TCU's ALICE tender-analysis robot captures R$426m in 2021

AI system

System: ALICE, expanded as 'Análise de Licitações e Editais' in 2017 and 2018 press reports and named 'Sistema de Análise de Licitações e Contratos (Alice)' in the TCU's July 2020 release, with companion tools SOFIA and MONICA; evolved into ALICE 360 by 2026
Model: Not assessed
Release: Not assessed
Application developer: Tribunal de Contas da União (TCU), in-house
Role: Not assessed
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: Daily text-mining of the Diário Oficial and Comprasnet notices by Alice with e-mailed alerts to auditors. Low-cost follow-up by auditors (phone call, e-mail, ofício) or formal acórdão when needed. Agencies withdrawing, annulling or redoing flawed notices after contact.
Human oversight: Not assessed
Relation to AI: The TCU attributes the 2021 figure to control actions supported by Alice under its benefit-measurement method. It says “o segundo de fato reflete o benefício calculado, que resultou da ação de controle e foi medido”. No independent evaluation or counterfactual was found. The 2017 DNIT/BR-153 example rests on a trade-press interview with a TCU auditor. The TCU acórdão was not available. Developer and deployer are the same public body.
AI contribution: Not assessed

Benefits

Outcome identifiers: Not assessed
Benefit: captured R$426m in 2021 (R$194m in 2020)
Category: Reduced Risk
Status: Not assessed
Beneficiary groups: People. Organizations. Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: captured R$426m in 2021 (R$194m in 2020)
Comparison: Not assessed
Results: captured R$426m in 2021 (R$194m in 2020)
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: Tribunal de Contas da União (TCU), in-house. Operator: Tribunal de Contas da União (TCU); also shared with DNIT and the Ministry of Infrastructure under a 2020 agreement.
Jurisdictions: Benefit: Brasília, Brazil. Developer: Brazil. Operator: Brazil.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Structured publication of notices on federal systems. Auditor capacity to act on the daily list. A benefit-measurement method that records actual, not potential, benefits.
Replication attempts: 3 follow-up entries. Benefit recurred: 2. Outcome unresolved: 1.
Later findings: 6 May 2026. A TCU plenary ruling of 6 May 2026 (Acórdão 1.145/2026) records that the tool, now called Alice 360, has moved to intensive use of AI, with LLM-based query screens; no outcome figures are given.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0083-source-0001–0006
Citations: Numbered references below.
Language: Not assessed
Source relationships: Developer, Independent
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: People. Organizations. Society.
Depth: Not assessed
Reasons: People: Aggregate program figure with no discrete attributable event. Organizations: The R$426 million is an annual total of control benefits credited to the Alice system, not a single verified event. Society: Aggregate. The TCU's own benefit-measurement methodology, not an independent evaluation.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently reported

Entry details

Kind

Not assessed

Related entries

IRS audit-selection algorithm, an audit body's targeting tool like TCU's ALICE, skewed audits toward Black taxpayers, harming the taxpayers the audit function serves.

Field

Public administration: external audit and procurement control. Supreme audit institution.

Purpose

Daily automated screening of federal procurement notices to direct external-control auditors to likely irregularities

Dates

Date basis unassessed: 2021 (precision unassessed). Updated: 21 October 2026.

Stage

Not assessed

Version

SHA-256: 4201a828e804…

References

Other identifiers: AI Incident Database: 461.
Category mappings: Not assessed
Brief: Not assessed
Revisions: 9 revisions. Latest: 21 October 2026.

Sources

News report Independent author

Inforchannel (2017). TCU cria robô para varredura diária de irregularidades em editais, 27 Sep 2017 (quoting TCU auditor Valéria Nakano).
News report Independent author

Gomes, H. S. (2018). Como as robôs Alice, Sofia e Monica ajudam o TCU a caçar irregularidades em licitações. G1, 20 Mar 2018, republished by Ministério Público de Contas do Espírito Santo.
Press release Developer

Tribunal de Contas da União (2022). Atuação do TCU gera recorde de benefícios financeiros em 2021. TCU press office, 4 Mar 2022.
Press release Developer

Tribunal de Contas da União (2020). TCU, Ministério da Infraestrutura e DNIT firmam acordo para aperfeiçoar fiscalização e prevenir fraudes em licitações. TCU press office, 10 Jul 2020.
Press release Developer

Tribunal de Contas da União (2024). Uso de inteligência artificial aprimora processos internos no Tribunal de Contas da União. TCU press office, 22 Feb 2024.
Other Independent author

Zênite (2026). TCU aponta irregularidades em contratações públicas e destaca avanço do Alice 360 com IA (summary of TCU Acórdão 1.145/2026-Plenário, rel. Min. Jorge Oliveira, 6 May 2026). Blog da Zênite, 14 May 2026.