DISCO connects Cecilia AI with discovery and witness preparation

DISCO brings AI into the legal evidence process through Cecilia, alongside discovery software and tools for preparing witness testimony. Its current commercial model combines data-based platform pricing with included Cecilia access, while automated review has a separate charging basis. The useful evaluation follows evidence from discovery into a checked witness outline, rather than treating an impressive answer as a finished case analysis.

01 / ProductEvidence analysis continues beyond document search

DISCO Ediscovery provides collection-processing and review capabilities including search, filters, tagging, clustering and productions. Its Early Case Assessment and Active Review spaces work with the same database. This gives the AI discussion an operational setting: the question is how analysis helps the team understand and handle the evidence already inside its review process.

Cecilia AI supports evidence questions, document summaries, contextual definitions, timelines and deposition work. These functions address different points in a matter. A term’s meaning may need the surrounding record; a chronology needs reliable dates; a witness outline needs a clear distinction between what the witness said and what other documents suggest.

The agentic Cecilia page describes Advanced Research as a multistep approach to investigating facts and relationships, with reasoning and supporting references. That can help a user pursue a complex question, but the presence of several research steps does not establish that every relevant document was found. The resulting account remains something the legal team must examine.

Case Builder adds deposition transcripts, linked exhibits, video and collaborative preparation. It describes source-linked summaries and shared annotations. This makes the handoff to witness work particularly relevant: a useful insight should remain connected to the precise testimony or exhibit that another lawyer needs to inspect.

02 / AudienceWitness preparation creates a concrete test of usefulness

A litigation team preparing a witness on a disputed sequence of events has a clear evaluation target. It needs to find relevant documents, understand chronology and identify points requiring clarification. An investigation team can use a similar approach when planning an interview. Both benefit from a record that makes it easy to challenge a proposed factual inference.

The fit is weaker if the task is drafting a general legal memo from a small set of already understood sources. A discovery platform brings operational responsibilities around data, permissions and review. Those responsibilities are worthwhile when the evidence collection is a substantial part of the work, but they should not be ignored in a comparison with a simpler assistant.

OpenText is a useful comparison for broader information and discovery requirements. Harvey provides a contrast for teams considering a wider legal AI workspace. Compare the work that must happen after an answer appears: opening evidence, preparing testimony, handling corrections and maintaining the matter record.

03 / WorkflowA proposed witness outline separates testimony from inference

Choose an authorised historical dispute with a known chronology and a completed deposition. Include an email chain, a contemporaneous attachment and testimony that describes the same event differently. This is a proposed pilot, not a claim of testing. The supervising lawyer should prepare an independent list of the important discrepancies before the AI-assisted exercise begins.

First establish the evidence scope. Record which custodians, periods and document types are included, and check that the key exhibits have been processed. Ask a narrow factual question about a decision or communication. Keep the initial keyword or filtered search as a comparison point so the team can see whether the AI uncovers useful additional material or simply restates familiar documents.

Follow the source references for the answer and separate three categories: what a document directly records, what a witness testified and what the reviewer infers from the combination. Those categories should remain visible in the proposed outline. A message sent before a meeting may establish its timing without proving that every participant read or understood it.

Use the case-preparation tools to build a short witness section around a specific discrepancy. Place the relevant testimony beside the supporting exhibit and draft a neutral question that seeks clarification. Avoid asking the system to turn every inconsistency into a conclusion about credibility. A changed recollection can have several explanations that the record does not yet resolve.

Give the outline to another lawyer who did not conduct the initial search. Ask them to locate the source, confirm the quoted passage and identify any missing context. Then introduce a corrected date or additional exhibit and update the outline. This tests whether the team can maintain an accurate working document as the factual record changes.

Measure important omissions, unsupported inferences, time to locate the source and effort required to revise the outline. An answer that saves search time but creates lengthy verification work may have limited value. The goal is a more efficient route to well-prepared questioning, with the lawyer retaining responsibility for how the evidence is used.

04 / PricingPlatform access and automated review use different billing units

This distinction prevents a common budgeting mistake. Included access to an evidence assistant does not make every form of automated document review part of the same price. Ask the account team to show a complete scenario containing initial data, later evidence, user work and any Auto Review pass. The contract should make clear which meter applies at each stage.

The Auto Review page describes both client-managed and DISCO-managed approaches, with Tag Tuner helping refine review criteria. That choice affects the staffing plan as well as the invoice. A team running the process itself needs a subject-matter expert and an agreed validation method; a managed engagement still needs someone authorised to settle the review instructions.

05 / DistinctionCase preparation gives AI findings a practical destination

The most interesting connection is between discovering a fact and using it in witness preparation. A summary is only an intermediate product. Its value grows when a colleague can inspect the source, add an annotation and incorporate the point into an outline without losing the distinction between testimony and interpretation. That is a practical workflow benefit a small pilot can test.

The Cecilia product page identifies particular capabilities as available in the UK and Europe. Availability should still be checked for the precise feature and proposed account, especially when a demonstration spans several products. Treat a region label as useful product information, not a substitute for the contractual data-processing and hosting arrangements.

Automated review offers another possible extension, but it should be evaluated as a separate decision process. The team may use an assistant to explore an issue and then develop formal coding criteria. Moving from exploratory questions to classification changes the consequence of error, so the sampling, escalation and acceptance rules need to change with it.

06 / QuestionsCheck whether the evidence supports the strength of the conclusion

The first question concerns source coverage. Ask the team to inspect a result that appears convincing but depends on material outside the selected scope. A timeline can omit a decisive event; a summary can emphasise one passage and miss a qualification. Evaluate those failure cases explicitly instead of treating citations as proof that the account is complete.

The Auto Review description discusses explanations, protocol refinement and validation metrics. Use those to ask for a matter-specific evaluation with known examples and independent checking. Vendor throughput or superiority claims are not evidence of performance on the organisation’s collection, and this article does not reproduce them as established results.

The next question is operational: who can correct a result, who approves the final witness material and how a later change reaches everyone using it? Demonstrate that process with the permissions intended for the actual team. This Blueprint reviewed public sources without operating a customer account; extraction quality, research completeness, regional access and negotiated commercial terms remain open evaluation questions.

07 / DecisionChoose a handoff where source access matters

DISCO is worth evaluating when a substantial evidence collection must turn into practical case preparation. Start with a witness topic that has a known factual record and a second reviewer who can challenge the outline. Scale when source access, correction work and the separate cost of automated review are all understandable.