Billing Narratives That Survive Client Review

A polished billing narrative can still fail client review. It may describe the wrong matter, combine tasks the client requires to be separate, use a prohibited term, contradict its billing code, or include a purpose that sounds plausible but was never part of the work.

The opposite can fail too. A perfectly accurate entry such as Review documents may tell the reviewer too little to assess what was done and why the charge belongs on the matter.

The goal is not maximum detail or a universally “client-ready” sentence. It is the smallest accurate explanation that connects the work performed to the correct matter and the rules governing that matter. Anything the writer or software cannot safely know should be confirmed by the timekeeper, not filled in with persuasive-sounding language.

This article offers operational guidance, not legal advice. Billing requirements vary by client, matter, engagement, and jurisdiction. Use the current agreement and outside counsel guidelines for the matter.

There is no universal perfect narrative

Public billing guidelines show why a stock phrasebook is unreliable. The FDIC Outside Counsel Deskbook asks for enough information to identify the task, its purpose, and the relevant document or issue, and it expects distinct services to be itemized. By contrast, Zscaler’s outside counsel billing guidelines combine narrative requirements with their own coding, approval, block-billing, and compensability rules.

Those are examples of two clients’ requirements, not industry-wide templates. The Association of Corporate Counsel describes outside counsel guidelines as a way for a legal department to express its particular operating expectations. A narrative that passes one client’s review may need different detail, terminology, or task separation for another.

Good narrative writing therefore has two layers:

1. Describe the work truthfully. State what occurred using facts supported by the work record and the timekeeper’s knowledge.
2. Evaluate the entry in context. Check the description, matter, duration, codes, staffing, approvals, and other relevant requirements against the applicable rules.

Better prose cannot make prohibited work compensable, supply an approval that does not exist, or cure a staffing or rate problem. Treat those as separate tests.

Five elements to inspect

A useful review does not need a rigid sentence formula. It does need enough information to answer five questions.

1. Action

What discrete work occurred? Verbs such as analyze, draft, revise, prepare, and confer usually communicate more than labels such as work on or attention to. The verb must still match the evidence. Do not turn a review into an analysis simply because the latter sounds more valuable.

2. Object

What document, communication, record set, proceeding, or issue received the work? “Review correspondence” is more useful when the permitted description identifies the correspondence’s subject or role in the matter.

3. Purpose

Why was the task undertaken? Purpose can help a reviewer connect work to the matter, but it is also where unsupported detail often enters. Include purpose only when the work record or timekeeper supports it. If the system saw a document being reviewed but could not know the intended next step, it should not invent one.

4. Scope

What bounded detail lets the reviewer understand the charge? That might be a record category, motion, deposition topic, or stage of a transaction. Scope is not a demand for a play-by-play. It should be only as specific as the applicable guideline and confidentiality obligations permit.

5. Constraint

Does the whole entry satisfy the client’s rule, not merely read well? Check the matter, task boundary, duration, code, approval, and supported entry-level billing rules independently.

Specific enough does not mean revealing everything

Specificity and appropriate disclosure are not opposites. A useful narrative can identify the kind of work, its subject at a safe level, and its supported purpose without reproducing privileged communications, legal advice, confidential facts, or litigation strategy. Privilege, work-product protection, and ethical confidentiality are related but distinct; a billing template should not pretend to resolve those legal questions for every matter.

The right level depends on the audience, agreement, matter, and governing law. ABA Model Rule 1.6 addresses protection of information relating to a representation; it does not prescribe invoice wording and is not the law in every jurisdiction. Firms should apply their own professional-responsibility guidance when deciding what an invoice may disclose.

For example:

• Overexposed: Analyze claimant’s private medical history to develop impeachment strategy for Dr. [name].
• Safer level of detail: Analyze designated medical records concerning issues identified for expert examination.

The second version is not automatically correct. It works only if those facts are supported and the client’s rules consider that level of detail sufficient. The point is to preserve a meaningful action and object while removing unnecessary sensitive detail.

This matters even more when AI assists with drafting. ABA Formal Opinion 512 emphasizes understanding generative AI’s limitations, protecting client information, supervising its use, and reviewing its output. A fluent draft is not a substitute for professional judgment.

Before-and-after examples

The examples below use fictional matters and generic facts. They illustrate a review method, not approved billing language.

Example 1: a thin activity label

Observed work: The timekeeper reviewed a defined set of maintenance records. The records concern a notice timeline, but the intended next step is not evident.

Before: Review documents.

Better supported draft: Analyze maintenance records concerning the reported notice timeline.

Do not add without confirmation: ...to prepare cross-examination of the site manager.

The improved version identifies the action, object, and supported issue. It does not invent a strategic purpose. If the client requires purpose, the timekeeper must supply or confirm it before approval.

Example 2: communication without a case biography

Observed work: The timekeeper prepared an email to opposing counsel about proposed deposition dates.

Before: Email regarding case.

Better supported draft: Prepare correspondence to opposing counsel regarding proposed deposition dates.

Naming every recipient or reciting the negotiation is unnecessary unless the client’s rules require it and disclosure is appropriate. A participant’s role may provide enough context without exposing personal details.

Example 3: one remembered block, three observed tasks

Observed work: Research, drafting, and a conference occurred as distinct segments with their own evidence and durations.

Before: Research discovery issue; draft response; conference with client regarding strategy: 2.8 hours.

Better approach: Preserve separate proposed entries:

• Research governing standard for the disputed discovery request.
• Draft response to disputed discovery request.
• Conference with client representative regarding response approach.

Each entry should retain its supported duration. Do not divide a remembered 2.8-hour total into invented allocations. Whether related work may be grouped is a client-specific rule; a useful outside counsel guideline operating process makes that rule available before approval.

Example 4: language cannot create prior approval

Observed work: A task was performed in a category requiring prior approval, but no approval reference is available.

Before: Prepare detailed research memorandum concerning jurisdictional issue.

Better approach: Keep the accurate description and flag the missing approval. Add an approval reference only if approved matter data or the timekeeper confirms it.

The problem is not the sentence. Rewriting it cannot make the underlying charge authorized.

Review the evidence before editing the sentence

Narrative review is most reliable when it happens while the work context still exists. A timekeeper should be able to compare the proposed entry with the underlying work, confirm the matter and duration, inspect the applicable rule, and then edit or approve it.

Hourglass drafts entries from observed evidence about work performed in supported applications. That evidence can inform the duration, matter, narrative, and billing codes. When the available evidence is not strong enough to support a matter, narrative, task code, or activity code, Hourglass can leave the field blank for the timekeeper rather than manufacture certainty.

Approved, configured entry-level rules can then be applied at the firm, client, or matter level. Depending on the administrator’s configuration, Hourglass can explain an issue, suggest a correction, warn the reviewer, or block release. A person must still review and explicitly approve the entry before it is exported to the firm’s billing system.

That boundary is important. Hourglass checks supported entry-level conditions; it does not control the invoice after export or claim to resolve rates, budgets, expenses, invoice formatting, or substantive billing disputes. The production data trail ends when the approved entry is sent to the billing system. For a broader view of what can be caught earlier, see why legal invoice deductions happen and the Hourglass billing-compliance workflow.

A nine-question approval checklist

Before releasing a draft, ask:

1. Does this entry describe work that actually occurred?
2. Is it assigned to the correct client and matter?
3. Are the task boundary and duration supported by the work record?
4. Does it identify enough action, object, and context for this client?
5. Did the draft add a purpose, participant, issue, result, or next step that was not known?
6. Does the description agree with the selected task and activity codes?
7. Does it satisfy the applicable approved entry-level rules?
8. Does it disclose only the appropriate amount of information?
9. Has the timekeeper resolved visible uncertainty and approved the result?

Billing leaders can use the same structure to diagnose recurring cleanup. Track substantive narrative edits separately from matter corrections, duration changes, coding repairs, and rule warnings. A single “quality” score hides which control failed and where the process should improve.

The most durable billing narrative is not the most elaborate one. It is the one that can be traced to real work, says enough for the applicable reviewer, respects the matter’s disclosure limits, and leaves the final decision with the person responsible for the entry.