TAAP points French state accountants to risky payment requests

The Cour des comptes found that the DGFiP's TAAP software let accountants check 31.9% fewer payment requests, while errors found on AI-analyzed requests rose from 1.65% to 8.53%.

  • 2 an organization

    Accountants checked 31.9% fewer requests and found errors in 8.53% of AI-analyzed requests, against 1.65% in the year before TAAP.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Reported benefit.
Evidence

foundstudiedreportedclaimed

An independent audit report supports the benefit.

Compared with

Errors found in the year before TAAP. Measured.

Without the system
1,65 %
With it
8,53 %
The system’s part
Not isolated. The figures compare years before and after TAAP and come from the DGFiP's own assessment. The rate on AI-analyzed requests separates part of the AI's effect, but the system also applies business rules. The Cour does not say whether it checked the DGFiP's figures.
Limits
Not yet checked.
Costs and harms
Not yet checked.
Later
Not yet checked.

Download data Cite Suggest a correction

Benefit type
Improved Task Accuracy
System
Traitement automatisé de l'analyse prédictive (TAAP)
Made by
Direction générale des Finances publiques (DGFiP)
Used by
DGFiP public accountants who control state spending
Where
France (national)
Stage
In routine use
Date
Kind
Aggregate result
Broad benefit
Analyzing Data or Information, Monitoring and Controlling Resources
Collections
Scrutiny and findings
Last revised
31 October 2026

Sources

  1. Audit report, by authors with no tie to the system.

    Cour des comptes. Mieux suivre et valoriser les gains de productivité de l’État issus du numérique. Observations définitives S2024-1570, adopted 5 December 2024 and published 21 January 2025, pp. 60-61.

    Sur le seul périmètre des DP analysées par l’IA, le taux d’erreurs détectées est très supérieur à celui de l’année pré-TAAP (8,53 % contre 1,65 %, soit un facteur multiplicatif de 5,2). Le logiciel est donc un auxiliaire performant des services comptables.

Details

AI system

Workflow

Business rules flagged categories of requests with frequent errors or high financial stakes. A model trained on three years of past checks ranked the other requests and kept learning.

Follow-up

Conditions for replication

Accountants must concentrate their checks on the requests that the system flags.

Entry details

Sector

Public finance (control of state payments).

Purpose

The DGFiP built TAAP to point public accountants to the riskiest payment requests, so that they check fewer requests and find more errors.

Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.

Cite this record

Peter H. Vartanian. (2026). TAAP points French state accountants to risky payment requests (AIBD-0690). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0690.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)

BibTeX Data Field definitions