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

Government companies in Chile are squarely in the spotlight as the Chilean AI Policy 2021–2030 frames public-sector transformation around three practical pillars - enabling factors, development and adoption, and ethics/regulation - creating a roadmap for cost cuts and efficiency gains through targeted pilots and capacity building (Chilean AI Policy 2021–2030).

Backed by world‑class infrastructure and an updated 2024 strategy that positions Chile as a regional AI leader - with 58 data centers and dedicated research funding - public agencies can realistically run low‑risk automation pilots to speed procurement and citizen services (Chile: Leader in Artificial Intelligence in Latin America).

As regulation in the region shifts to risk‑based, human‑centred rules, practical upskilling is essential: short, work-focused courses such as Nucamp's AI Essentials for Work bootcamp teach prompt writing and on-the-job AI use cases that help teams move from theory to measurable savings - one successful municipal records pilot can free staff weeks of repetitive work.

Stanford's task-level generative AI study, produced in close collaboration with Chilean partners, turns the conversation from

what if

to

where to start

: by breaking the 100 most common occupations into tasks it finds that roughly 4.7 million Chilean workers could significantly accelerate more than 30% of routine tasks, a scale equivalent to a measurable national productivity boost (Stanford study on the impact of generative AI on work in Chile).

The analysis - echoed in Stanford GSB's coverage that nearly half of tasks in common jobs are

accelerable

- points to pragmatic, low‑risk pilots for public bodies: about 31% of public‑sector tasks, 65–75% of teachers' administrative tasks, and 44% of SME tasks are good targets for quick wins such as ChatGPT‑style tools for data entry, report visuals, customer support and information retrieval (Stanford GSB analysis: AI and productivity in common jobs).

The takeaway for Chilean government companies is distinct and actionable: prioritize pilots that free staff from repetitive work so human time can be redirected to service delivery and higher‑value public tasks - a concrete path from pilots to measurable savings outlined in the study and in practical guides for priority sectors (Guide to priority AI pilot sectors for Chilean government).

Practical quick wins for Chilean government companies start with small, low‑risk pilots that free staff from repetitive work while keeping people - not models - in charge: SUSESO's use of machine‑learning to triage medical claims, for example, shows how a predictive model can speed processing without replacing human judgment (the agency had roughly 20,000 claims awaiting decisions), and that procurement checklists requiring bias, transparency and data‑protection documentation can push vendors to higher standards (SUSESO AI procurement case study - World Privacy Forum).

Quick technical wins include automating data entry and report visuals, deploying prompts that turn raw datasets into crisp policy briefs for Ministerio de Ciencia teams, and piloting assistant tools for municipal administrative assistants to reclaim hours of form‑work (Complete guide to using AI in Chile's government (2025), Top 10 AI prompts for government policy visualization in Chile).

Pairing these MVPs with simple procurement requirements and human‑in‑the‑loop rules protects vulnerable citizens and delivers measurable efficiency gains fast.

“Success might be defined another way; in situations affecting people's wellbeing and livelihoods, use of a well‑designed and assessed model in support of speedier‑yet‑thoughtful human decisions can constitute success.” - Moya, SUSESO

SUSESO's rollout is a practical how‑to for Chilean public procurement: facing roughly 200,000 medical‑leave claims last year and about 20,000 still awaiting decisions, the agency turned to machine‑learning to triage cases and speed processing, but discovered that procurement rules shape outcomes as much as models do (World Privacy Forum case study on Chile's SUSESO AI deployment).

A mandatory ChileCompra bidding template (since adjusted) forced vendors to submit bias, transparency and data‑protection analyses - pushing suppliers to build documentation and internal buy‑in - but it also revealed a persistent tension: traditional weight on price and competition can under‑reward vendor experience in fairness, explainability and security.

Working with GobLab UAI, SUSESO used algorithmic transparency and bias‑measurement tools to evaluate a gradient‑boosting claims model and an audit of a mental‑health classifier, showing that procurement design can nudge safer deployments if it prioritizes capability and clear human‑in‑the‑loop rules over lowest price alone (GobLab Ethical Algorithms Project: methods and findings, ChileCompra standard bidding terms for AI projects).

“Success might be defined another way; in situations affecting people's wellbeing and livelihoods, use of a well‑designed and assessed model in support of speedier‑yet‑thoughtful human decisions can constitute success.” - Moya, SUSESO

Chile has moved quickly from principles to procurement: GobLab's “Ethical, Responsible and Transparent Algorithms” initiative at Universidad Adolfo Ibáñez - backed by IDB Lab and working with ChileCompra, the Council for Transparency and the Ministry of Science - packages ethics into tools that procurement teams can actually use, not just signal about.

The project published an Ethical Formulation guide (Aug 2022) and helped create ChileCompra's Standard Bidding Terms for AI (Jan 2023), and it now offers audit-ready instruments such as an Algorithmic Transparency Report Card and a Statistical Bias & Fairness Measurement toolkit that were piloted across agencies like FONASA and IPS; the effort even surveyed more than 800 public organisations as it pushed the first Latin‑American regulation on algorithmic transparency in 2023.

For Chilean government companies this means procurement can require bias reviews, transparency reporting and human‑in‑the‑loop rules up front - turning tenders into an active safeguard that protects citizens while unlocking speed and efficiency (see GobLab's project overview and a regional summary of the initiative).

“This not only improves the efficiency of government acquisitions but also strengthens public trust in government management and fosters equal opportunities for suppliers and contractors.” - María Paz Hermosilla

Chile's new “Development and Management of a National Supercomputing Infrastructure Specialized in AI” program is a practical lever for government companies and local tech partners: it offers co‑funding of up to US$7 million per project (covering up to 80% of costs), targets technology companies, AI transfer centres and universities, and gives winners up to five years in two stages to build capacity that can train large models for mining, energy, agriculture and more - effectively expanding limited existing facilities like the NLHPC and CENIA while anchoring an AI campus in renewable‑rich, well‑connected zones to boost digital sovereignty and competitiveness (InvestChile program overview for the National Supercomputing Infrastructure Specialized in AI).

For procurement and IT teams this means a concrete path to on‑shore supercomputing, stronger public‑private collaboration, and a better chance to retain talent and attract investment rather than shipping workloads overseas - a national bet that turns data into domestic capability rather than an exported commodity (Telecompaper regional reporting on the Chile supercomputing program).

“Countries with advanced supercomputing capabilities have demonstrated improvements of over 10% in productivity in sectors such as mining and agribusiness.” - José Miguel Benavente, CORFO

Chile's sector hotspots for practical AI adoption are already clear: education is the standout - Stanford's deep dive finds that roughly 65–75% of teachers' tasks (mostly administrative) are ripe for automation, a change that can free classroom time for richer student interaction (Stanford report on generative AI impact in Chile); large, low‑friction initiatives like Caja Los Andes and ChileMass's free six‑session course for 4,000 educators show how fast upskilling can scale teacher readiness (ChileMass AI teacher training program for 4,000 educators).

SMEs also stand to gain - Stanford flags ~44% of SME tasks as optimisable - so lightweight AI assistants and templated prompts make immediate productivity wins possible.

Social‑service delivery and public bodies should mirror pilots that combine human‑in‑the‑loop rules with simple automation (triage, form processing, report generation), and experiments like AI career‑guidance agents (which increase initial engagement in counseling pilots) illustrate how bots and humans can complement each other at scale (IDB report on scaling AI career-guidance agents in Chile).

The throughline is pragmatic: target high‑time, repetitive tasks, train people fast, and measure time reclaimed so policymakers can see concrete savings.

Start small, start measured: a beginner's roadmap for Chilean government companies begins with a task‑level diagnostic (the Stanford deep dive shows roughly 31% of public‑sector tasks are accelerable) and a short list of “quick wins” - data entry, report visuals, information retrieval and customer support - that free staff for higher‑value work; piloting one tightly scoped assistant with clear human‑in‑the‑loop rules and procurement requirements borrowed from GobLab's ethics tools creates early wins and legal cover for scale.

Pair each MVP with short, role‑specific training (Chile's new 45‑hour virtual AI course for 600 pedagogy students is a concrete model for rapid upskilling), measure time reclaimed in hours or days, and bake audit and bias checks into contracts so vendors deliver explainability as standard.

Use local capabilities where possible (national research centres and CENIA collaborations can help tailor models to Chilean Spanish and contexts), track mission outcomes not just cost saved, and iterate: small pilots with measurable time savings pave the path from experimentation to institutional adoption.

See the Stanford study for task‑level guidance and the teacher training example for scaled, short‑course reskilling models.

“This not only improves the efficiency of government acquisitions but also strengthens public trust in government management and fosters equal opportunities for suppliers and contractors.”

Chile's path forward is practical, not hypothetical: the Stanford deep‑dive shows generative AI could accelerate nearly half of tasks across the 100 most common jobs - about 4.7 million workers - and unlock efficiency equivalent to almost 12% of GDP if well implemented, with the public sector alone seeing roughly 31% of tasks ripe for quick wins like data entry, report visuals and information retrieval (see the Stanford study for task‑level guidance).

Start small, measure hours reclaimed, and pair pilots with short, role‑specific training so gains stick; for example, short applied courses such as Nucamp's AI Essentials for Work bootcamp teach prompt craft and on‑the‑job AI use cases that help teams capture those wins.

Combine these pilots with clear procurement and human‑in‑the‑loop rules, lean on local governance tools and audits, and report outcomes in hours or service quality so policymakers and citizens can see concrete value and trust grows alongside efficiency (Stanford: Deep dive on generative AI in Chile).

“How much of that is achievable really depends on how well we take advantage of this opportunity.” - Gabriel Weintraub