Dictating a Privilege Log in Document Order, Field by Field

The privilege log is where a discovery production pays its real tax: hundreds of fields, hundreds of pages, and one wrong privilege basis in any of them is enough to open the whole log. The failure mode is not the big mistake; it is the small one repeated, the basis phrase that the paralegal attached to a document it does not actually cover. Dictating a privilege log in document order with a fixed field sequence is the method that makes the log right the first time, because the information is captured in the order the documents sit on the desk rather than being reconstructed later from a spreadsheet of notes.

The log's difficulty is structural, not intellectual. Each entry needs the custodian, the date, the author, the recipients, the subject, the privilege basis, and the description, and the fields are so similar from line to line that the eye stops checking after the fiftieth row. The eye drift is what the spreadsheet imports: by row three hundred, the basis column holds a value the screener would not have chosen at row ten. The voice method defeats the drift by keeping the entry attached to the document, since the document is what the screening decision was made against, not the row.

The tax is paid twice when the log is wrong. First in the rework, where the whole log gets a review pass to catch the basis errors; second in the fight, where the adversary challenges a sample, finds one inconsistency, and asks for the log to be re-pulled under a different methodology. The court's patience with privilege logs is short because it has seen the same inflation pump. The working standard is verifiable accuracy, and verifiable accuracy starts with a field order spoken the same way for every document, so the log's consistency is built in before the export is ever run.

Fix the field order and use it for every entry: custodian, date, author, recipients, subject, privilege basis, description, in one spoken line per document. The order is chosen so the fields fall the way the screener actually reads a document, and it is the same order the court expects in the log's column layout, which means the export needs no re-ordering later. The spoken line for one document is short: custodian, Rodriguez. Date, March 3. Author, Dana Wu. Recipients, Alvarez and Cohn. Subject, draft settlement analysis. Basis, attorney client communication. The entire entry is one breath, and the breath keeps the entry attached to the document.

The date field follows a strict spoken format, since dates are the most commonly challenged log field: say the day, the month, and the year as separate words, and never abbreviate. The author and recipients fields are where the description of the log can veer toward overstatement: dictate the author as the person named on the document, and the recipients as they appear in the copy line, rather than as the sender's team. The temptation to widen the recipients is the same temptation that widens the basis, and it is the one the spot check is designed to catch. The field order enforces the discipline by making the recipients a distinct spoken unit.

The description field is the last and the one that carries the most argument. The description says what the document is, not what it proves: analysis of the settlement range, not discussion of plaintiff's weakness. Dictate the description from the document's own title or first line, and keep it to a phrase the court could check against the attachment. A description that argues is a description that hands the adversary a reason to look. The description is also where the voice method pays, because the drafter sees the document and speaks its content immediately, rather than describing it from memory of a note taken three days ago.

The basis phrases are the log's vocabulary, and the vocabulary is standardized before the log is started. The core phrase is attorney-client communication, and its variants are the ones the courts recognize: attorney-client communication and work product, attorney work product prepared in anticipation of litigation, and the rarer opinion work product that gets a separate marker. The team writes the approved list on one page, and the basis spoken in the dictation must match a phrase on the page. No synonyms, no on-the-fly descriptions, because the adversary's challenge reads the bases the way a contract reads definitions. Keep the predicate for every entry in one vocabulary, so that the log reads like a ship's manifest rather than a pile of lawyer notes.

The spoken basis discipline has a rule of one: one basis per document, chosen from the list, and if a document genuinely carries two bases, the entry says the two approved phrases in the agreed order. The rule matters because a log with nineteen different descriptions of the same privilege is a log that looks inflated, and the inflation argument is the cheapest one to make. It also matters because the noun phrase attorney-client communication is a phrase the voice workflow should treat as fixed text: say the trigger, let the software insert the exact wording, and never spell it from memory in each row.

The variants carry conditions. Attorney-client communication requires the communication to be between the attorney and the client or their agents, and the dictated entry must reflect that: the author is the attorney, the recipient is the client, or the recipient is the client's agent. The work product basis requires the anticipation of litigation, which is why the date field and the subject matter matter more here; a subject that is routine business drafting under a work product basis is the entry the spot check will flag. When the standards are attached to the phrases, the basis column becomes a prompt instead of a guess.

The spot check is the log's quality gate, and it targets the ten files that decide the whole log rather than a random sample. The ten are chosen on purpose: the two most senior partners' files, because their privilege claims are the widest; the files that passed through outside counsel, because the agency analysis is hardest there; the files with the largest attachment counts, because the attachment is where the basis drifts; and any file whose date falls in the dispute's hottest month, because the anticipation of litigation analysis is most contested there. Ten files is enough to reveal the pattern, and the pattern is what the adversary will also find.

The check is verification, not review: open each file, read the entry the log assigned, and confirm the author, the recipients, and the basis against the document itself. A log entry that passes for nine files and fails on the tenth tells the team where the methodology slipped, and the slip is usually the field order drifting out of sequence rather than a decision that changed. Dictate the correction into the log in the same line format, and the correction becomes part of the log's edit history, which is the record that answers the methodology question later. Those who review the log after the dispute should be able to reconstruct every change, and a dictated workflow leaves a breadcrumb trail that manual typing often erases.

The spot check earns its name because it also tests the checker. The person who screens and the person who checks should not be the same person, since the screener who spent four thousand rows on the same basis will defend the same basis in the check. The checker reads the documents cold and compares, and the discrepancy that results is the one the court would have found. The check is run after the export and before the production, and the result gets a one-line summary in the matter notes: ten checked, the pattern found, the fix applied, and the answer to the question the log may still face.

The export converts the voice log into the court's format, and the conversion is where a clean log becomes a clean filing. Define the column mapping once: the spoken fields map to the log's columns in the order they were spoken, so the custodian, date, author, recipients, subject, basis, and description occupy the columns the standing order names. The export script runs the mapping, and a table of the field order is attached to the matter notes so the next export is a repetition, not a re-derivation. The mapping is also the place where the basis list gets enforced, since a phrase that does not match the approved list flags in the export.

The column check runs after the export and before the certificate. Spot-check the columns across the log's start, middle, and end: does the date column hold dates in one format, does the basis column hold only approved phrases, does any row have an empty custodian, does any description exceed the court's page or line limits. The check is simple enough for the paralegal to run, and it catches the structural failures that read as carelessness: the swapped column, the imported header, the row that lost its date. Each finding is fixed in the source log and re-exported, since the source is the log of record.

The waiver question closes the process. The log's attention to privilege bases is what defends the privilege, but the log itself, once produced, becomes the record of what was claimed, and the claim is evaluated by what was not claimed, too. The entries where the basis was doubtful should carry the same discipline as the confident ones, since a log that claims everything and verifies nothing is a log that invites the motion to compel that the next case pays for. The team's habit is the waiver audit: the excluded documents get the same field order, the same basis list, and the same spot check, and the exclusion is documented, because the excluded list is the part of the log that survives the redaction fight.