Artificial Intelligence for Public Sector Professionals Training Course

Course overview

Public-sector teams are under pressure to improve service access, process high-volume information, identify risk earlier and respond consistently to citizens, while protecting statutory duties around fairness, privacy, transparency and records management. Artificial intelligence can support these aims, but poorly framed pilots can introduce bias, unsupported automated decisions, uncontrolled data sharing and vendor lock-in. This course equips professionals to assess, procure, govern and deploy AI use cases that stand up to operational, legal and public scrutiny.

Participants examine machine learning, generative AI, retrieval-augmented generation and decision-support systems through public-sector scenarios such as benefits triage, planning enquiries, casework summarisation, fraud detection and citizen-contact analysis. They learn to define a problem statement, map stakeholders and affected groups, assess data readiness, apply the NIST AI Risk Management Framework, set human-oversight controls, test for bias and hallucination, and specify measurable service outcomes. The course also covers AI procurement requirements, model documentation, audit trails, privacy impact assessment inputs and monitoring arrangements.

Teaching combines instructor-led briefings with structured workshops, policy and procurement case studies, prompt-testing labs and peer challenge sessions. Participants work in teams on a realistic public-service AI proposal and leave with an AI implementation pack: a use-case canvas, risk register, data and governance checklist, evaluation plan, procurement question set and 90-day pilot roadmap. This makes the course suitable for professionals who need to move from AI interest or isolated experiments to controlled, evidence-based adoption across a department, agency or local authority.

Course objectives

By the end of this course, participants will be able to:

• Define an AI use-case canvas linking a public-service problem to users, decisions, benefits and measurable service outcomes
• Classify proposed AI systems by automation level, decision impact, data sensitivity and required human oversight
• Apply the NIST AI Risk Management Framework to identify, assess and treat AI risks across the system lifecycle
• Evaluate data readiness using provenance, quality, representativeness, retention and access-control criteria
• Design human-in-the-loop workflows with escalation rules, override authority and auditable decision records
• Test generative AI outputs for hallucination, bias, harmful content and retrieval-grounding failures
• Draft AI procurement requirements covering supplier evidence, model documentation, security, interoperability and exit arrangements
• Produce a 90-day AI pilot roadmap with governance gates, evaluation metrics, monitoring controls and accountable owners

Benefits of attending

Target competencies

Who should attend

• Digital Transformation Managers — who must convert AI opportunities into governed service-improvement initiatives
• Policy Managers — who need to assess how AI affects policy delivery, fairness and accountability
• Service Delivery Managers — who oversee high-volume citizen-facing processes suitable for assisted automation
• Data and Analytics Leaders — who must establish reliable data, model assurance and monitoring practices
• Procurement Officers — who evaluate AI supplier claims and need defensible contractual requirements
• Information Governance and Risk Professionals — who must address privacy, records, security and algorithmic accountability

Requirements and prerequisites

Participants should have experience of a public-sector service, policy, operational, digital, data, procurement or governance function and be comfortable discussing process maps, service metrics and risk controls. Familiarity with spreadsheets, structured data concepts, information governance and basic project delivery is useful. Participants should bring one candidate service problem or AI proposal from their organisation where possible. Programming, statistics beyond simple averages and percentages, machine-learning model building, and prior use of Azure AI Studio are not required. The course explains core technical concepts in operational language before applying them to governance and implementation decisions.

Training methodology

The course uses instructor-led explanation to establish technical and governance foundations, then applies them through public-sector cases involving citizen services, enforcement, grants and internal operations. Participants map a live or realistic use case, inspect sample model outputs, test prompts and retrieval results, complete risk and data-readiness assessments, and compare supplier evidence. Small-group review panels challenge each team’s assumptions on fairness, privacy and accountability. The final two days are structured as an implementation workshop, producing a governed pilot plan that participants can adapt with sponsors and assurance teams.

Course outline

Day 1: Public-sector AI opportunities and boundaries

• AI capability categories: predictive, generative, optimisation and computer vision
• Public-service value hypotheses and measurable service outcomes
• Decision-support versus automated decision-making classifications
• High-impact use cases in benefits, licensing, casework and contact centres
• Public value, equity and administrative burden considerations
• AI lifecycle from problem definition to retirement
• Use-case prioritisation with impact, feasibility and risk scoring

Workshop: Participants create a scored AI opportunity map for a public-service process and select one candidate use case.

Day 2: Data foundations for accountable AI

• Data inventories, ownership and lawful access
• Data provenance and lineage for training and retrieval sources
• Data quality dimensions: completeness, accuracy, timeliness and consistency
• Representativeness and coverage analysis for affected populations
• Personally identifiable information and sensitive-category data handling
• Records retention, deletion and archival implications
• Data-readiness assessment templates and evidence logs

Workshop: Participants complete a data-readiness assessment and evidence log for their selected AI use case.

Day 3: Machine learning and generative AI for decision-makers

• Supervised learning, classification and regression in service operations
• Generative AI, large language models and token-based output generation
• Retrieval-augmented generation architecture and source grounding
• Training, fine-tuning, prompting and inference distinctions
• Model performance measures: precision, recall, false positives and false negatives
• Hallucination, prompt injection and data leakage failure modes
• Appropriate task selection for predictive and generative systems

Workshop: Participants compare predictive and generative AI design options for a casework and citizen-enquiry scenario.

Day 4: Responsible AI, fairness and human oversight

• Fairness concepts: disparate impact, equal treatment and procedural fairness
• Bias sources in historical data, labels, proxies and model deployment
• Explainability requirements for staff, citizens and oversight bodies
• Human-in-the-loop, human-on-the-loop and manual fallback models
• Escalation thresholds and override authority design
• Accessibility and inclusion requirements for AI-enabled services
• Documenting reasons, appeals and contestability mechanisms

Workshop: Participants design a human-oversight workflow with escalation, override and appeal routes for an AI-assisted eligibility decision.

Day 5: AI risk management and assurance

• NIST AI Risk Management Framework Govern, Map, Measure and Manage functions
• Risk taxonomy for accuracy, bias, privacy, security and operational resilience
• AI risk registers with likelihood, impact, controls and ownership
• Model cards and system cards as assurance documentation
• Pre-deployment testing protocols and acceptance criteria
• Incident response for harmful outputs and model failures
• Independent review, audit trails and assurance reporting

Workshop: Participants build an AI risk register and assurance evidence plan using the NIST AI RMF structure.

Day 6: Privacy, security and information governance

• Privacy impact assessment inputs for AI-enabled services
• Purpose limitation, data minimisation and access-control design
• Microsoft Purview data classification and sensitivity labels
• Security threats: prompt injection, model extraction and insecure plugins
• Supplier data-processing roles and cross-border data considerations
• Knowledge-base governance for retrieval-augmented generation
• Information lifecycle controls for prompts, outputs and logs

Workshop: Participants produce a privacy, security and records-control checklist for an internal generative AI assistant.

Day 7: Building and evaluating public-sector AI prototypes

• Azure AI Studio project structure and model deployment concepts
• Prompt templates, system instructions and role-based constraints
• Grounded responses using approved retrieval sources
• Test-set creation from representative service scenarios
• Evaluation criteria for accuracy, relevance, safety and citation quality
• Red-team testing for harmful, misleading and adversarial prompts
• Version control for prompts, data sources and evaluation results

Workshop: Participants test a grounded assistant in Azure AI Studio and document evaluation results against a public-service test set.

Day 8: AI procurement and supplier governance

• Outcome-based AI requirements and evaluation criteria
• Supplier due diligence for training data, model limitations and performance claims
• Contract clauses for audit rights, incident notification and change control
• Interoperability, API access and portability requirements
• Service-level measures for availability, accuracy and response handling
• Intellectual property, data ownership and reuse provisions
• Supplier exit plans and continuity arrangements

Workshop: Participants draft a supplier evaluation scorecard and procurement question set for an AI case-management solution.

Day 9: Operating model, monitoring and adoption

• AI governance roles: service owner, model owner, data steward and assurance lead
• ISO/IEC 42001 AI management system principles
• Approval gates from discovery through pilot and scaled deployment
• Operational monitoring for drift, error rates and unequal outcomes
• User training, acceptable-use guidance and staff consultation
• Citizen communication, transparency notices and feedback channels
• Benefits realisation dashboards and review cadences

Workshop: Participants define an AI operating model with RACI responsibilities, monitoring measures and decision forums.

Day 10: Pilot planning and executive decision making

• Pilot scope definition and minimum viable control set
• Baseline measurement and success-metric design
• Pilot cohort selection and phased rollout planning
• Implementation dependencies, budget assumptions and resource estimates
• Go-live readiness reviews and stop-go decision criteria
• Executive briefing structure for AI investment decisions
• Ninety-day action planning and stakeholder engagement

Workshop: Participants present a complete AI pilot implementation pack and receive a peer review against governance, value and feasibility criteria.

Tools & standards covered

Azure AI Studio, Microsoft Purview, NIST AI Risk Management Framework 1.0, ISO/IEC 42001

A typical training day

08:30 – 10:30	First session
10:30 – 10:45	Refreshment break
10:45 – 12:30	Second session
12:30 – 13:30	Lunch and networking
13:30 – 15:00	Third session
15:00 – 15:15	Refreshment break
15:15 – 16:30	Workshop and daily review

Live online deliveries follow the same structure in the East Africa Time zone, with shorter screen blocks and longer breaks.

What the fee includes

How you can take this course

Certification

Participants who complete the full five days receive the Skillset Development Certificate of Completion, stating the course title, course code, dates and delivery format — suitable for professional-development records and employer reimbursement.

Frequently asked questions