AI Answer Summary

Legal billing automation does not let a firm bill more. On hourly work the conduct rules require billing the time you spent, so compressing the work compresses the bill. What automation can raise is the share of billed work that survives review by clients, carriers, and courts.

The realization gap breaks into four components: rate concessions negotiated with the client, discretionary write-downs applied by a billing partner before the invoice goes out, line-item reductions applied after it goes out, and collection lag. Automation can affect the middle two. It cannot touch the first or the last.

The strongest move is counterintuitive. The 2026 Report on the State of the US Legal Market records associate realization averaging 85.6% and argues that the gap creates room to automate work clients were not paying for in the first place.

What Legal Billing Automation Can and Cannot Change

Automation cannot increase what you bill on hourly work, but it can increase what you keep.

If a tool turns a four-hour task into forty minutes of prompting and review, the hourly bill is forty minutes. Firms evaluating billing automation as a revenue play have the logic reversed, and the ethics section below explains why there is no way around it.

The automation can reduce the amount that gets cut from an invoice after the work is done. To see whether that is worth anything at your firm, you need to know where your own gap comes from.

Pull those four numbers before you evaluate a single tool. Two of the rows cannot be improved by software at all, and firms regularly buy software expecting them to move. For insurance defense practices the third row usually dominates. For firms with a heavily negotiated institutional client base, the first row does, and no billing system will help.

Billing Realization, Collection Realization, and the Number to Track

Three different calculations share the name realization, and most published figures do not say which one they mean.

Same firm, three different percentages and the third is the number a partner board argues about.

Worth knowing before you go looking for a benchmark: published realization figures for US law firms range from the mid-seventies to the low nineties, mostly in consultant and vendor content, often with no named dataset and no statement of which metric is being reported. Some of that spread reflects real variation by practice area and firm size. Some of it is three different calculations being reported as one number.

Our recommendation(and we are labelling it as a recommendation rather than a standard): report the overall figure to your board, and track the pre-bill write-down component separately, by practice group. That component is the one an operations change can shift, and it is the one nobody measures.

What the Rules Allow When Automation Compresses the Work

On hourly work you bill the time you spent. Efficiency does not create billable hours.

ABA Formal Opinion 512, issued in July 2024, is direct about this, and it rests on Opinion 93-379 from 1993, which remains the foundational billing authority. Two points from that older opinion still do the work:

A lawyer who has agreed to bill by the hour does not meet her ethical duty by billing a client for more time than she spent. If she turns out to be particularly efficient at reaching a result, she still cannot charge for hours she did not work. And in its most quoted line, a lawyer who spends four hours on behalf of three clients has not earned twelve billable hours.

Opinion 512 applies that to generative tools with a worked example. A lawyer who spends fifteen minutes entering information into a tool to draft a pleading may bill those fifteen minutes, plus the time spent reviewing the draft for accuracy and completeness. Nothing beyond that.

One further point matters for any firm moving away from hourly billing. Opinion 512 signals that charging the same flat fee may become unreasonable if a tool materially compresses the work. Jurisdictions have not landed in the same place on this. Virginia has read Rule 1.5 as leaving more room for value-based pricing, and North Carolina, Texas and DC have each addressed billing for AI-assisted work. If you are pricing non-hourly work, this is a question for your own state's guidance rather than the ABA's.

Why Billing Narratives Get Written Down

Reviewers cut entries they cannot assess. The defect is vagueness, not the grouping of tasks.

The clearest public record of narrative quality being priced sits in fee applications, where a judge reviews a bill line by line and states the reduction. That is a different setting from ordinary client billing, so read what follows as instructive rather than identical. The reasoning transfers even where the procedure does not.

The detail worth carrying from Welch is the part that gets misquoted. The district court had applied a 20% across-the-board reduction to all of the hours requested, taking 20% as the middle of the California bar committee's range. The Ninth Circuit vacated it. Barely more than half the submitted hours had been block billed, so cutting the total by 20% worked out as roughly a 40% penalty on the entries that had been block billed, well beyond what the range supported.

So the authority to reduce is settled. The size of the reduction has to be justified against the specific entries.

Two other reductions in the same case survived on appeal, and both are worth knowing. The court affirmed a 20% cut for billing in quarter-hour increments, because it was tied to evidence of over-billing: the time records were full of fifteen and thirty-minute charges for phone calls, emails and short letters. And it affirmed cutting a thirteen-hour fee motion to four hours, because the language was recycled from the firm's submissions in other cases. Reused text was found not to justify original time.

The record also contains a textbook example of the problem. One entry claimed 8.75 hours for reviewing insurance documents, preparing discovery and deposition notices, emailing opposing counsel and holding an intra-office conference. Four hours came off it.

That entry is the same thing a billing partner strikes on a pre-bill every month. The difference is that the partner never records why.

What Makes a Time Entry Survive Review

An entry survives when a reader who was not there can tell what was done, why it was needed, and whether the time fits the work.

Four properties do that:

• One task per line. Or, if tasks are grouped, each one inside the block carries its own detail and its own time
• The subject named rather than gestured at. Which motion, which agreement, which witness, which section
• The purpose visible, so a reviewer can judge necessity instead of guessing at it
• Time proportionate to the described work, since proportionality is what the reviewer is testing

Compare two versions of the same afternoon. The first: "Review and revise brief and analyze key documents and correspondence re same, 3.3." The second: "Revise argument section III of summary judgment brief to address new deposition testimony, 1.8. Review Henderson deposition transcript pp. 40 to 95 for admissions on notice, 1.0. Correspond with co-counsel on exhibit list, 0.5."

Same work, same total. The second one tells a reader what they are paying for.

Nobody fixes this by asking associates to write better entries. Firms have been asking for thirty years and the pre-bill still arrives full of the first version. It gets fixed by putting a step between time entry and the pre-bill, which is what the sequencing section covers.

Automate the Work You Were Already Writing Off

The safest work to automate is the work clients were already refusing to pay for.

This argument comes from the market rather than from us. The 2026 Report on the State of the US Legal Market, published jointly by Thomson Reuters and the Center on Ethics and the Legal Profession at Georgetown Law, records associate realization averaging 85.6%, with associate work already being written off at significant rates. The report's own conclusion is that this creates a buffer in which AI can absorb the inefficient portions without touching collected revenue, letting firms automate the work that was not getting paid for while keeping associates on higher-value tasks.

Our reading of that, stated as interpretation: it resolves the tension in the first section of this article. Automating billable drafting reduces the bill. Automating written-off work costs nothing in collected revenue and returns capacity. Those are opposite ends of the same decision, and most of the market is selling the first one.

A caution from the same publisher, because it cuts against the obvious response. The Law Firm Rates Report 2026, from the Thomson Reuters Institute with the True Value Partnering Institute, found that firms taking very different approaches to discounting and realization discipline end up collecting roughly the same amount per hour. If that holds, tightening realization policy is a weaker lever than firm leadership tends to assume, which points back at narrative quality and written-off work.

None of this is a new problem. Collection realization across the market fell from 95% in 2007 to 88% by 2015, so a firm looking at its own decline is looking at a long structural drift rather than a recent shock.

Where to Start: Sequencing and What to Measure

Start where the write-down is discretionary and recorded, because that is the only component you can prove you changed.

Four steps, in this order:

1. Pull the four component numbers from the first table for the last four quarters, split by practice group. The split matters, since a litigation group and a transactional group usually lose money in different rows
2. Sample fifty pre-bill adjustments and record the stated reason for each. The common finding is that no reason was stated, which is itself the finding
3. Fix the narrative step before buying anything that captures time faster. Faster capture of vague entries produces more vague entries
4. Set the measurement before deployment: pre-bill write-down percentage by practice group, month on month, against the baseline from step one

Best for whom. A narrative cleanup step suits firms where partners are editing pre-bills by hand and the reasons go unrecorded. A time-capture tool suits firms where hours are going unrecorded entirely, which is a different problem needing a different product. Firms often have both and buy only the second.

One honest limit. Nothing here moves a rate concession negotiated by a client's procurement team, and no automation collects a receivable that has not moved in ninety days. If your gap sits mostly in those two rows, the answer is a pricing conversation or a collections process, and you should spend the budget there.

How Codebridge Builds Billing Narrative Workflows

We build the step between time entry and the pre-bill. Entries get checked against what a reviewer needs to see, flagged where the subject or the purpose is missing, and rewritten as a suggestion for the timekeeper or the billing partner to approve or reject.

The lawyer approves every change. A time entry is a representation about that lawyer's own work, so nothing goes onto an invoice without them accepting it. The approval gets recorded, which gives the firm something it usually lacks: a record of which adjustments were made and why, by practice group, that can be read six months later.

One workflow goes live in three weeks, wired into the practice management and billing systems the firm already runs. Your firm owns the repository, the prompts, and the configuration from day one.

Our founding team spent more than a decade at KPMG, which is where we learned what a partner will and will not put their name to.

If you want to look at where your write-downs are coming from, book a 20-minute call and we will go through the four components with you.