May 14–15, 2025 — New York City
Green IO held its first U.S. conference in New York this May, bringing together a technically savvy, action-oriented audience. Hosted by the Green IO podcast and Apidays, the event focused on practical, production-ready strategies for reducing digital systems’ environmental impact—while simultaneously improving their intelligence, scalability, and resilience.
The main themes of the event were clear: APIs are becoming critical infrastructure for AI and agent-based systems, and sustainability is increasingly a core engineering and product strategy, not a side concern.
APIs and AI & Designing for Agents and Automation
A recurring insight throughout the conference was that APIs are now foundational for intelligent systems. Large language models (LLMs), retrieval-augmented generation (RAG), and autonomous agents rely on clean, stable, well-documented APIs to reason, plan, and take action. As AI adoption accelerates, teams that treat API design and governance as a first-class concern will have a measurable advantage.
Business Value and System Design
APIs serve as the connective tissue for agents. When designed well, they allow agents to dynamically orchestrate tools, access data, and interact with services in real-time. Companies that treat their APIs as internal products—complete with semantic clarity, observability, and versioning—are better positioned to scale AI workloads safely and efficiently.
“Agent Experience” (AX) was introduced as a rising counterpart to user experience (UX). Just as UX designers consider clarity, responsiveness, and accessibility for human users, developers building agent-facing APIs must consider consistency in schema, descriptive error messages, and minimal latency.
Best Practices for AI-Ready Infrastructure

Speakers emphasized the importance of:
- Auditing existing APIs for latency, schema depth, and compatibility with AI tools.
- Standardizing interaction protocols, such as the Model Context Protocol (MCP), to let LLMs interact with APIs predictably.
- Keeping microservice patterns simple and composable—for instance, using a FastAPI wrapper around a compact LLM with structured outputs.
- Defining production-readiness to include observability, authentication, rate limiting, and reproducibility—not just uptime.
Risks and Warnings
Deploying LLMs without adequate guardrails is risky. Attendees shared real-world examples of prompt injection, runaway token costs, brittle toolchains, and failed deployments caused by insufficient logging or schema mismatches. The LLM ecosystem remains fragmented and relatively immature, which makes careful tooling decisions even more important.
Sustainability in Tech & Practical Pathways to Impact

The second major theme of the conference focused on sustainability—not as a branding exercise, but as a measurable, financially aligned component of software and systems engineering. Sessions covered everything from emissions tracking and infrastructure choices to carbon-aware design practices and emerging regulations.
Sustainability as Cost Control and Risk Mitigation
Several speakers made the business case for integrating sustainability into everyday tech decisions. Mastercard, for example, reduced both cost and environmental impact by aligning emissions metrics with financial dashboards, automating cloud shutdowns, and embedding sustainability into their FinOps practices—resulting in $28M in savings and 800 metric tons of CO₂ emissions avoided.
Other companies shared experiments like routing traffic to greener cloud regions or replacing large general-purpose models with smaller, task-specific ones. These choices not only reduced carbon impact but also improved performance and reduced cloud spend.
Best Practices and Tools
Engineering teams were encouraged to:
- Track emissions using open-source tools such as Cloud Carbon Footprint, CodeCarbon, SCI, CCF, and Boavista.
- Optimize software assets by eliminating wasteful features, minimizing bundle sizes, and refactoring queries.
- Apply the Climate Product Leaders playbook to embed sustainable practices into product management, AI workflows, infrastructure decisions, and cross-functional strategy.
- Blend FinOps and GreenOps so that carbon becomes a first-class metric alongside cost, uptime, and latency.
Regulations and Governance
A growing body of international policy and standards—like the EU Ecodesign Directive, Web Sustainability Guidelines from the W3C, and ISO accessibility rules—are pushing digital sustainability into the realm of compliance. Companies that start embedding sustainability into their governance frameworks now will be better positioned to meet upcoming mandates and avoid technical debt.
The overlap between sustainability and inclusion also surfaced: lightweight, accessible systems tend to be more energy-efficient, user-friendly, and equitable by design.
Key Recommendations
Across sessions, speakers returned to a few core recommendations:
1. Prepare APIs for AI.
Treat APIs as strategic infrastructure. Design for agents. Add observability, authentication, semantic structure, and clear error handling.
2. Secure and monitor AI deployments.
Use tools and protocols that reduce risk. Avoid deploying LLM-based systems without clear safeguards for reproducibility, authorization, and cost control.
3. Track what matters.
Use emissions tracking tools like Cloud Carbon Footprint and align environmental KPIs with business outcomes.
4. Build for sustainability from the start.
Think beyond carbon offsets. Start with low-impact architectures, cleaner infrastructure, and intentional design choices.
5. Stay ahead of regulation.
The sustainability standards space is evolving quickly. Build governance structures now to future-proof your stack.
Resources and Next Steps
To learn more about the topics covered at Green IO New York 2025, explore the following:
- Green IO Podcast – Deep dives into sustainable digital practices
- Cloud Carbon Footprint – Open-source tool for cloud emissions tracking
- Climate Product Leaders Playbook – A practical framework for product teams
- Web Sustainability Guidelines (W3C) – Standards for digital sustainability and accessibility
- Hugging Face Open LLM Leaderboard – Open models ranked by performance and sustainability, helping teams make informed, low-impact model choices
Conclusion
Green IO New York 2025 showed how technical strategy, intelligent systems, and sustainability are converging. From robust APIs that support agents to carbon-aware cloud choices that reduce cost and impact, the tools and patterns are available. What matters now is adoption—and leadership.