engineer, AI Architect

Related on AIssential

• Karpathy’s March of Nines shows why 90% AI reliability isn’t even close to enough — Reliability in AI agentic workflows compounds failure, demanding disciplined engineering to achieve enterprise-grade dependability.
• Decomposing the Agent Orchestration System: Lessons Learned — Agents can reliably recover from infrastructure failures if given the right context and capabilities.
• LAI #127: The Infrastructure Layer of AI Is Becoming the Product — Robust AI systems require careful architectural design, especially for agentic applications, to prevent silent failures and ensure reliability.
• From Chaos to Choreography: Multi-Agent Orchestration Patterns That Actually Work — Sandipan Bhaumik — Scaling multi-agent AI systems requires distributed systems thinking and robust architectural patterns to avoid production failures.
• LAI #113: The Engineering Work That Decides Whether AI Holds Up — Operational discipline is key to successfully deploying and scaling AI systems in production.
• Build Long-running AI agents that pause, resume, and never lose context with ADK — Long-running AI agents require durable state machines, event-driven resumption, and multi-agent delegation to manage complex, multi-day workflows.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.