Uploaded June 2026 | Updated September 2026, 2 weeks ago
What does AI agent adoption actually look like in production?
Drawing on insights from thousands of teams building with Mastra, this session explores the patterns emerging across the AI landscape as organizations move from prototypes to production systems. Learn which types of agents are succeeding, where companies are seeing the most value, and what separates successful deployments from stalled experiments.
The talk breaks down the three major categories of production agents—customer-facing agents, internal enterprise agents, and developer platform agents—and shares real-world examples from companies including Indeed, Factorial, MongoDB, and others. Along the way, you’ll learn practical lessons about evaluation, rollout strategies, cost optimization, organizational adoption, and how teams can accelerate their path from proof-of-concept to production.
Timestamps
00:00 Introduction and State of Agent Adoption
00:54 What Thousands of Production Agent Teams Have Learned
01:29 The Three Types of Production AI Agents
01:49 Customer-Facing Agents Explained
02:32 Factorial’s HR Assistant Case Study
03:55 Indeed’s AI Career Counselor
05:01 The Challenges of Cost and Accuracy
06:35 Early Access Programs and Agent Rollouts
07:19 Why Experienced Agent Builders Move Faster
08:22 Avoiding Reinventing the Wheel with Agent Frameworks
09:03 Internal Enterprise Agents
10:07 Enterprise Search and Knowledge Discovery
10:43 Process Automation in Large Organizations
11:22 Aligning Leadership and Engineering Teams
12:23 Why Leaders Should Build Agents Themselves
13:09 Developer Platform Agents
13:28 AI-Powered Site Reliability Engineering (AISRE)
14:20 Internal Agent Platforms at Scale
15:24 Rapid Iteration and Workflow Automation
16:00 The Future of Agent Development
Key Takeaways
• Most production AI applications fall into three categories: customer-facing agents, internal enterprise agents, and developer platform agents.
• The biggest challenges teams face when deploying agents are accuracy, evaluation coverage, and managing inference costs at scale.
• Early-access programs and phased rollouts help teams gather real-world feedback before exposing agents to all users.
• Organizations that have team members with prior production agent experience consistently move faster and avoid common pitfalls.
• Internal developer platforms and reusable agent frameworks allow organizations to scale agent development across hundreds or thousands of engineers.
--
#AIAgents #Mastra #AgentEngineering #LLMOps #AIEngineering #EnterpriseAI #DeveloperTools #AgenticAI #ProductionAI #ArizeObserve
🔗 Try Arize AX & Phoenix OSS: arize.com
🔔 Subscribe for weekly content on LLMs, agents, and evaluation: youtube.com/@arizeai?sub_confirmation=1
What does AI agent adoption actually look like in production?
Drawing on insights from thousands of teams building with Mastra, this session explores the patterns emerging across the AI landscape as organizations move from prototypes to production systems. Learn which types of agents are succeeding, where companies are seeing the most value, and what separates successful deployments from stalled experiments.
The talk breaks down the three major categories of production agents—customer-facing agents, internal enterprise agents, and developer platform agents—and shares real-world examples from companies including Indeed, Factorial, MongoDB, and others. Along the way, you’ll learn practical lessons about evaluation, rollout strategies, cost optimization, organizational adoption, and how teams can accelerate their path from proof-of-concept to production.
Timestamps
00:00 Introduction and State of Agent Adoption
00:54 What Thousands of Production Agent Teams Have Learned
01:29 The Three Types of Production AI Agents
01:49 Customer-Facing Agents Explained
02:32 Factorial’s HR Assistant Case Study
03:55 Indeed’s AI Career Counselor
05:01 The Challenges of Cost and Accuracy
06:35 Early Access Programs and Agent Rollouts
07:19 Why Experienced Agent Builders Move Faster
08:22 Avoiding Reinventing the Wheel with Agent Frameworks
09:03 Internal Enterprise Agents
10:07 Enterprise Search and Knowledge Discovery
10:43 Process Automation in Large Organizations
11:22 Aligning Leadership and Engineering Teams
12:23 Why Leaders Should Build Agents Themselves
13:09 Developer Platform Agents
13:28 AI-Powered Site Reliability Engineering (AISRE)
14:20 Internal Agent Platforms at Scale
15:24 Rapid Iteration and Workflow Automation
16:00 The Future of Agent Development
Key Takeaways
• Most production AI applications fall into three categories: customer-facing agents, internal enterprise agents, and developer platform agents.
• The biggest challenges teams face when deploying agents are accuracy, evaluation coverage, and managing inference costs at scale.
• Early-access programs and phased rollouts help teams gather real-world feedback before exposing agents to all users.
• Organizations that have team members with prior production agent experience consistently move faster and avoid common pitfalls.
• Internal developer platforms and reusable agent frameworks allow organizations to scale agent development across hundreds or thousands of engineers.
--
#AIAgents #Mastra #AgentEngineering #LLMOps #AIEngineering #EnterpriseAI #DeveloperTools #AgenticAI #ProductionAI #ArizeObserve
🔗 Try Arize AX & Phoenix OSS: arize.com
🔔 Subscribe for weekly content on LLMs, agents, and evaluation: youtube.com/@arizeai?sub_confirmation=1
