Platforms for Humans and Machines: Engineering for the Age of Agents — Juan Herreros Elorza

2026-04-08 · Source: AI Engineer · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Intermediate, long

Summary

Banking Circle, a global cross-border payments provider processing over €1 trillion annually for 700+ financial institutions, established a platform engineering team to support its 250+ technical staff. This team developed "Atlas," an internal developer platform designed to abstract infrastructure complexity and enable developers to focus on core banking systems, APIs, data science, and integrations. Atlas comprises sub-platforms for compute (Kubernetes-based), infrastructure (blob storage, databases, secret management), messaging, and observability. The platform aims to resolve common developer pain points, such as manual deployment processes and dependency provisioning, which become critical bottlenecks for AI agents. The core idea is to make the platform self-service, API-based, and local-first, supported by robust documentation and encouraging contributions.

Key takeaway

For AI Engineers and MLOps teams building or integrating AI agents into development workflows, prioritize platform self-service, API-driven interfaces, and local-first execution. This approach minimizes manual dependencies and enables agents to operate autonomously, significantly boosting productivity. Leverage the current focus on AI to advocate for and implement these foundational best practices, ensuring your platform is agent-ready and measurable for success.

Key insights

Platform engineering principles like self-service and API-first design are crucial for enabling AI agent productivity.

Principles

• Best practices are amplified by AI agents.
• Shift left to fail early and iterate quickly.
• Measure platform success with DORA and support metrics.

Method

Design internal developer platforms with self-service, API-first interfaces, and local-first execution. Provide structured, accessible documentation and observability for AI agents to close feedback loops.

In practice

• Expose platform capabilities via well-defined APIs.
• Keep documentation near code or in a centralized API-accessible hub.
• Use `agents.md` or skills to guide AI agent behavior.

Topics

• Banking Circle
• Atlas Platform
• Platform Engineering
• AI Agents
• Self-Service APIs

Best for: AI Engineer, MLOps Engineer, AI Architect

Related on AIssential

• Measuring Agents in Production — Production AI agents are basic, human-supervised tools, not fully autonomous systems, delivering tangible value.
• Top 10: AI Automation Platforms — AI automation platforms are evolving into sophisticated orchestrators, democratizing complex workflows through natural language interfaces.
• Podcast: Engineering Stable, Secure and Scalable Platforms: A Conversation with Matthew Liste — Platform engineering prioritizes stability, security, and scalability, using customer journeys to manage risk and adapt to AI's accelerating impact.
• The Organization Is the Bottleneck — Organizational maturity and existing engineering foundations dictate AI coding agent success, not the tools themselves.
• Spec-driven development: the AI engineering workflow at Notion | Ryan Nystrom — AI agents streamline engineering workflows, automating meeting prep, code generation, and spec-driven development to boost velocity.
• From backend support to AI orchestrators: Reimagining the way the Digital Workplace team operates — Agentic AI orchestrates complex IT workflows, transforming manual Digital Workplace operations into scalable, autonomous systems with human oversight.

Open in AIssential →

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