General description

SyRCI-IA is an artificial intelligence solution implemented by Madrid City Council to improve the initial classification and routing of citizen suggestions, complaints and compliments.

The solution forms part of the Madrid City Council Suggestions, Complaints and Compliments Management System (SYRCI) and is functionally led and governed by the Directorate-General for Transparency and Quality – Subdirectorate-General for Quality and Evaluation.

SyRCI-IA combines three complementary mechanisms: deterministic routing rules, specialised artificial intelligence and human oversight. Existing rules remain in use where structured information is sufficient to determine the destination reliably. AI is applied when semantic interpretation of the citizen’s free-text communication is required.

The AI component supports classification by matter, submatter and responsible municipal unit. It does not assess the merits of a complaint, determine the content of the response, issue an administrative decision or replace the competent municipal unit. The receiving unit validates the routing and may reject it where it considers itself not competent. Returned cases are reviewed and manually reassigned by the supervisory team. Human administrative responsibility therefore remains embedded throughout the process.

Needs addressed

Madrid City Council receives tens of thousands of suggestions, complaints and compliments every year concerning a wide variety of municipal services. Before substantive processing can begin, each communication must reach the unit with the relevant competence.

Incorrect initial routing can generate returns, internal consultations, duplicated work and delays before the competent unit begins dealing with the case. Before SyRCI-IA, approximately half of incoming communications could be routed using deterministic rules, while the remainder required manual reading, interpretation and classification.

The solution was created to reduce repetitive manual classification, improve the speed and consistency of initial routing and increase operational capacity while preserving traceability and human control.

Key features

SyRCI-IA uses a hybrid routing model combining deterministic rules and artificial intelligence. The AI component consists of three specialised classifiers based on BERT (Bidirectional Encoder Representations from Transformers), fine-tuned on approximately 90,000 historical records to support classification by matter, submatter and responsible municipal unit.

The solution includes human validation and manual correction within the ordinary administrative workflow, confidence thresholds adapted to different management units, monitoring of incorrect and returned assignments, model comparison, error analysis and retraining.

The AI service is integrated with SYRCI through structured exchanges using FastAPI and JSON. MLflow supports model lifecycle management and Kubernetes supports deployment within Madrid City Council’s municipal infrastructure.

By 9 May 2026, AI had intervened in 38,221 initial routings. Of these, 30,225 were validated as correctly assigned, representing a 79.08% validated accuracy rate. AI accounted for 47.06% of automatic assignments, while deterministic rules and AI together accounted for 97.46% of assignments within the measured scope.

Intended audience

SyRCI-IA is relevant to public administrations and other public-sector organisations receiving large volumes of unstructured communications that must be classified and routed to specialised organisational units.

Potential reuse contexts include suggestions and complaints systems, citizen enquiries, incident-management systems, public information requests, internal service desks and other administrative workflows requiring semantic classification and organisational routing.

Reuse

The principal reusable asset is not Madrid’s specific organisational taxonomy, personal data or trained production models, but the implementation and governance pattern.

Another public administration can adapt the approach by structuring its own historical data and administrative taxonomy, retaining deterministic rules for simple cases, applying text-classification AI where semantic interpretation is required, defining confidence thresholds, maintaining human validation and correction, recording errors and returns, retraining models using validated operational evidence and integrating the classifier as a modular service with its existing case-management system.

Replication does not require access to Madrid’s personal data or trained production models. Each administration can use its own historical records, vocabulary, competences and organisational structure.

Madrid City Council also publishes relevant open datasets, data structures, field definitions and matter/submatter vocabularies. A dedicated SyRCI-IA Reuse and Implementation Blueprint provides further guidance for public administrations interested in adapting the approach.

Interoperability and standards

SyRCI-IA can be described across all four dimensions of the European Interoperability Framework.

Legal interoperability is supported by maintaining AI as an internal routing function without automated decisions producing legal effects. Organisational interoperability is based on common routing criteria, defined responsibilities, supervision and shared monitoring mechanisms across municipal units. Semantic interoperability is supported through common administrative concepts — including matter, submatter, category, responsible unit and territorial references — with relevant structures and vocabularies publicly documented through Madrid City Council’s Open Data Portal. Technical interoperability is achieved through the separation of the AI service from the transactional SYRCI application and structured service interfaces.

The supporting municipal infrastructure operates within the Spanish National Security Framework. SyRCI-IA has also undergone formal security and AI compliance assessment.

Policy contribution

SyRCI-IA contributes to the objectives of the European Interoperability Framework and the wider Interoperable Europe agenda by showing how an AI capability can be embedded in an established public-sector workflow while preserving human accountability, common semantics, modular technical integration and organisational governance.

It also demonstrates a proportionate approach to responsible public-sector AI: technology is restricted to the function for which it creates demonstrable value; its performance is measured using real operational evidence; human supervision is retained; risks are formally assessed; and the system is continuously monitored and improved.

By publishing the solution and its reuse guidance through the Interoperable Europe Portal, Madrid City Council aims to facilitate knowledge sharing and adaptation by other European public administrations.