AI Fundamentals for Policy Makers
A short interactive course that builds AI literacy from definitions up to drafting a basic adoption plan.
- Audience
- Government leaders and policy professionals
- Duration
- 20 minutes
Fondamentaux de l'IA pour les decideurs publics
Un mini-cours interactif qui explique l'IA en langage simple, montre des cas d'usage publics et aide a cadrer un premier plan d'adoption.
- Public
- Dirigeants publics et professionnels des politiques
- Duree
- 20 minutes
Fundamentos de IA para decisores publicos
Um curso curto e interativo que desenvolve literacia em IA, mostra casos de uso no setor publico e orienta os primeiros passos de adocao.
- Publico
- Lideres governamentais e profissionais de politicas publicas
- Duracao
- 20 minutos
AI readiness brief
From concept to policy checklist
A visual walk-through of the first four decisions leaders make before launching an AI pilot.
Brief de preparation IA
Du concept au premier cadre de decision
Un apercu visuel des decisions essentielles a prendre avant de lancer un pilote IA dans le service public.
Resumo de prontidao em IA
Do conceito ao primeiro quadro de decisao
Uma leitura visual das decisoes essenciais antes de lancar um piloto de IA no setor publico.
What you will be able to do
- Define artificial intelligence in plain terms
- Explain common public sector AI use cases
- Identify risks and benefits of an AI initiative
- Distinguish between narrow AI and general AI
- Evaluate an AI procurement proposal using a basic checklist
- Outline the first steps of an AI adoption plan
Ce que vous saurez faire
- Definir l'intelligence artificielle avec des mots simples
- Reconnaitre des cas d'usage utiles dans le secteur public
- Evaluer risques, benefices et garde-fous d'une initiative IA
- Distinguer l'IA specialisee de l'IA generale
- Poser les bonnes questions avant un achat ou un pilote
O que sera capaz de fazer
- Definir inteligencia artificial em linguagem simples
- Explicar casos de uso no setor publico
- Identificar riscos, beneficios e salvaguardas
- Distinguir IA estreita de IA geral
- Esbocar um primeiro plano de adocao
Adoption snapshot
Start with one narrow workflow
Keep the pilot limited enough to audit end to end.
Use data you already trust
Check completeness, representativeness, and ownership early.
Keep a human in the loop
Review outputs before any high-impact action reaches a citizen.
Apercu de l'adoption
Commencer par un seul flux de travail
Le pilote doit rester assez etroit pour etre observe et audite de bout en bout.
S'appuyer sur des donnees deja fiables
Verifiez tot la qualite, la representativite et la responsabilite sur les donnees.
Maintenir une revue humaine
Aucun resultat sensible ne devrait atteindre un citoyen sans verification.
Panorama da adocao
Comecar por um fluxo de trabalho estreito
O piloto deve ser pequeno o suficiente para ser observado e auditado de ponta a ponta.
Usar dados ja confiaveis
Verifique cedo qualidade, representatividade e responsabilidade sobre os dados.
Manter revisao humana
Nenhum resultado sensivel deve chegar ao cidadao sem verificacao.
Typical AI pilot effort mix
Most of the work is not model building. It is governance, data preparation, and review design.
Repartition type d'un pilote IA
La majeure partie du travail concerne la gouvernance, les donnees et l'organisation de la revue, pas seulement le modele.
Distribuicao tipica do esforco num piloto de IA
Grande parte do trabalho esta na governanca, nos dados e no desenho da revisao, nao apenas no modelo.
Reflect and flip
Each card starts with a prompt. Flip it to reveal the coaching cue or model answer.
Reflechir et retourner
Chaque carte commence par une question. Retournez-la pour voir la piste de reponse ou le modele de reponse.
Refletir e virar
Cada cartao comeca com uma pergunta. Vire-o para ver a orientacao ou a resposta-modelo.
Course content
What is Artificial Intelligence?
Artificial intelligence refers to computer systems that learn from data and use that learning to make predictions, recommendations, or decisions. Unlike traditional software — which follows rules a developer writes explicitly — an AI model improves as it is exposed to more examples. Think of a spam filter that gets better the more emails it sees, or a translation tool that improves as it processes more text.
There are two broad categories worth knowing. Narrow AI does one task well: recognising speech, ranking search results, or flagging unusual transactions. General AI — capable of reasoning across many domains like a person — does not yet exist outside research labs. Everything in public service today is narrow AI.
Consider a service your department delivers. Write down one repetitive, data-heavy task in that service. Could a system that learns from past examples make that task faster or more consistent? Keep that example in mind as you work through the rest of this course.
Public Sector Use Cases
Governments worldwide are piloting AI in four broad areas. Document processing uses machine learning to read, sort, and extract data from forms, applications, and reports — cutting processing times from weeks to hours. Citizen-facing assistants answer common questions around the clock using information drawn from published policies. Predictive analytics helps teams anticipate demand — for hospital beds, social services, or infrastructure maintenance — before a crisis emerges. Fraud and anomaly detection flags unusual patterns in grants, benefits, or procurement data for human review.
The common thread is augmentation, not replacement. The AI handles volume and pattern-matching; the official handles judgement, exceptions, and accountability.
A social welfare office processes 4,000 applications a month. A vendor proposes an AI tool to pre-screen applications and flag incomplete or inconsistent submissions for staff review. The system will not approve or reject — it will prioritise the queue. Think about: what data would the system need? Who reviews its recommendations? What happens when it is wrong?
Risks and Governance
AI introduces risks that differ from traditional software. Bias is the most discussed: if the training data reflects past unfair decisions, the model will reproduce them at scale. A hiring tool trained on historical CVs may disadvantage certain groups not because a developer chose that outcome, but because the data carried that pattern.
Explainability matters in public services because decisions affecting citizens must be defensible. A model that produces a score without a reason is hard to challenge or audit. Procurement teams should ask vendors: can this system explain, in plain terms, why it produced a given output?
Data quality is a practical risk. Garbage in, garbage out still applies. Before adopting AI, assess whether your data is complete, accurate, and representative. Pilot small, measure carefully, and keep a human in the loop for any decision with significant consequences.
Name one decision in your area of responsibility where an incorrect AI recommendation could harm a citizen. Now name the safeguard you would put in place before deploying any tool in that decision pathway.
Planning Your First AI Initiative
A good first initiative has four characteristics: it is narrow in scope, it uses data you already have, success is measurable, and a human reviews outcomes. Start with a problem statement, not a technology. 'We want to use AI' is not a problem statement. 'Staff spend 40% of their time manually sorting applications that could be categorised by rule' is.
From there, the steps are: map the data you have and check its quality; define what a correct output looks like and how you will measure it; run a small pilot with real cases; audit the results for errors and bias; only then consider scaling. Build the oversight mechanism — the human review process — before you build the tool.
Contenu du cours
Comprendre l'IA
L'intelligence artificielle designe des systemes capables d'apprendre a partir de donnees pour produire une prediction, une recommandation ou un classement. Dans les administrations, il s'agit aujourd'hui d'IA specialisee, centree sur une tache precise.
Notez une tache repetitive dans votre service. Si les agents utilisent toujours les memes indices pour la traiter, il existe peut-etre un point de depart pour un pilote.
Cas d'usage publics
Les usages les plus realistes sont le tri de dossiers, l'assistance aux usagers, la detection d'anomalies et la prevision de la demande. Le principe cle reste le meme: l'outil accelere le volume, l'agent garde le jugement.
Un fournisseur propose un systeme pour prioriser les demandes sociales incompletes. Quelles donnees alimentees? Qui verifie les recommandations? Comment un citoyen peut-il contester une erreur?
Gouvernance et risques
Les risques principaux sont le biais, l'absence d'explication claire et des donnees de mauvaise qualite. Dans le secteur public, un resultat utile ne suffit pas: il doit aussi etre defendable et verifiable.
Citez une decision de votre perimetre ou une recommandation erronee pourrait nuire a un citoyen. Quel garde-fou imposeriez-vous avant tout deploiement?
Lancer un premier pilote
Commencez par un probleme concret, des donnees deja disponibles et un indicateur simple de succes. Pilotez petit, mesurez, auditez, puis decidez si l'echelle superieure est justifiee.
Conteudo do curso
O que e IA?
IA refere-se a sistemas que aprendem com dados para apoiar previsoes, recomendacoes e decisoes. Na administracao publica, o uso pratico atual e de IA estreita, voltada para tarefas especificas.
Pense em um processo do seu departamento que gere volume e repeticao. Esse processo poderia ser triado ou classificado com apoio de exemplos historicos?
Casos de uso no setor publico
Os casos mais comuns incluem processamento documental, assistentes para cidadaos, analise preditiva de demanda e deteccao de fraude. O valor esta em acelerar volume sem retirar responsabilidade humana.
Um escritorio de beneficios quer usar IA para priorizar pedidos incompletos. Que dados seriam necessarios? Quem revisa os resultados? O que acontece quando o sistema erra?
Riscos e governanca
Bies nos dados, baixa explicabilidade e qualidade fraca de dados podem comprometer a justica e a confianca. Em servicos publicos, transparencia e revisao humana nao sao opcionais.
Em qual decisao do seu contexto um erro de IA teria maior impacto para o cidadao? Qual salvaguarda deve existir antes do piloto?
Planeando a primeira iniciativa
Comece com um problema mensuravel, dados disponiveis e um fluxo de revisao humana. Teste em pequena escala, audite resultados e so depois pense em expandir.
Knowledge check
Answer each question.
1. Which of these statements about artificial intelligence are accurate? Select all that apply.
Select all that apply
2. A department wants to reduce time spent answering repeat citizen questions. Which of these are reasonable first AI use cases? Select all that apply.
Select all that apply
3. Which of these are genuine risks to plan for before adopting AI in public services? Select all that apply.
Select all that apply
4. Which of these statements about explainability requirements in public sector AI is correct?
Select all that apply
5. What should happen before scaling an AI pilot in government? Select all that apply.
Select all that apply
6. Which of these statements correctly describe 'narrow AI' as used in public services today? Select all that apply.
Select all that apply
You scored /
This is the kind of immediate, low stakes feedback that keeps learners moving. Every sample course is built the same way: outcomes first, then practice, then feedback.
Request a course like thisVerification des acquis
Repondez a chaque question. La correction est faite cote serveur et la bonne reponse n'est pas envoyee au navigateur a l'avance.
1. Lesquelles de ces affirmations sur l'intelligence artificielle sont exactes ? Selectionnez toutes les reponses correctes.
Select all that apply
2. Un departement veut reduire le temps passe a repondre aux questions repetitives des citoyens. Lesquels de ces cas d'usage sont raisonnables comme premier projet IA ? Selectionnez toutes les reponses correctes.
Select all that apply
3. Lesquels de ces elements sont de veritables risques a anticiper avant d'adopter l'IA dans les services publics ? Selectionnez toutes les reponses correctes.
Select all that apply
4. Laquelle de ces affirmations sur les exigences d'explicabilite de l'IA publique est correcte ?
Select all that apply
5. Que devrait-il se passer avant de faire passer un pilote IA a plus grande echelle ? Selectionnez toutes les reponses correctes.
Select all that apply
6. Lesquelles de ces affirmations decrivent correctement l'« IA etroite » utilisee aujourd'hui dans les services publics ? Selectionnez toutes les reponses correctes.
Select all that apply
Votre score est de /
Ce type de retour immediat et sans enjeu excessif aide les apprenants a poursuivre. Chaque cours exemple suit la meme logique : objectifs, pratique, puis feedback.
Demander un cours similaireVerificacao de conhecimentos
Responda a cada pergunta. A avaliacao acontece no servidor e a resposta correta nao e enviada antecipadamente ao navegador.
1. Quais destas afirmacoes sobre inteligencia artificial sao corretas? Selecione todas as opcoes corretas.
Select all that apply
2. Um departamento quer reduzir o tempo gasto a responder a perguntas repetidas dos cidadaos. Quais destes casos de uso sao razoaveis como primeiro projeto de IA? Selecione todas as opcoes corretas.
Select all that apply
3. Quais destes sao riscos genuinos a considerar antes de adotar IA nos servicos publicos? Selecione todas as opcoes corretas.
Select all that apply
4. Qual destas afirmacoes sobre os requisitos de explicabilidade da IA publica esta correta?
Select all that apply
5. O que deveria acontecer antes de escalar um piloto de IA? Selecione todas as opcoes corretas.
Select all that apply
6. Quais destas afirmacoes descrevem corretamente a IA estreita usada atualmente nos servicos publicos? Selecione todas as opcoes corretas.
Select all that apply
A sua pontuacao foi /
Este tipo de feedback imediato e de baixo risco ajuda o formando a continuar. Cada curso de exemplo segue a mesma estrutura: objetivos, pratica e depois feedback.
Pedir um curso como este