Using Agents in Production: Past Present and Future // Euro Beinat

2026-03-01 · Source: MLOps.community · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, E-commerce & Digital Commerce · Depth: Advanced, long

Summary

Prozus, a global e-commerce technology company operating in 100 countries, is deploying 30,000 AI agents across its operations by March. These agents serve two primary purposes: transforming e-commerce into a more "agentic, personalized, and ecosystemic" experience, and creating an "AI agentic workforce" to enhance employee productivity, quality, independence, and agility. The company has developed an in-house AI agent builder called Tokan, which started in 2019 and became agent-based in 2024, enabling employees across various departments to create agents. Prozus categorizes agents by seniority (intern, junior, intermediate, senior) based on their access to tools and integrations. Examples include a senior agent for restaurant account management, an intermediate data analyst agent, and a junior newsletter reader agent. The company emphasizes that beyond technical capabilities, successful AI agent adoption is an organizational and cultural challenge, driven by reducing barriers, upskilling, and fostering a "bottom-up" collective discovery process through initiatives like internal competitions.

Key takeaway

For CTOs and VPs of Engineering aiming for widespread AI agent adoption, recognize that technical solutions alone are insufficient. Your strategy must prioritize organizational and cultural shifts, actively reducing perceived barriers for non-technical employees to create agents. Implement programs like internal competitions and upskilling initiatives to foster a "bottom-up" approach, enabling collective discovery and embedding agent creation as a standard practice across the enterprise.

Key insights

Large-scale AI agent adoption is primarily an organizational and cultural challenge, not just a technical one.

Principles

• AI adoption requires bottom-up collective discovery.
• Agent seniority correlates with tool and integration access.

Method

Prozus developed an in-house agent builder, Tokan, to enable non-engineers to create agents, fostering adoption through barrier reduction, upskilling, and internal competitions like "Pros Got Talent."

In practice

• Categorize agents by capabilities (e.g., "intern" to "senior").
• Measure agent impact across productivity, quality, agility, and independence.

Topics

• AI Agents in Production
• AI Agentic Workforce
• E-commerce AI Applications
• AI Agent Development
• AI Adoption Strategy

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

Related on AIssential

• From Multi Agent Systems to Institutional Learning in the Enterprise - with Papi Menon of Outshift by Cisco — Scaling multi-agent AI requires a shared cognitive layer for collective learning and contextual understanding, moving beyond isolated agent performance.
• Most AI Agents Fail in Production Because They’re Built Backwards — Production AI agents require a bottom-up, distributed systems architecture, not a top-down, model-centric approach.
• If more companies block bots and dial down bloat, AI adoption doesn’t collapse. It matures. — AI adoption is shifting from ubiquitous integration to constrained, trust-centric deployment, prioritizing governance and utility.
• How Endava is redesigning software delivery around AI agents — Organizations can achieve AI-native operations by embedding AI agents as the primary problem-solving approach across all workflows.
• The Organization Is the Bottleneck — Organizational maturity and existing engineering foundations dictate AI coding agent success, not the tools themselves.
• Box CEO on AI Agents & Why Enterprise Can't Keep Up | a16z — Enterprise AI adoption is hindered by integration complexities, architectural paralysis, and a mismatch between tech innovation and organizational readiness.

Open in AIssential →

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