Agent workflows · 2026-06-22

UNIDES AI project ideas: examples across 11 policy areas

Explore UNIDES AI project ideas across 11 policy areas and select a feasible concept using social impact, human oversight, data protection, and measurement criteria.

Where should AI fit in a UNIDES project?

AI should be a tool that addresses a demonstrated student problem, not the sole justification for a UNIDES project. Define the audience and need first, then add AI only when automation, accessibility, information access, or personalization creates real value.

Strong concepts deliver measurable youth outcomes instead of a technical demonstration. People review model output, sensitive data is minimized, and ownership after the project is defined from the start.

Problem definition comes before technology.
Compare AI with simpler alternatives.
Require human oversight for high-impact decisions.
Measure outcomes for young people, not only model performance.

Examples for the 11 policy areas

These ideas are not ready-made applications. Each community must validate the local need, define its audience, and adapt the concept to the eligibility requirements in the official guidance.

A project may touch several areas, but its core problem and expected result should stay focused. A limited pilot around one measurable benefit is more feasible than a broad platform promise.

Disaster resilience: a campus emergency assistant grounded in verified sources.
Family and values: an archive that classifies intergenerational oral-history records.
Science and technology: a safe generative-AI and prompt-literacy lab.
Environment and climate: a system that classifies campus waste reports and extracts trends.
Education: an accessibility tool for summaries and study questions from course materials.
Youth information: a RAG assistant limited to approved university sources.
Youth health and sport: a resource navigator that does not produce diagnoses.
Volunteering and participation: opportunity matching by skills, location, and schedule.
Employment and entrepreneurship: anonymized résumé feedback and interview practice.
Social inclusion: conversion of visual, text, and audio material into accessible formats.
International youth work: a multilingual partner-information and cultural-preparation assistant.

Data protection, security, and human oversight

Student records, health data, disability information, résumés, and interviews may be sensitive. Collect only necessary data, keep identifiers out of prompts, and define retention and access controls.

AI must not make high-impact decisions about scholarships, health, discipline, recruitment, or support eligibility on its own. Source display, error reporting, human approval, and safe fallback belong in the prototype.

First consider solving the problem without collecting personal data.
Never embed an API key in a browser or mobile client.
Separate tenant and community data on the server.
Constrain model output with sources, scope, and human approval.

Turn the idea into an application and prototype

Evaluate evidence of need, audience size, the ability to build a pilot in eight to twelve weeks, data access, technical ownership, and sustainable cost together. The first version should prove one core flow reliably.

For idea selection, application structure, model/API comparison, and prototype scope, send your university, community, target audience, and technical need to info@llmtr.com. LLMTR is not an official program operator; the relevant public authority makes evaluation and funding decisions.

Frequently asked questions

Can a UNIDES project include AI?

An applicable and safe AI component can support a demonstrated youth need across policy areas including science and technology; final eligibility must be checked against the next official guidance.

What is the best UNIDES AI project idea?

There is no universal best idea. The strongest candidate solves a locally validated need within a limited period, with available data and measurable benefit for young people.