The OECD.AI Policy Navigator

Our policy navigator is a living repository from more than 80 jurisdictions and organisations. Use the filters to browse initiatives and find what you are looking for.

Intelligent Complaint Analysis System (SIAR)


Added by:   OECD analyst
Added on:   16 Nov 2026
Updated by:   OECD analyst
Updated on:   16 Nov 2026

undefined
The award-winning Intelligent Complaint Analysis System (SIAR) automates analysis of mass consumer complaints, including during power outages. It helps identify severe impacts and reduce processing time, enabling faster institutional responses and stronger evidence for fair compensation. SIAR was developed to overcome the unsustainable bottleneck created by manually processing thousands of complaints.

Name in original language

Sistema Inteligente de Análisis de Reclamos (SIAR)

Initiative overview

Consumer complaints filed with businesses through our institution are our primary input for identifying non-compliance and infringements of the Chilean Consumer Protection Law. However, their manual analysis is slow and resource-intensive. During mass-impact events with a high volume of complaints (typically between 3,000 and 12,000), manual processing becomes unsustainable, compromising the timeliness and effectiveness of our institutional response. Traditionally, an official requires eight to 31 business days to analyse this volume. This limitation was starkly exposed during the 2024 power outage that affected nine regions nationwide due to adverse weather, generating an unprecedented volume of complaints against electricity companies—specifically 32,753 in a matter of days. Even with a dedicated team of four officials, manual analysis would have taken at least 21 business days.

The core problem lies in the inability to quickly assess the true impact of these emergencies, identify specific patterns of harm, and generate timely evidence for protection actions. This bottleneck historically left consumers vulnerable and without prompt solutions. The main challenge was to overcome this structural limitation: we needed a way to rapidly process massive complaint volumes to expedite decisions and ensure compliance with consumer protection regulations.

To solve this, we developed the Intelligent Complaint Analysis System (SIAR, for its Spanish acronym). SIAR leverages artificial intelligence to automate complaint analysis. Using semantic analysis, Python, and specialised natural language AI models, SIAR deepens the evaluation process to quickly detect new issues and severe cases (such as the loss of food or medicine in a power outage). Crucially, the model was trained directly by expert officials using established legal criteria and consumer protection regulations, ensuring the AI's outputs are highly accurate and institutionally reliable.

Regarding SIAR’s future evolution, its high adaptability drives two strategic lines of action: internal scaling and external replicability. Internally, the tool has already transcended its original emergency-response purpose. Currently in production, SIAR now features a user-friendly interface for non-technical staff, removing the need for programming skills. As we prepare to launch internal training, the system will be integrated into daily administrative workflows, automating routine tasks to ensure high-standard analysis during both national contingencies and normal operations. Externally, its replicability has been actively explored in other public institutions that, as part of their core business, must analyse large volumes of text-based data—such as the National Migration Service. This dual expansion demonstrates its immense potential to scale, transform mass case management, and ultimately deliver faster, data-driven justice across the broader public sector.

Other relevant details

Results, outcomes and impacts: SIAR was evaluated using quantitative and qualitative indicators, showing drastically improved efficiency. It reduced analysis time for approximately 33,000 complaints from 21 business days to just five days, increasing the processing rate from one complaint per minute to up to 17 complaints per minute. Qualitatively, it expanded problem characterisation from four to 12 categories, precisely identifying severe cases like food or medicine loss as a result of the power outage. This generated robust evidence for extrajudicial negotiations, securing over USD 19.2 million in compensation for 2.5 million affected consumers. Where agreements failed, this evidence supported class-action lawsuits—a success validated by the 2025 Transformative Institutions Award. Challenges and lessons learned: A key challenge was rapidly adapting an existing, generic internal AI tool during a national emergency to accurately process specialised consumer texts. We learned that interdisciplinary collaboration is fundamental: involving technical, legal, and consumer analyst teams is crucial to define ethical and accurate training criteria. Furthermore, training the AI with specific regulatory criteria greatly improves the relevance and sensitivity of the analysis. To mitigate risks, constant validation and human review proved indispensable for ensuring accuracy, traceability, and institutional trust in critical decisions. Finally, for an initiative like this to succeed, institutional urgency combined with leadership committed to evidence-based decision-making is necessary to accelerate development and scalability.

About the policy initiative


Category:

  • AI policy initiatives, programmes and projects

Initiative type:

  • AI use cases/projects in the public sector

Status:

  • Active

Start Year:

  • 2024

Target Sectors:


OECD AI Principles:

—

Images: