AI System for Preparing Client Document Request Lists

The problem

In day-to-day operations, document requests may be sent in several separate messages. When the matter type, lawyer's checklist, document index and communication preference are spread across different screens, documents or team records, the issue may not be noticed in time.

This fragmented data delays preparation of the document request draft and forces staff to search for supporting records again by hand. Before deciding, the matter lawyer must complete missing fields and verify each inconsistency individually.

A common exception is that the same content may have been uploaded under a different name or in a different format. If this context is not checked, an incomplete or incorrect automation could request unnecessary data or overlook an important document, so the output cannot be acted on directly.

What we set out to improve

• Number of document items changed by the lawyer
• Share of outputs corrected or rejected by the matter lawyer
• Time between the source record and the review output

How the system works

1. 01 Define the review scope and the authorised matter lawyer role
2. 02 Collect the inputs: matter type, lawyer's checklist, document index and communication preference
3. 03 Check the source identifier, date, version and data freshness
4. 04 Match documents against the checklist
5. 05 Turn missing items into a request draft
6. 06 Check the exception: the same content may have been uploaded under a different name or in a different format
7. 07 Place low-confidence or conflicting records in a separate human review queue
8. 08 Write the source identifier, rule, output and timestamp to the audit trail
9. 09 Have the matter lawyer approve, correct, defer or reject the output

Methods we used

• Matching against a lawyer-approved checklist and document types
• Source, format, version and required-field checks for the matter type, lawyer's checklist, document index and communication preference
• A missing-data alert instead of an estimate when inputs are absent or conflicting
• Separate handling of exceptions and the main rule
• Joint retention of the source identifier, rule result, document request draft and human decision

Where people stay involved

Data and security

Access to the matter type, lawyer's checklist, document index and communication preference is limited to the minimum permissions needed for the task. Identifiers and sensitive fields are masked where possible, and only the necessary section is sent to the model rather than the full raw file. Access, output and the matter lawyer's decision are logged; production use does not begin until the organisation approves the retention period and data location.

Who this suits

Frequently asked questions