Marape: Government will use AI and ICT to strengthen merit-based decisions in 2026
Prime Minister James Marape says the Government will deploy artificial intelligence (AI) and information and communications technology (ICT) next year to improve governance, transparency and efficiency across the public service.
He framed AI as the "engine room" for decision support in core government functions while keeping human oversight in place. "Technology will help us remove subjectivity and strengthen fairness in how decisions are made," he said.
Where AI will support public service decisions
- Public service recruitment and selections
- Contract awards and vendor due diligence
- Project proposal assessments and prioritisation
- Law and policy development support
- Performance monitoring, compliance checks and system efficiency
Marape said the objective is to move decisively to a merit-based society by reducing human bias and long-standing issues like personal preferences, corruption, nepotism and manipulation of systems. AI will assess qualifications, experience, performance and compliance against clear criteria, with human accountability retained.
2026: Transition year with frameworks, safeguards and pilots
Marape noted this direction has been flagged before and will now shift from intent to implementation. The priority in 2026 is to establish the guardrails and test what works before scaling.
- Ethics and risk standards aligned to proven guidance such as the OECD AI Principles and the NIST AI Risk Management Framework.
- Clear, measurable criteria for merit-based assessments; documented data sources and quality checks.
- Algorithmic transparency: decision logs, explainability, and auditability.
- Human-in-the-loop controls, conflict-of-interest checks and an appeals pathway for impacted parties.
- Procurement guardrails: security, privacy, bias testing, performance benchmarks and vendor accountability.
- Independent oversight and periodic reviews to protect democratic processes.
What agencies can do in Q1
- Map high-volume, rules-based decisions where AI can assist without removing human judgment (e.g., shortlisting against objective criteria).
- Define objective, measurable merit criteria and thresholds for each decision type.
- Audit and clean data; remove sensitive attributes that create bias; document data lineage.
- Draft a standard operating procedure for AI-assisted workflows, including approvals and escalation.
- Select one low-risk pilot with a clear success metric (accuracy, time-to-decision, cost per case, variance reduction).
- Set up model monitoring for drift, error rates and fairness across demographic groups.
- Train relevant teams on prompt quality, verification steps and accountability controls. For structured upskilling by role, see AI courses by job.
Practical guardrails for day-one use
- Do not fully automate high-stakes decisions; require human review before final outcomes.
- Publish plain-language summaries of criteria used in AI-assisted decisions.
- Mandate dual-control for procurement and major recruitment outcomes.
- Keep complete audit trails: inputs, model version, prompts, outputs and human approvals.
- Schedule quarterly bias tests and accuracy reviews; decommission tools that fail standards.
- Create an incident response process for erroneous or harmful outputs.
Why this matters for trust
Marape stressed that technology will assist decision-makers, not replace democratic processes. "This transition is about building trust in public institutions. When systems are fair, transparent and data-driven, confidence in government increases."
The signal is clear: move from policy talk to practical delivery. Start small, measure results and expand what proves fair, fast and reliable.
Questions for department heads
- Which decisions can be made more objective with clear criteria and quality data?
- What data do we need to clean, label or integrate before any pilot?
- What risks (privacy, security, bias) must be mitigated, and how will we test them?
- Where do we place human approval, and what's our appeals process?
- What outcomes will we track to prove value (e.g., time saved, error reduction, fairness)?
Implementation starts with disciplined pilots, strong oversight and measurable wins. Do that, and the public will feel the difference where it counts: fair opportunities, cleaner processes and decisions that stand up to scrutiny.
Your membership also unlocks:
