There is no reliable page-per-hour rule for every deposition. The time depends on transcript length, density, the reviewer’s familiarity with the matter, the required summary format, and the intended use.

A short factual deposition may be straightforward. A long corporate or expert examination with exhibits, technical language, and issue-specific analysis can take substantially longer.

Treat any unqualified speed number as a claim that needs context. A processing timer does not measure transcript preparation, issue selection, citation verification, corrections, export cleanup, or professional review—the steps that determine when a draft becomes usable.

What affects the time

Length matters, but complexity often matters more. Repetitive testimony may be faster to compress than a shorter transcript with changing terminology, disputed exhibits, and frequent qualifications.

The output format changes the workload. An exhaustive page-line digest is different from a selective issue-based summary. The first emphasizes sequential coverage; the second emphasizes access to material topics and requires careful scoping so the reader does not mistake selection for completeness.

  • Number and complexity of disputed issues
  • Need for chronology or exhibit cross-references
  • Quality and searchability of the PDF
  • Depth of page-line citation required
  • Whether the reviewer knows the matter already
  • Level of professional verification

Manual review

A careful manual summary includes reading, issue coding, drafting, citation capture, and a verification pass. For a long transcript, that can occupy a meaningful portion of a workday or more.

The time is not wasted if the reviewer is also learning the case. The question is whether every part of that first pass requires the same level of professional attention.

Do not use a generic estimate as a staffing promise. Run a baseline on representative transcripts and record the reviewer’s experience, transcript length, format, issue complexity, and review standard. A fair baseline should include the same deliverable quality expected from the assisted workflow.

AI-assisted review

AI can reduce some of the initial organizing and drafting work. The legal team still needs time to review the result and verify material citations.

The honest comparison is therefore not “AI time” against “human time.” It is the complete AI-assisted workflow against the complete manual workflow.

The ABA’s Formal Opinion 512 discusses competence, confidentiality, supervision, candor, and reasonable fees among the duties implicated by generative-AI use. Those considerations do not disappear because a draft arrives quickly.

Measure six stages separately

A stage-by-stage time log shows where the workflow helps and where it creates rework. Use the same clock definitions for manual and assisted tests, and pause the timer only under a rule established in advance.

  • 1. Prepare and validate the transcript file
  • 2. Read enough of the matter context to define scope
  • 3. Generate or draft the first version
  • 4. Verify material findings and citations
  • 5. Correct omissions, wording, chronology, and formatting
  • 6. Complete professional review and export the deliverable

Where time savings may come from

Potential time savings come from reducing repetitive organization: collecting related passages, drafting a chronology, grouping testimony by issue, and formatting a consistent first version.

Professional time can then be directed toward context, strategy, credibility, and the significance of the testimony.

Savings are likely to vary by transcript. A searchable record with clear speakers and stable page-line numbering is a different input from an image-only scan, damaged text layer, dense expert examination, or transcript with extensive exhibits. Report the distribution of results, not just the fastest run.

Use a small benchmark table

For each test, record transcript pages, file type, summary format, manual baseline, processing time, verification time, correction time, total assisted time, material errors, citation errors, and reviewer. Separate provider processing time from total reviewed-work-product time.

Repeat the same controlled transcript after material model or workflow changes. A result from one configuration does not prove future performance, and a synthetic test does not establish performance on every client record.

Use accuracy as the constraint

A faster summary is not useful if the team cannot trust or verify it. Measure time saved only after including citation checks and corrections. For important uses, accuracy and reviewability remain the controlling requirements.

LegalBrief does not promise a fixed completion time. The homepage phrase “in minutes, not hours” describes the intended benefit of an assisted first draft, not a service-level guarantee. Actual processing and review time varies with transcript length, format, and complexity.