Irish court backs predictive coding for 680,809 bank documents

In March 2015 the High Court of Ireland approved predictive coding for 680,809 documents over the defendants' objection. It found that the method was at least as accurate as manual review.

  • Not rated an organization

    The review had not run when the court ruled, and the saving rests on sworn estimates that the parties' experts disputed.

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

Status
Intended benefit.
Evidence

foundstudiedreportedclaimed

An independent court document supports the benefit.

Compared with

Manual (linear) review of the 680,809 documents. Assumed without a measurement.

Without the system
9 months at a cost of €2 m by the bank's solicitor's estimate, or 113 days at €220,000 by the defendants' expert's
With it
less than 10% of the 680,809 documents would need to be manually reviewed
The system’s part
Not isolated. The court approved the review before it ran. The accuracy finding rests on studies that the bank's expert exhibited, and the saving rests on sworn estimates. The judgment describes a model that learns from an expert's coding of training sets.
Limits
Not yet checked.
Costs and harms
Not yet checked.
Later
Not yet checked.

Download data Cite Suggest a correction

Benefit type
Improved Organizational Efficiency
System
Technology assisted review using predictive coding
Made by
Not named in the judgment (the bank's expert took his target f-measure from Symantec, the provider of the Clearwell system)
Used by
Irish Bank Resolution Corporation and its solicitors, in discovery
Where
Dublin, Ireland
Stage
Proposed
Date
, when the source was published
Kind
Benefit created
Broad benefit
Processing Information, Getting Information
Collections
Scrutiny and findings
Last revised
31 October 2026

Sources

  1. Court document, by authors with no tie to the system.

    High Court of Ireland (Fullam J.). Irish Bank Resolution Corporation Ltd & ors v Quinn & ors [2015] IEHC 175, judgment of 3 March 2015. BAILII copy, read through the Internet Archive.

    The evidence establishes, that in discovery of large data sets, technology assisted review using predictive coding is at least as accurate as, and, probably more accurate than, the manual or linear method in identifying relevant documents.

Details

AI system

Workflow

A keyword search and the removal of duplicates narrowed 1.7 million documents to 680,809. An expert trains the model on samples until it predicts relevance over consecutive samples. Documents below the cut-off are sampled to find relevant documents that the model missed.

Follow-up

Conditions for replication

The process is transparent, with notice of the training threshold and a chance to challenge it. The data set is large, which the bank's expert put at 500,000 documents or more. Senior lawyers review the documents above the cut-off.

Entry details

Sector

Legal services / e-discovery.

Purpose

A vendor that the judgment does not name built the predictive coding software to give each document a relevance score, with algorithms that learn from an expert reviewer's coding.

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). Irish court backs predictive coding for 680,809 bank documents (AIBD-0829). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0829.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)

BibTeX Data Field definitions