Check email recipients and attachments

Practice relevance

Cross-practice idea · A general operational method. Review the stated tool and jurisdiction limits before use.

The problem

Compare a proposed send with an approved list of recipient roles and document IDs.

Why it matters: A wrong recipient or attachment can expose information the firm did not intend to share. Checking an approved send manifest adds a review step, while the sender still verifies real addresses and file versions in the mail client.

Sources & evidence

Proposed / not tested. The public discussion supports the problem; the method source supports only the documented features. No outcome from this adapted recipe has been verified.

What’s been verified

The linked sources support the complaint and documented method components. We have not run this recipe or verified its results. The example below is synthetic, not a customer result.

What to measure: During the pilot, count findings confirmed by the reviewer, false alarms, and missing-input cases. Do not infer time or money saved.

Pain evidence: Worst mistake at work discussion ↗Worst mistake at work discussion, comments dated 2026-10-09, recent. A commenter recounts a colleague sending an internal strategy report to opposing counsel rather than the client; other comments describe accidental client inclusion. These are anecdotes in comments, not verified incident reports.
Method: Outlook MailTips ↗Outlook MailTips provides recipient-related warnings in supported Exchange accounts. It does not validate legal permission or fully inspect every attachment; the manifest check below is a separate proposed control.

Source checked: October 15, 2026. Pain reports are self-reported and identities are unverified. Method documentation supports components, not a tested end-to-end workflow. We wrote these prompts. The adapted workflow hasn’t been tested.

Suggested approach

Start when: External message is ready for review

compare recipient roles and attachment IDs against an approved distribution manifest
Keep row or page references and mark missing or ambiguous evidence. Do not apply the suggested changes.
The responsible staff member checks the draft against the source records and approves any separate action.

Limits and review

The assistant drafts or reports only. Check source records and review the output before a human performs any separately approved send or change. No automatic legal decision, money movement, account connection or publication is included.

Who reviews it: Sender verifies exact real addresses, document versions and authorization inside the mail client. Test an opposing-counsel recipient substituted for client and a similarly named attachment from another matter.

What you need

Your firm-approved AI assistant · prepared files

A selected high-risk unsent message, sender-approved exact distribution/attachment manifest and secure access to the actual recipient addresses and file versions.
Outlook MailTips is supplemental and requires a supported Exchange account; unavailable MailTips cannot be treated as an all-clear.
Existing firm-approved Acrobat Studio, AI Assistant add-on, or another plan including Acrobat AI; online services/cloud storage, language and region must be permitted. No purchase, login or connection is part of this verification. The manual route works without AI.
Preserve original evidence unchanged in restricted storage. Use synthetic material first. Prepare a separate, explicitly approved minimized AI input; aliases alone do not anonymize narrative or attachments. Do not upload raw originals or an alias key merely because another tab/file is sanitized.
Use synthetic inputs first. Keep real data in firm-approved accounts and processing, with appropriate access, confidentiality and retention controls. Aliases alone do not anonymize records. Never paste credentials, API keys or unnecessary client details.

How to use

Documentation-verified component path; proposed adaptation, not tested

Before you start

Manually preparing metadata for every ordinary email may cost more than a careful checklist. Use selected consequential sends or an already maintained approved manifest.
No universal draft-recipient/attachment CSV export is verified. Role aliases are assigned by staff; actual address expansion, link access and attachment contents must be checked inside the secure mail client.

Check access and setup

Actual account access, source versions/completeness, privacy approval and reviewer sign-off were not observed.

Inspect the actual unsent message
Outlook desktop and sender-approved manifest · Open the saved draft; inspect To, Cc, Bcc and every attachment/link; read available MailTips
Read the actual expanded recipient addresses in all recipient fields, including reply-all additions or groups. If a group cannot be completely resolved, block it. Open each proposed attachment from the draft securely and compare its exact version/matter to the approved manifest; inspect link targets/access through the approved system. Manually record recipient aliases/roles and attachment IDs without copying strategy content.
Expected output
Observed proposed-send list and separate approved-send list, with exact secure source references.
Next step
Create a minimized role/attachment-ID comparison input; keep real addresses, sensitive filenames and document content local.
Check before continuing
MailTips warns about some recipient situations; it does not establish permission or attachment correctness. Changes to the draft invalidate the snapshot.
Step documentation 1 ↗Step documentation 2 ↗
Load only the approved AI input
Acrobat PDF Spaces · Acrobat homepage > PDF Spaces > Create a PDF Space > upload files or Paste copied text > Add to PDF Space
Add the minimized approved-send, proposed-send and attachment-version/shareability records. Use clear source/version labels and verify the Files panel. Do not add originals that have not separately been cleared for Adobe processing.
Expected output
Visible approved source set in the PDF Space; no account integration is implied.
Next step
Use the AI Assistant message field in this same space for the next prompt.
Check before continuing
Confirm every named source is actually loaded and readable. PDF Spaces supports up to 100 files, 100 MB and 600 pages per file; unsupported/password-protected/handwritten material blocks affected extraction.
Step documentation 1 ↗Step documentation 2 ↗Step documentation 3 ↗
Run the bounded prompt and inspect citations
Acrobat AI Assistant · The PDF Space AI Assistant message field
Paste the prompt below. Open each source-number citation, check it, and use Copy to transfer the answer into a new local Word document or text file. Correct errors using the originals before reliance.
Acrobat AI Assistant promptCompare only the approved and proposed send manifests. Return message_id,matter_alias,recipient_role_or_attachment_id,exception,evidence_refs,required_review. Flag unexpected recipients, reply-all expansion, unknown group membership, internal-only attachments, wrong-matter IDs, wrong versions and missing approvals. Never infer consent or an alias-to-person mapping. Incomplete metadata means manual verification required. Do not inspect or repeat strategy content, change permissions, alter the draft or send anything.
Expected output
Recipient/attachment mismatches and approval blockers.
Next step
Save the checked draft as send-check-review.docx; carry source IDs, versions and reviewer corrections into the next review step.
Check before continuing
Human verifies the real exact addresses and actual attached files again immediately before any send; alias agreement alone is insufficient.
Step documentation 1 ↗
Human review and final handoff
Final human review
Responsible staff reviewer and lawyer/process owner · Local reviewed report beside the frozen source set
Sender checks the final draft’s exact To/Cc/Bcc addresses, recipient-group expansion, attachment bytes/version and any linked-file permissions against the approval. Resolve all blockers and record disposition.
Expected output
Version-specific send checklist.
Next step
Sending remains a separate authorized human action; this is not a confidentiality guarantee.
Check before continuing
All unresolved or unreadable evidence stays open. No inferred approval or external action follows from AI output.
Step documentation 1 ↗

What you’ll get

send-review.csv: message_id, check, approved_value, proposed_value, attachment_id, finding, human_check_required. No live changes.

Inputs and a sample test Synthetic example · not a measured result

Input files

approved-send.csv: message_id, matter_alias, approved_recipient_role_ids, approved_attachment_ids. proposed-send.csv: message_id, matter_alias, proposed_recipient_role_ids, proposed_attachment_ids. attachments.csv: attachment_id, matter_alias, shareability_label, approved_version_id.

Synthetic example

MSG1: CLIENT1 approved, OPPOSING_COUNSEL1 proposed
MSG2: internal-only attachment proposed
MSG3: approved IDs, exact real addresses not yet checked

Expected output

MSG1: recipient mismatch
MSG2: attachment blocker
MSG3: sender still verifies real addresses and file versions

Test a failure

Sender verifies exact real addresses, document versions and authorization inside the mail client. Test an opposing-counsel recipient substituted for client and a similarly named attachment from another matter.

Try with the sample files

Follow the tool steps above and use the inputs named in each step. Where provided, native-* files are blank scaffolds and NATIVE-INPUT-MAPPING explains their fields. Original examples and the optional full review prompt use a separate prepared-input schema. Keep them separate unless the mapping explicitly connects them. Downloading a kit does not connect accounts or prove the workflow works.

Download inputs and checklist 10 files

Instructions, blank templates and separate synthetic examples; these files do not install an automation. Blank templates still need verified source data.

Download starter kit All inputs, synthetic examples and review notes in one ZIP. No automation is installed.
Start here: get the sources and prepare the review Documentation-checked acquisition, explicit manual mapping, exact prompts and human review limits
Setup and review checklist Required inputs, file use, review steps and limits
Approved send blank CSV Blank input headers for approved-send.csv
Approved send synthetic example Synthetic test records for approved-send.csv
Proposed send blank CSV Blank input headers for proposed-send.csv
Proposed send synthetic example Synthetic test records for proposed-send.csv
Attachments blank CSV Blank input headers for attachments.csv
Attachments synthetic example Synthetic test records for attachments.csv
Synthetic example expected results Compare a pilot result with the planted exceptions

Optional full review prompt for prepared inputs

Prepared-input review prompt with Your firm-approved AI assistant · prepared files

Compare this message's deidentified recipient roles and attachment IDs with the approved send manifest. Flag unexpected recipients, reply-all expansion, internal-only attachments, wrong-matter IDs and missing approval. Do not inspect or repeat strategy content, infer consent, change permissions or send. If metadata is incomplete, mark ‘manual verification required.’

INPUTS
approved-send.csv: message_id, matter_alias, approved_recipient_role_ids, approved_attachment_ids. proposed-send.csv: message_id, matter_alias, proposed_recipient_role_ids, proposed_attachment_ids. attachments.csv: attachment_id, matter_alias, shareability_label, approved_version_id.

OUTPUT
send-review.csv: message_id, check, approved_value, proposed_value, attachment_id, finding, human_check_required

Use only the supplied synthetic or firm-approved inputs. Cite source row or page IDs for each finding. Keep unknowns unresolved. Return a draft for the responsible staff reviewer. Do not send messages, connect accounts, change records, move money, or make legal decisions. Keep source files unchanged. If an input or permission is missing, list what is needed and stop that check.

Proposed / not tested · editorial prompt. This prompt uses the original prepared-input schema shown in the sample test. Step prompts may use different tables. Follow NATIVE-INPUT-MAPPING where supplied; do not mix the original examples with native-step inputs without mapping and checking every field.

Read original method Read pain evidence