How AI Is Helping Government Companies in Portugal Cut Costs and Improve Efficiency

Portugal's AI Portugal 2030 strategy is turning abstract promises into practical savings for public services by prioritising workforce upskilling, shared computing and experimentation: the plan funds AI education and regional upskilling, creates sandboxes like LabX and InnoLabs, and even adds supercomputing muscle - the Vision system at the University of Évora delivers up to 10 petaflops for machine‑learning tests - so agencies can prototype automation before broad rollout (Portugal AI 2030 strategy report - European AI Watch).

That focus on skills plus shared infrastructure aims to reduce call‑centre loads, speed decision making and cut repetitive costs while preserving human oversight, a balance highlighted in policy summaries (AI Portugal 2030 policy overview - Digital Watch Observatory).

Public managers and staff can build the practical prompts-and-tools skills to make those pilots stick by taking short, work‑focused courses such as Nucamp's Nucamp AI Essentials for Work syllabus.

Bootcamp	Length	Early bird cost	Registration
AI Essentials for Work	15 Weeks	$3,582	Register for AI Essentials for Work bootcamp

Portugal's national AI strategy is deliberately collective and governance‑forward: AI Portugal 2030 was launched in February 2019 after six public discussion meetings and sets out seven strategic pillars - from boosting AI skills and research to using the country as a “living laboratory” for smart cities, energy and biodiversity - while promoting public‑private networks and sandboxes for experimentation (AI Portugal 2030 national strategy - OECD AI Observatory).

Responsibility is split across specialised bodies so policy is both coordinated and reviewable: the Foundation for Science and Technology (FCT) steers the strategy, INCoDe.2030 coordinates delivery with partners such as ANI, Ciência Viva and AMA, and annual reviews are carried out by the Ministries of Science, Technology and Higher Education and of Economy and Digital Transition - mechanisms designed to keep pilots, ethics work (including an AMA guide) and regulatory design in lockstep with citizen engagement and market uptake (AI Portugal 2030 national strategy report - European Commission AI Watch).

Body	Role
Foundation for Science and Technology (FCT)	Primary responsibility for developing the national AI strategy
Ministry of Science, Technology & Higher Education	Primary responsibility for implementation; conducts annual reviews
INCoDe.2030 Coordination Structure (with ANI, Ciência Viva, AMA)	Promotion, inter‑governmental coordination and stakeholder mobilisation

Portugal's approach to building AI-ready public services hinges on concrete skilling pipelines such as the UPskill programme, a partnership led by IEFP, APDC and polytechnic institutes that combines six months of intensive ICT training with a three‑month on‑the‑job internship and even a €1,200 salary guarantee for placements - an arrangement designed to turn unemployed candidates into deployable talent while matching firms' needs (participants have included Accenture, Microsoft and OutSystems).

Launched under the Digital Transition Action Plan and advertised via streamlined channels (applications are handled through a chatbot), UPskill has scaled across multiple editions - by 2023 some 1,683 trainees had participated and the 2024 edition added eleven new technical courses like Cloud (Azure), Java and embedded systems.

Complementary measures such as the Industry Qualify scheme and rural-focused digital training show how Portugal blends employer co‑design, short intensive courses and paid internships to keep pipelines flowing; for a concise programme dossier see the UPskill programme dossier at CEDEFOP and the UPskill open applications announcement at IEFP.

Element	Detail
Training model	6 months intensive ICT courses + 3 months internship
Lead partners	IEFP, APDC, CCISP (polytechnic institutes)
Salary guarantee	€1,200 for some internship hires
Trainees (by 2023)	1,683 participants

“Creating opportunities for lifelong learning will be key.” - Branka Modrusan, PwC's Academy

Portugal's cost-cutting play for AI leans on shared compute: the RNCA (National Network for Advanced Computing) stitches together HPC, HTC, cloud and operational centres so research, industry and public bodies can request access to advanced computing services rather than buying and running their own clusters (RNCA National Network for Advanced Computing (FCCN) - Portugal official site); at the same time INCD's new identity as the CNCA positions a single national facilitator to manage RNCA initiatives and deepen Portugal's role in EuroHPC and international projects like iMagine, strengthening coordination and technical support for users (INCD to CNCA rebranding announcement - LIP press release).

The national roadmap already names nodes such as UC‑LCA, INCD/CNCA and MACC and even the EuroHPC Deucalion as part of this ecosystem, with a PetaScale machine approved for MACC - an approach that concentrates expertise, avoids duplicated capital outlay, and lets agencies scale model training and analytics on shared systems instead of buying their own supercomputer.

Automation and process optimisation in Portugal increasingly blends legal reform, workflow redesign and digital assistants so citizens and staff both see faster outcomes: the Simplex legislative package moves many urban‑licensing checks from “prior control” to the execution or completion phase and requires municipalities to use a standardised electronic planning platform, allowing developers in some cases to submit project documents and start construction without prior validation (Simplex legislative package explainer - Ecovis Portugal); complementary evaluation of Simplex argues that debureaucratisation is a 360º change demanding cross‑cutting KPIs, staff development and political commitment rather than only ICT fixes (Effectiveness of Simplex - GEE working paper).

At the service level, Portuguese public bodies are pairing rule changes with automation: conversational agents like the ePortugal Sigma virtual assistant deliver 24/7 plain‑language support and measurably reduce call‑centre load, while RPA pilots target backlogs in claims and permitting workflows rather than purely cost cutting - a contrast noted in comparative studies that also flag staff buy‑in and ease of implementation as common bottlenecks (ePortugal Sigma virtual assistant - government AI use case (Portugal)).

The result is a pragmatic stack: legal simplification, standardised digital forms, targeted automation and human oversight combine so approvals and routine citizen requests become noticeably quicker without sacrificing accountability.

“Seek advice on real estate transactions, because building regulations have not become any easier. Gonçalo Areia, Managing Partner, RBMS – Member of ECOVIS International, Lisbon, Portugal”

Portugal's banks and public bodies are turning to real‑time AI to stop scams before money moves, and Lisbon‑born Feedzai is a visible example: its 2025 industry report highlights that more than half of modern fraud now uses AI and that nine in ten banks already deploy AI to detect scams and speed investigations (Feedzai 2025 AI Fraud Trends report), while its new Feedzai IQ product promises privacy‑preserving, network‑wide intelligence so institutions can share risk signals without exposing customer data (Feedzai IQ privacy-preserving network intelligence announcement).

That mix of behavioural analytics, device signals and federated learning is already delivering concrete wins in Portugal - detecting complex mule networks and reducing false positives - which makes the case for public sector adopters who need to protect citizens and cut manual triage costs while keeping human oversight in the loop; after all, when scams are engineered with cloned voices and photo‑real fakes, speed matters as much as accuracy.

Metric / Result	Value
Consumers protected (global)	1 billion
Events processed per year	70 billion
Payments secured annually	$8 trillion
Novobanco result using TrustScore	43% ↑ fraud detection; 41% ↑ value detection; 0.66% alert rate

“Feedzai TrustScore helped us uncover and fight new fraud patterns that would either have gone undetected or been much harder to effectively block.” - Duarte Pupo Correia, Novobanco

Portugal is already piloting AI where it tangibly cuts costs and keeps services moving: predictive‑maintenance models trained on local datasets help metros and fleets spot wear before a breakdown, while smart routing and aerial options shrink delivery times and reach isolated or disaster‑hit areas.

A Porto‑based effort - the MetroPT benchmark dataset - provides the kind of real operational telemetry that makes eXplainable Predictive Maintenance work for urban rail operators (MetroPT predictive-maintenance benchmark dataset on Zenodo), and a national study of consumer attitudes shows that drone parcel delivery is viable if services prove useful and easy to use: Portugal's drone delivery market is projected to grow 1.8% annually to $7.6 million by 2028, with perceived usefulness the strongest driver of adoption among 1,108 surveyed citizens (Portugal drone parcel delivery acceptance study - Future Business Journal).

The takeaway for public managers is practical: couple local datasets, explainable models and clear user journeys (and privacy safeguards) to move from pilots to cost‑saving scale - imagine fewer out‑of‑service trains and parcels tumbling down from the sky to a mountain hamlet when a road is blocked.

Dataset	Creators / Affiliations
MetroPT	Bruno Veloso (U. Portucalense & INESC TEC), João Gama (FEP - UP & INESC TEC), Rita Ribeiro (FCUP - UP & INESC TEC), Pedro Pereira (INESC TEC)

Portugal's most concrete sector pilots live in health, where AI-powered triage and conversational agents are already bending patient flows and staff time in measurable ways: hospital group CUF reports that Ada's digital triage nudged 49% more patients toward primary care, cut uncertainty and completed 53% of assessments outside normal hours, while conversational and pre‑triage voice agents from Infermedica (available in Portuguese) shave 3–4 minutes off calls and have helped a call centre reach average interviews as short as 4 minutes 57 seconds - turning a typical 15‑minute triage into a calm five‑minute handover and freeing clinical time for complex cases.

These pilots prove a practical playbook - structured symptom intake, clinical reasoning that's explainable and 24/7 coverage - that reduces unnecessary ER referrals, smooths EHR handoffs and creates data streams public managers can reuse for faster emergency prioritisation and situational awareness.

Metric	Value / Source
Increase in patients choosing primary care	49% - Ada at CUF (Ada digital triage case study at CUF hospital group)
Assessments completed outside hours	53% - Ada at CUF (Ada digital triage case study at CUF hospital group)
Typical reduced triage time	4 min 57 sec (case) / Pre‑triage voice agent saves 3–4 minutes - Infermedica (Infermedica pre‑triage voice agent for call centers)
Symptoms & conditions coverage	Collects ~4× more symptoms and covers ~2× more conditions vs. traditional RBTP - Infermedica

“We needed a clinical triage tool that could effectively map to the services we offer and fulfill the whole patient journey, at scale, 24/7.” - Dr Micaela Seemann Monteiro, CUF

Portugal's nearshore ecosystem is a real cost‑and‑time accelerator for public–private AI collaboration: a growing startup scene that taps U.S. funding, know‑how and networks and many U.S. companies establishing business and service centres means governments can run pilots with local teams who speak multiple languages and understand European regulation (U.S. State Department 2024 Investment Climate Statement for Portugal).

That proximity makes it simpler to co‑design conversational assistants, fraud analytics or shared service hubs with vendors who can iterate overnight rather than across time zones; the payoff is lower procurement friction and faster deployment of useful services such as the ePortugal Sigma virtual assistant, which already shows how 24/7 plain‑language support reduces call‑centre load (ePortugal Sigma virtual assistant case study).

Picture a Lisbon hub routing bilingual citizen queries while a partner startup tunes models the same afternoon - a vivid example of nearshore collaboration turning experiments into scaled public services without the usual cross‑border drag.

“many U.S. companies”

Metric	Value (Source)
U.S. FDI inflows (2023)	€2.1 billion - U.S. State Dept.
Unemployment rate (2023)	6.5% - U.S. State Dept.

Portuguese public bodies tend to assemble a pragmatic stack of off‑the‑shelf and custom tools - large language models and NLP for document summarisation and extraction, private‑hosted deployments to keep sensitive records inside controlled environments, conversational virtual assistants for 24/7 citizen support, and targeted RPA to clear backlogs - so repetitive work is automated while oversight stays human.

Case work such as the Withum example shows how fine‑tuned LLMs and extraction pipelines can pull names, qualifications, licences and even security‑clearance details from resumes and scale recruitment workflows without leaking data (Withum case study: AI-powered resume management), and Portugal's ePortugal Sigma virtual assistant is a local example of conversational AI reducing call‑centre load with plain‑language answers (ePortugal Sigma virtual assistant case study).

As agencies adopt these tools, the AIA phased timeline and milestones provide a checklist for safe rollout between 2025–2026, reminding managers that technology choices must match governance, privacy and skills plans (AIA phased timeline and milestones), so the result is faster services without sacrificing control - think fewer manual form checks and more time for complex cases.

Beginners tackling AI for Portuguese public services should keep a tight checklist of high‑impact, local examples: the national AI Portugal 2030 strategy lays out seven pillars from education to “living labs” for cities, energy and biodiversity (see the OECD summary for the strategy), while the European AI Watch report highlights concrete wins such as Ciência Viva's outreach (270 high‑school students in a two‑week AI summer school) and a MOOC that drew about 2,000 participants - showing that skills and awareness are already scaling; link these training efforts to shared compute capacity like the Vision supercomputer (up to 10 petaflops) and RNCA access, where 18 AI‑related projects out of 129 won advanced‑computing support, and the pathway from lab to market becomes visible.

Add a few practical case studies to the list - AI Moonshot Challenge funding (EUR 500k) for satellite+AI environmental work and fast drone deliveries by Connect Robotics during COVID - plus local rollout examples such as the ePortugal Sigma virtual assistant that reduces call‑centre load - and beginners will have a compact, actionable map of where skills, compute and pilots meet to cut costs and speed services in Portugal (OECD summary of AI Portugal 2030 national strategy, European AI Watch report on Portugal AI strategy, Nucamp case study: ePortugal Sigma virtual assistant).

Data point / Case	Value / Note
AI Portugal 2030 pillars	7 strategic pillars (education, research, living labs, public services, etc.)
Ciência Viva summer school (2019)	270 high‑school students (two‑week AI programme)
AI MOOC first session	~2,000 participants
RNCA advanced computing access (call)	18 AI projects out of 129 awarded supercomputing/cloud access
Vision supercomputer (U. Évora)	Maximum performance: 10 petaflops
AI Moonshot Challenge	EUR 500,000 prize for AI + space sustainability solutions

Start with a short, practical roadmap: map your AI use cases and run proportionate AI impact assessments tied to GDPR and the AIA's phased timeline (key provisions applied from February 2025 with full compliance expected by August 2026), then harden data governance, logging and human‑oversight rules before any rollout; Portugal's designated supervisory authorities (ANACOM coordinates the list) should be in the loop early to avoid surprises and administrative fines under the AIA that can reach up to €35 million or 7% of global turnover for the gravest breaches (Portugal AI legal guide - Chambers & Partners (Artificial Intelligence 2025 AIA compliance and penalties)).

Use national assets and sandboxes to de‑risk pilots - LabX, InnoLabs and RNCA nodes (including the Vision supercomputer) let teams test models on shared infrastructure before procurement or scale‑up (Portugal AI 2030 strategy and infrastructure - European AI Watch (LabX InnoLabs RNCA)).

Lock contracts around IP, data rights and liability, plan conformity assessments for high‑risk systems, pair deployments with staff upskilling and public consultation, and treat transparency, audit trails and cybersecurity (NIS2/AIA intersections) as non‑negotiable - so pilots move from promising demos to reliable services that cut costs and protect citizens (Nucamp AI Essentials for Work syllabus - AIA phased timeline and compliance milestones).

Portugal's path from pilots to payoff is tangible: a recent Implement Consulting Group study on the economic opportunity of generative AI in Portugal estimates generative AI could lift Portugal's GDP by EUR 18–22 billion (about +8% in a peak year) and sees 60% of jobs working alongside generative AI - but harnessing that requires stronger skills, clearer rules and data readiness (Implement Consulting Group study).

The public sector still lags - only 26% of organisations have integrated AI despite 64% recognising cost‑saving potential - so practical moves matter: start with tight data governance, sandboxed pilots on shared compute, and short, work‑focused training to build prompt and tool skills.

For managers seeking immediate, usable training, the Nucamp AI Essentials for Work 15-week bootcamp syllabus teaches prompt writing and workplace AI use in 15 weeks, while benchmarking adoption and risks is helped by recent reporting on public sector progress (TechMonitor report on AI adoption in the public sector (EY survey)).

Treat skills, governance and modest pilots as the core trio - turn routine tasks into reclaimed time for higher‑value public service delivery and make the economic opportunity real.

Metric	Value / Source
Projected GDP boost from generative AI	EUR 18–22 billion (+8% peak year) - Implement Consulting Group
Jobs working with generative AI	60% - Implement Consulting Group
Public sector organisations with AI integrated	26% - TechMonitor (EY survey)
Public sector seeing cost‑savings potential	64% - TechMonitor (EY survey)
Public sector generative AI adoption	12% - TechMonitor (EY survey)

“The initial focus has paid off for pioneers who have developed a more effective digital and data foundation... They have made faster progress in embedding data capabilities organisation‑wide.” - Permenthri Pillay, EY Global Government & Public Sector Digital Modernisation Leader