Portugal to Use AI for Faster Licensing, Human Oversight Intact and Tacit Approvals on Delays

AI in Licensing: Faster Decisions, Clear Accountability

Portugal's minister of State Reform, Gonçalo Matias, signalled a shift: bring AI into licensing to cut decision times while keeping a human at the end of the decision. The plan leans on prior notifications, tacit approval when deadlines are missed, and stricter accountability for public employees.

For government teams, this is not about replacing people. It's about moving staff from repetitive checks to higher-value work and enforcing service deadlines that the public can rely on.

How the new licensing model works

• Prior notification: most activities can start without a license unless explicitly listed as exceptions.
• Tacit approval: if the administration misses the legal deadline, authorization is granted by default.
• Human-in-the-loop: AI accelerates analysis; human reviewers close out sensitive or high-impact cases.
• Accountability: clear responsibility if legal standards or timelines are not met.

What this means for your team

• Process redesign: map end-to-end steps, remove redundant checks, and set where AI screens for completeness and risk.
• Deadlines as product features: timer-driven workflows, alerts for aging cases, and escalations before deadlines hit.
• Data quality first: structured forms, validation at intake, and clean registries to reduce manual back-and-forth.
• Transparency: log every AI recommendation, decision rationale, and override for audit and appeals.

Guardrails to put in place now

• Risk-tiering: low-risk = straight-through via notification; medium = AI triage + human review; high = full human assessment.
• Legal and privacy: document legal basis, data sources, and retention; run a Data Protection Impact Assessment (DPIA).
• Procurement controls: require model performance thresholds, bias testing, security certifications, and incident reporting.
• Appeals and redress: publish clear channels, timelines, and escalation paths; pause timers during active appeals.

Metrics that matter

• Median time to decision by license type
• Share of cases within statutory deadlines
• Percentage of tacit approvals (target this to fall as responsiveness improves)
• Appeal and rework rates, split by AI-assisted vs. manual cases

Workforce shift: from paperwork to higher-value work

Expect roles to tilt toward complex case evaluation, compliance inspections, policy analysis, and vendor oversight. Equip staff to work with AI outputs, interpret risk signals, and manage exceptions with confidence.

Implementation checklist (first 90 days)

• Pick 1-2 high-volume, low-risk license types for a pilot with prior notification.
• Define statutory timers and publish a service charter that aligns with tacit approval rules.
• Stand up an intake portal with structured data and automated completeness checks.
• Launch dashboards for case aging, escalations, and tacit approval warnings.
• Train reviewers on AI-assisted workflows, bias risks, and audit logging.

For policy context on safe AI deployment in the public sector, see the European Commission's overview of the AI Act here. For practical upskilling options for public-facing roles, browse role-based AI courses here.