AUT-001

KAI System – AI-based Semi-Automatic Reimbursement of Medical Services Fees

Download PDF
Austria Europe & Central Asia High income Full Production Deployment Confirmed

IT-Services der Sozialversicherungen GmbH (IT-SV); Österreichische Gesundheitskasse (ÖGK); Dachverband der österreichischen Sozialversicherungsträger (Federation of Austrian Social Insurance Carriers)

At a Glance

What it does Perception and extraction from unstructured inputs — Operational and process automation
Who runs it IT-Services der Sozialversicherungen GmbH (IT-SV); Österreichische Gesundheitskasse (ÖGK); Dachverband der österreichischen Sozialversicherungsträger (Federation of Austrian Social Insurance Carriers)
Programme KAI System – AI-based Semi-Automatic Reimbursement of Medical Services Fees
Confidence Confirmed
Deployment Status Full Production Deployment
Key Risks Operational and system integration risks
Key Outcomes Two-thirds (66%) of reimbursement requests processed semi-automatically by October 2022 (ÖGK Jahresbericht 2022).
Source Quality 4 sources — News article / media, Report (multilateral / development partner), Government website / press release, +1 more

The KAI system (Künstliche Intelligenz in der Kostenerstattung) is an artificial intelligence platform deployed by Austria's social insurance system to semi-automatically process reimbursement claims for medical services provided by elective (non-contracted) doctors and therapists. The system was developed by IT-Services der Sozialversicherungen GmbH (IT-SV), the shared IT service provider for Austria's social insurance carriers, with an initial project budget of EUR 2,808,120 and a project duration of 15 months. The primary operational user is the Österreichische Gesundheitskasse (ÖGK), Austria's largest statutory health insurance fund, which processes the majority of cost reimbursement requests nationwide.

In Austria's statutory health insurance system, insured persons who receive treatment from elective doctors or therapists (Wahlärzte/Wahlbehandler) — providers who do not have a direct contract with the health insurance fund — must initially pay for these services out of pocket. They can subsequently submit invoices to their responsible social insurance carrier and receive a partial reimbursement of costs (Kostenerstattung). As of 2017, approximately 5.4 million reimbursement applications were submitted annually across Austria, with volumes increasing at a rate of approximately 8 percent per year. This rising volume, combined with the high manual processing burden, created significant operational challenges: backlogs in processing, high complexity in individual cases, substantial personnel requirements, and the risk of inconsistent decision-making where identical applications could receive different outcomes depending on the individual caseworker.

The KAI system was designed to address these challenges by automating the end-to-end reimbursement workflow from invoice receipt through to approval or rejection, without requiring manual intervention for straightforward cases. The AI pipeline encompasses four core processing steps that were previously performed manually: (1) capture and recognition of submitted documents, including scanned medical invoices, forms, and correspondence from elective practitioners; (2) creation of the processing case in the administrative system; (3) recognition and encoding of medical service positions, including automated ICD-10 diagnosis encoding; and (4) plausibility checks to verify the consistency and correctness of extracted data, including invoice amounts and payment details such as IBAN numbers. The system employs machine learning for document recognition, optical character recognition (OCR) for data extraction, and supervised learning algorithms for diagnosis encoding and plausibility verification. The models are continuously trained and refined based on caseworker feedback and correction data.

The procurement process for the KAI system followed a two-stage competitive procedure. In a preliminary project, IT-SV conducted a negotiated tendering process that identified six bidders considered to be leaders in artificial intelligence. Framework agreements were concluded with all six companies. In the subsequent main project, a renewed call for competition was issued to these six pre-qualified vendors, of which five submitted bids. The best bidder was selected for implementation. The project consortium included IT-SV as project lead, the Hauptverband der österreichischen Sozialversicherungsträger (now the Dachverband der Sozialversicherungsträger), and several regional health insurance funds including the Kärntner Gebietskrankenkasse (KGKK), the Oberösterreichische Gebietskrankenkasse (OÖGKK), the Sozialversicherungsanstalt der gewerblichen Wirtschaft (SVA), the Wiener Gebietskrankenkasse (WGKK), and the Versicherungsanstalt für Eisenbahnen und Bergbau (VAEB). The project was initiated in 2019.

By October 2022, according to ÖGK's annual report (Jahresbericht 2022), two-thirds (66 percent) of reimbursement requests were being processed semi-automatically through the KAI system. The system was intended to enable the social insurance carriers to handle the continuously rising application volumes without requiring proportional increases in staffing. Additional expected benefits included standardised decision-making across cases, improved fairness for insured persons through consistent treatment of identical applications, shorter waiting times for reimbursement, and the ability for individual caseworkers to be relieved of monotonous tasks while receiving support for more complex cases. IT-SV has indicated that the ongoing quantified benefits — primarily time savings in processing reimbursement applications — substantially exceed the ongoing operational costs, including expenditure on further development and continued model training, allowing the initial investment to be amortised through efficiency gains.

However, the system's real-world performance has attracted scrutiny. A 2024 investigation by AlgorithmWatch reported that doubts exist as to whether the KAI system has actually accelerated reimbursement processing. One documented patient case showed that average reimbursement times increased from 27 days in 2021 (based on 15 submitted invoices) to 54 days in 2023 (based on 17 submitted invoices). An ÖGK representative attributed delays to the system's machine learning nature, stating that it was still in a learning phase. IT-SV declined to comment publicly on the system's performance, citing ongoing procurement policies. AlgorithmWatch also noted that Austria's freedom of information law, which was passed in January 2024, would not apply to entities such as IT-SV until 2025, limiting public transparency regarding the system's operational performance and decision logic.

The KAI system operates with a human-in-the-loop oversight model. Controls are built into the processing pipeline at defined value thresholds, for flagged critical cases, and through sampling-based quality checks. Initially, human oversight interventions were frequent and were designed to be gradually reduced as system accuracy improved over time. Caseworkers retain the ability to review and override the system's outputs at multiple points in the workflow. The decision criticality is assessed as moderate: while the system directly affects the speed, accuracy, and consistency of financial reimbursement decisions for insured persons, the amounts involved are partial reimbursements for individual medical services rather than core benefit eligibility determinations, and human review is retained for critical and high-value cases.

The system processes sensitive personal data, including medical invoices containing patient health information, ICD-10 diagnosis codes, and financial data such as IBAN numbers and reimbursement amounts. This data is classified as special category data under the EU General Data Protection Regulation (GDPR) and Austrian data protection law. The system operates under Austria's data protection framework, which implements the GDPR. IT-SV has noted that the KAI platform is also designed for extensibility, with further AI-supported use cases within the social insurance system planned to build upon the existing AI components developed for cost reimbursement. In 2023, IT-SV initiated a new procurement for AI services across six lots via framework agreements, with an execution period from February 2024 to February 2028, reflecting the broader institutional commitment to scaling AI within Austria's social insurance infrastructure. IT-SV's total planned investment in artificial intelligence amounts to EUR 52.5 million.

Classifications follow the DCI AI Hub Taxonomy. Hover over field labels for definitions.

Social Protection Functions

Implementation/delivery chain
Management of contributions and withdrawals primaryProvision of payments/services
SP Pillar (Primary) The social protection branch: social assistance, social insurance, or labour market programmes. Social insurance
Programme Name KAI System – AI-based Semi-Automatic Reimbursement of Medical Services Fees
Programme Type The type of social protection programme, classified under social assistance, social insurance, or labour market programmes. View in glossary Health Insurance
System Level Where in the social protection system the AI is applied: policy level, programme design, or implementation/delivery chain. View in glossary Implementation/delivery chain
Automation Subtype For operational automation cases: (a) document processing and generative staff assistance, or (b) workload and resource forecasting. (a) Document processing and generative staff assistance
Programme Description Kostenerstattung (cost reimbursement) process within Austria's statutory health insurance system, administered by the Österreichische Gesundheitskasse (ÖGK). Insured persons who receive treatment from elective (non-contracted) doctors and therapists pay out of pocket and subsequently submit invoices for partial reimbursement. The KAI system automates the processing of these reimbursement applications, encompassing document recognition, data extraction, ICD-10 diagnosis encoding, and plausibility checking.
Implementation Type How the AI output is produced: Classical ML, Deep learning, Foundation model, or Hybrid. Affects validation, compute requirements, and governance profile. View in glossary Classical ML
Lifecycle Stage Current stage in the AI lifecycle, from problem identification through to monitoring, maintenance and decommissioning. View in glossary Monitoring, Maintenance and Decommissioning
Model Provenance Origin of the AI model: developed in-house, adapted from open-source, commercial/proprietary, or accessed via third-party API. View in glossary Commercial/proprietary
Compute Environment Where the AI system runs: on-premise, government cloud, commercial cloud, or edge/device. View in glossary Not documented
Sovereignty Quadrant Classification of data and compute sovereignty: I (Sovereign), II (Federated/Hybrid), III (Cloud with safeguards), or IV (Shared Innovation Zone). View in glossary Not assessed
Data Residency Where the data used by the AI system is stored: domestic, regional, or international. View in glossary Not documented
Cross-Border Transfer Whether data crosses national borders, and if so, whether documented safeguards are in place. View in glossary Not documented
Decision Criticality The rights impact of the decision the AI supports. High criticality requires HITL oversight; moderate requires HOTL; low may operate HOOTL. View in glossary Moderate
Human Oversight Type Level of human involvement: Human-in-the-Loop (active review), Human-on-the-Loop (monitoring), or Human-out-of-the-Loop (periodic audit). View in glossary HITL
Development Process Whether the AI system was developed fully in-house, through a mix of in-house and third-party, or fully by an external provider. View in glossary Mix of in-house and third-party
Highest Risk Category The most significant structural risk source identified: data, model, operational, governance, or market/sovereignty risks. View in glossary Operational and system integration risks
Risk Assessment Status Whether a formal risk assessment, informal assessment, or independent audit has been conducted for this system. Not assessed
Documented Risk Events AlgorithmWatch (2024) documented a patient case where average reimbursement times increased from 27 days (2021) to 54 days (2023) following system deployment. ÖGK attributed delays to the system's ongoing learning phase.

Impact Dimensions

Accountability, transparency and redress
Autonomy, human dignity and due process
  • Data minimisation controls
  • Human oversight protocol
CategorySensitivityCross-System LinkageAvailabilityKey Constraints
Beneficiary registries and MISPersonalLinks data across multiple systemsCurrently available and usedPatient insurance records linked with submitted reimbursement claims for eligibility verification and payment processing
Financial and payments data: programme operationsSensitiveLinks data across multiple systemsCurrently available and usedPayment information including IBAN numbers, invoice amounts, and reimbursement calculations
Unstructured and text-based contentSpecial categorySingle source (no linkage)Currently available and usedMedical invoices, forms, and correspondence from elective doctors/therapists; contains ICD-10 diagnosis codes and patient health information classified as special category data under GDPR

AlgorithmWatch (2024). Austria's Social Security Invests Over €50m in AI – Just for Bookkeeping? Berlin: AlgorithmWatch. Available at: https://algorithmwatch.org/en/austrias-social-security-invests-massively-in-ai/ (Accessed: 24 March 2026).

View source News article / media

International Social Security Association (ISSA) (2025). AI applications in social security: Building evidence and insights from the TechByte. Geneva: ISSA. Available at: https://www.issa.int/node/281122 (Accessed: 30 March 2026).

View source Report (multilateral / development partner)

IÖB – Innovationsfördernde Öffentliche Beschaffung (n.d.). Artificial Intelligence (AI) in der Kostenerstattung der SV. Vienna: IÖB. Available at: https://www.ioeb.at/erfolgreiche-projekte-detail/artificial-intelligence-ai-in-der-kostenerstattung-der-sv (Accessed: 24 March 2026).

View source Government website / press release

Österreichische Gesundheitskasse (2023). Jahresbericht 2022. Vienna: ÖGK. Available at: https://www.gesundheitskasse.at/cdscontent/load?contentid=10008.779654&version=1689762923 (Accessed: 28 November 2025).

View source Report (government / official)
Deployment Status How far the system has progressed into real-world operational use, from concept/exploration through to scaled and institutionalised. View in glossary Full Production Deployment
Year Initiated The year the AI system was first initiated or development began. 2019
Scale / Coverage The scale and geographic or population coverage of the deployment. National — approximately 5.4 million reimbursement applications per year (2017 baseline, growing at ~8% per year); 66% processed semi-automatically as of October 2022
Funding Source The source(s) of funding for the AI system development and deployment. Austrian social insurance system (IT-SV project budget of EUR 2,808,120)
Technical Partners External technology vendors, academic partners, or development partners involved. Competitive tender with 6 pre-qualified AI vendors via framework agreements; best bidder selected from 5 submissions. Vendor identity not publicly disclosed. IT-SV declined to comment on vendor details citing procurement policies.
Outcomes / Results Two-thirds (66%) of reimbursement requests processed semi-automatically by October 2022 (ÖGK Jahresbericht 2022). Intended to handle rising volumes (~8% annual growth) without additional staffing. Standardised decision-making across cases. Ongoing quantified benefits (time savings) reported to exceed ongoing costs including model training (IÖB). The 2025 ISSA TechByte appendix adds that data recognition improved considerably over the first three years of operation as multiple AI models were adapted and retrained, though it does not provide a numeric KPI. However, independent reporting (AlgorithmWatch 2024) raised doubts about whether processing times actually improved.
Challenges AlgorithmWatch investigation (2024) found evidence of increased processing times post-deployment rather than improvement. ÖGK attributed to ongoing model learning. Limited public transparency: IT-SV declined comment; Austrian freedom of information law not applicable to IT-SV until 2025.

How to Cite

DCI AI Hub (2026). 'KAI System – AI-based Semi-Automatic Reimbursement of Medical Services Fees', AI Hub AI Tracker, case AUT-001. Digital Convergence Initiative. Available at: https://socialprotectionai.org/use-case/AUT-001 [Accessed: 1 April 2026].

Change History

Created 30 Mar 2026, 08:38
by v2-import (import)