Architecture & ModernizationLunch Talk40min
AI Doesn’t Click Buttons: Building APIs for the Agentic Age
This session explains how to design AI-first APIs for copilots and autonomous agents. It covers contract-first design, typed schemas, explainable errors, machine-readable actions, security, observability, OpenAPI, Java records, and legacy migration strategies, helping developers evolve existing APIs without rewriting applications.
talk.summaryAiDisclaimer
Kim LauwersQuality Coders BV
talkDetail.whenAndWhere
Wednesday, October 7, 13:05-13:45
Room 9
For decades, software has been designed around graphical user interfaces. The application was the product, the UI was the primary entry point, and APIs mainly existed to support it.
AI agents reverse that relationship. They do not click through screens, interpret disabled buttons or understand vague error messages. They interact with APIs. Yet many APIs expose data without meaning, errors without possible resolutions and operations without explaining what can happen next.
This session explores what it means to design an AI-first API: an API built not only for applications and integrations, but also for copilots, automation and autonomous agents.
We will move beyond the idea that adding a chatbot makes a system AI-ready. Through practical examples, we will cover contract-first design, strongly typed schemas, explainable errors, machine-readable actions, deterministic workflows, observability, security, OpenAPI, Java records, migration strategies for legacy systems.
This is not about allowing an LLM to make critical business decisions. It is about giving AI enough structure, context and guardrails to safely use existing business capabilities.
Attendees will learn how to recognise where current APIs fail agents and how to evolve them without rewriting an entire application.
The session is aimed at Java developers, architects, API designers and technical leads. No previous experience with AI agents is required.
AI agents reverse that relationship. They do not click through screens, interpret disabled buttons or understand vague error messages. They interact with APIs. Yet many APIs expose data without meaning, errors without possible resolutions and operations without explaining what can happen next.
This session explores what it means to design an AI-first API: an API built not only for applications and integrations, but also for copilots, automation and autonomous agents.
We will move beyond the idea that adding a chatbot makes a system AI-ready. Through practical examples, we will cover contract-first design, strongly typed schemas, explainable errors, machine-readable actions, deterministic workflows, observability, security, OpenAPI, Java records, migration strategies for legacy systems.
This is not about allowing an LLM to make critical business decisions. It is about giving AI enough structure, context and guardrails to safely use existing business capabilities.
Attendees will learn how to recognise where current APIs fail agents and how to evolve them without rewriting an entire application.
The session is aimed at Java developers, architects, API designers and technical leads. No previous experience with AI agents is required.
Kim Lauwers
Kim Lauwers is a technical designer, Java developer, Scrum Master and people leader with more than 20 years of experience in software development. He enjoys working where software architecture, complex business domains and team dynamics meet.
His interests include domain-driven design, clean architecture, API design and building systems that remain understandable and maintainable as they evolve.
His current work focuses on AI-first APIs, reusable agent skills, effective guardrails and reviewable AI-assisted development workflows.
Outside of technology, Kim practises martial arts and enjoys running—two activities that, much like software engineering, reward discipline, continuous improvement and the occasional willingness to take a hit and keep moving.
His interests include domain-driven design, clean architecture, API design and building systems that remain understandable and maintainable as they evolve.
His current work focuses on AI-first APIs, reusable agent skills, effective guardrails and reviewable AI-assisted development workflows.
Outside of technology, Kim practises martial arts and enjoys running—two activities that, much like software engineering, reward discipline, continuous improvement and the occasional willingness to take a hit and keep moving.