Two ways into our practice.
Below: the anonymised record of recent matters and the credentials that underwrite them, alongside our authoritative guide to how eDisclosure actually runs in the English courts.
- Matters in 2025
- 62
- Regulator deadlines met
- 100%
- Largest review
- 14M docs
- Median turnaround
- 11 days
Anonymised recent matters
Selected case notes from recent UK regulatory and contentious work, alongside our accreditations, certifications and professional memberships.
- FCA, CMA and public inquiry work
- IP, employment and commercial disputes
- ISO 27001 aligned, Cyber Essentials Plus
eDisclosure in UK Courts
What PD 57AD and CPR Part 31 actually require of corporate defendants, and how E-Discovery UK runs each stage of a disclosure exercise from legal hold to production.
- PD 57AD and the Disclosure Review Document
- The six stages, identify to disclose
- IP, breach of contract and commercial disputes
Analytics and machine learning, applied where they hold up to scrutiny.
Every technique below is used in a defensible, documented workflow, with the decisions recorded so they can be explained to the other side, to the court or to a regulator.
Technology-assisted review (TAR)
Defensible, accelerated prioritisation of documents, with the training and validation record kept so the approach can be defended in correspondence and at a CMC.
Machine learning classification
Models trained on the matter itself to surface relevance, likely privilege and the key themes running through a collection, so counsel see the substance early.
Email threading and near-duplicate detection
Advanced threading and near-duplicate grouping strip out repetition, cutting review volumes and cost without losing any unique wording.
Automated pattern recognition
Anomalies, communication relationships and hidden risks are identified across custodians and date ranges, pointing the investigation at what matters.
AI-supported early case assessment
AI review workflows applied from the outset give clarity on scope, exposure and cost before the full review begins.
1.9 million documents cut to 42,000 for review
A trade secrets claim against three departing engineers. Preservation covered laptops, two file servers, Microsoft 365 mailboxes, Teams and a GitHub organisation, across nine custodians and a four year date range.
- Stage 011,912,400Collected
Documents and messages forensically collected and hashed before any filtering.
- Stage 02608,700After de-duplication and threading
Exact and near-duplicates grouped, email threads reduced to inclusive messages only.
- Stage 03214,300After date, custodian and keyword filters
Agreed scope applied and recorded in the Disclosure Review Document.
- Stage 0442,100Prioritised by TAR for human review
Continuous active learning trained on 3,100 coded seed documents, validated by sampling.
What the analytics changed
- Review effort fell from an estimated 9,600 hours to roughly 1,400, with first pass completed in five weeks rather than five months.
- Elusion testing on the unreviewed set put estimated recall at 94 per cent, with the sampling record disclosed to the other side.
- Pattern analysis surfaced 214 files copied to a personal cloud account in the fortnight before resignation, which became the core of the injunction application.
- The TAR protocol, training log and validation statistics were agreed at the CMC and were not challenged at trial.
Figures anonymised and rounded. Custodian names, party names and document content withheld.
Bring us in early.
Defensibility is built, not retrofitted.
Whether you are responding to a regulator, preparing for disclosure, or scoping an internal investigation, start the chain of custody with a short, confidential conversation.