Labs - ONE Conference 2026

Skip to main content

RefundAI is a 3-hour, hands-on lab where intermediate and expert OutSystems ODC developers build a production-pattern Multi-Agent Refund & Retention Orchestrator from scratch. The system deploys four coordinated AI agents — a Central Orchestrator, a Policy Agent, a Loyalty & CLV/Loyality Agent, and a Sentiment Agent—that simultaneously and autonomously evaluate every refund request across three dimensions and route it to the optimal resolution path. Attendees will implement the full Supervisor/Worker orchestration pattern using ODC AI Agent Builder, configure Human-in-the-Loop (HITL) escalation using ODC Workflow, and publish a RefundDecisionMade Pub/Sub event to notify downstream enterprise systems. A pre-built data model and UI shell are provided so attendees spend 100% of lab time on agentic logic and governance.

This hands on lab guides participants through building a voice driven clinical documentation agent using OutSystems (ODC). Attendees will implement an AI powered workflow that converts dictated speech into structured medical data, automatically extracting key clinical entities and populating a standardized JSON schema.

The solution incorporates validation, normalization, and human in the loop review to ensure accuracy and compliance. Participants will also generate a formatted patient report from the structured data.

The lab highlights how OutSystems enables rapid development of agentic AI systems with enterprise grade governance, including auditability and role based access control.

By the end, attendees will have a working prototype of an autonomous clinical assistant that improves documentation efficiency, consistency, and integration readiness.

In this lab, participants will design and build an autonomous IT service desk agent using the OutSystems platform. The session will cover how to create AI agents capable of understanding user requests, managing context, and executing IT workflows such as ticket creation, status tracking, and knowledge base retrieval. Attendees will integrate Agents with enterprise systems to enable intelligent triage and automated resolution. Best practices for governance, and security in enterprise environments will also be addressed. By the end of the lab, participants will have a functional prototype of a service desk agent that reduces manual effort, improves response times, and enhances the overall support experience.

In this hands-on lab, participants will design and build an omnichannel AI agent using the OutSystems platform, leveraging agentic AI patterns to create autonomous digital teammates. The session will explore how to orchestrate agents, integrate enterprise data sources, and enable interactions across channels such as web and messaging platforms like Telegram. Attendees will implement core capabilities including intent handling, context management, and tool invocation, while applying best practices for scalability, governance, and security. By the end of the lab, participants will have a working prototype of an intelligent agent capable of executing tasks, collaborating with users, and adapting to dynamic business scenarios. This session is ideal for developers and architects looking to operationalize AI-driven automation within modern enterprise applications.

User Stories pile up in "Ready for Validation" while QA teams struggle to keep pace. In this hands-on Lab, participants build an intelligent two-layer agentic system that eliminates this bottleneck. The Orchestrator — an OutSystems AI Agent built with AI Agent Builder — reads User Stories from a project management tool, parses acceptance criteria into testable scenarios, and delegates visual test execution to OpenClaw via Model Context Protocol. OpenClaw navigates the running application, fills forms, clicks buttons, validates outputs, and captures screenshot evidence. The Orchestrator evaluates results, generates evidence reports, and updates workflow status autonomously. Every decision is governed by configurable approval gates and logged in a full audit trail. Participants leave with a fully functional, reusable orchestration pattern applicable beyond QA: compliance checks, data extraction, automated onboarding, and any scenario requiring governed delegation to external agents.

This lab will guide participants through building an agent system that, given an invoice, determines who in the organization needs to approve it, in what order and requests those approvals as needed.

Existing O11 applications manage employee info, organisational structure, as well as leave and approval responsibility delegation. The approval rules are highly dynamic, and captured in natural language.

The agent system will be implemented as a workflow loop that relies on an agent to identify the next approver in the approval chain – based on the organisational data, the invoice and the approval rules; passes that information to another agent that decides who is the final approver – using vacation data and delegation information and, finally; awaits for the approver to sign off on the invoice. When no more approvers are identified, the process concludes. A human-in-the-loop pattern can be (optionally) implemented to guarantee the correct approval chain.

ONE 2026 Labs

A hands-on track designed for builders

Refund AI: Building a Multi-Agent Refund & Retention Orchestrator

Build a Voice-Driven Clinical Report Agent (ODC + O11): From Speech-to-Text to Structured Patient Reports

Building an Autonomous IT Service Desk Agent

Designing an Omnichannel Agent in OutSystems

The Autonomous QA Agent: OutSystems Orchestrates OpenClaw for Intelligent Test Validation

Building an Dynamic Approval Chain using Agent Orchestration