A lawyer who knowingly makes a client pay for avoidable work is ripping off that client. The hours may be real. The waste is real, too. AI now gives the profession a practical test of whose interests it puts first.
A client needs a contract, a complaint, or an expert report. An available AI system could produce a useful draft in minutes and help identify omissions, test arguments, and improve the finished document. A lawyer who rejects that opportunity should have to explain what the client gains from the extra expense.
The client should not have to finance the lawyer’s refusal to modernize.
Consider an illustration. A lawyer expects to spend 12 hours preparing a finished document. AI produces a more thorough initial draft in 12 minutes. Suppose the lawyer actually spends 12 minutes preparing the AI inputs and another full hour verifying and revising the result. The billable work totals 1.2 hours.
| Illustrative method | Actual lawyer time | Fee at $500 per hour |
|---|---|---|
| Manual preparation and review | 12 hours | $6,000 |
| AI inputs plus professional review | 1.2 hours | $600 |
| Difference | 10.8 hours | $5,400 |
These are illustrative figures, not measurements of every legal assignment. They show why necessary human review does not, by itself, erase the possibility of enormous savings. Actual preparation, investigation, negotiation, or filing work must also be counted when the assignment requires it.
The quality argument has evidence behind it. A randomized study published in April 2026 tested upper-level law students using a legal AI tool, a reasoning model, or no AI. Both tools improved overall work quality and speed. The strongest gains appeared in complaint analysis and persuasive writing. The study found no meaningful overall benefit on its short nondisclosure agreement task, and accuracy did not improve consistently across tasks. The sensible conclusion is to evaluate the tool against the assignment. journals.sagepub.com
Evidence also extends to practicing lawyers. A September 2026 working paper studied 133 patent lawyers at eleven U.S. firms in a three-month randomized trial. Blinded expert attorneys judged the work. Access to a custom AI drafting assistant improved benchmark drafting quality at both ten and ninety days, with larger gains among junior lawyers. The study also found that lasting gains in unaided professional judgment were concentrated among senior lawyers. That supports combining AI with expertise and continued training. shapingwork.mit.edu
Better legal work can mean identifying a missing issue, exposing an unsupported assertion, or reorganizing an argument so that a judge can understand it. I would have lawyers test AI for those purposes and measure the result. A longer document earns no credit merely for being longer. Completeness means covering the relevant issues accurately.
Hourly billing creates a financial incentive that deserves scrutiny: fewer hours can mean less revenue from that client. Professional ethics must keep the client’s interests ahead of that incentive.
The clearest existing rule is straightforward. ABA Formal Opinion 512 says an hourly lawyer must charge actual time, including genuine review, rather than the hours the work would have taken without AI. A lawyer cannot turn a 12-minute drafting exercise into a fictional 12-hour entry. www.americanbar.org
Alabama’s 2026 guidance reinforces that distinction and generally treats AI subscription charges as office overhead unless the client agrees otherwise. Genuine supervision and legal judgment are compensable. Saved hours are not working hours. alabar.org
Knowingly inventing hours also implicates Pennsylvania Rule 8.4(c), which prohibits dishonesty and misrepresentation. A fabricated invoice deserves disciplinary scrutiny. www.padisciplinaryboard.org
The harder problem is the lawyer who actually spends twelve hours doing work manually when a safe, reliable method could accomplish the assignment far more efficiently. Pennsylvania Rule 1.5 prohibits illegal or clearly excessive fees. Its factors include the time and labor required, the difficulty, skill, customary charges, and results. Hours actually worked are therefore only part of the inquiry. www.padisciplinaryboard.org
My argument is that knowingly preserving unnecessary labor solely to enlarge an invoice should support an excessive-fee challenge. Current guidance does not establish a universal duty to use AI for every document. North Carolina expressly leaves the choice generally to professional judgment, while requiring client input when the decision to use or decline AI implicates fees. Those qualifications leave room for judgment; they should also require an explanation of avoidable expense. www.ncbar.gov
Applicable state rules govern discipline; ethics opinions guide their interpretation. The proposed duty to explain avoidable expense should be made explicit.
A fixed fee can reasonably reward expertise, responsibility, results, and efficiency. It does not remove the fee limits. Opinion 512 specifically warns that keeping the same flat fee after AI substantially reduces the work may be unreasonable. A lawyer should be able to explain the value provided at the agreed price. www.americanbar.org
Nor should the client ordinarily pay for the lawyer’s general education in using AI. Opinion 512 treats learning a tool regularly used in practice as part of maintaining competence, with a limited exception where a client specifically requests an unfamiliar tool and agrees to the training charge. www.americanbar.org
A recent federal decision also weakens a blanket objection that AI drafting necessarily destroys protection for litigation preparation. In Villanueva v. Las Vegas Metropolitan Police Department, No. 2:24-cv-00125-ART-DJA, Judge Anne R. Traum’s September 29, 2026 order partly reversed a demand for an expert’s entire ChatGPT history. storage.courtlistener.com
The expert’s AI-generated case summaries were discoverable because he considered them in forming his opinions. His prompts, deposition preparation outline, and report drafting material were protected. The court also held that his use of ChatGPT did not waive otherwise applicable protection in that instance. It treated ChatGPT as research, drafting, and editing software. This was a discovery ruling, not a certification of the expert’s accuracy or a decision about legal fees. storage.courtlistener.com
The lesson is to handle AI materials competently. Rule 26 distinguishes facts or data considered by a reporting expert from protected drafts. Counsel must plan for required preservation and disclosure while asserting applicable protection. Ordinary contract drafting does not automatically qualify for litigation work-product protection. www.law.cornell.edu
Accuracy and confidentiality are real obligations. Stanford researchers’ 2024 evaluation found that the particular commercial legal research tools tested still generated hallucinations in roughly 17 to 33 percent of queries. That is a historical finding about those systems and tests, not a current error rate for every AI product. It establishes why verification matters. RegLab
Lawyers should verify authorities, compare factual summaries with source records, review the provider’s data practices, and comply with the court’s applicable rules. Alabama’s guidance requires professional judgment and safeguards. These duties belong in a competent workflow. They warrant necessary review time and can justify declining a particular tool when the risks cannot be managed. alabar.org
I propose a stronger express professional standard:
When a reasonably available technology can materially reduce cost or improve legal work without compromising accuracy, confidentiality, or the client’s objectives, the lawyer should use it or explain to the client the reason for declining it.
Bar authorities should adopt that standard and examine fee complaints for unnecessary labor as well as invented hours. Law firms should compare completed work, including verification costs, and record where AI helps, where it fails, and how the client benefits.
Clients should ask for clear answers before agreeing to the price:
- What parts of this assignment can AI perform safely and effectively?
- What human investigation, judgment, and review remain necessary?
- How will the savings affect my hourly bill or fixed fee?
- If you decline AI, what specific benefit justifies the additional cost?
Lawyers can earn more by serving additional clients and offering work those clients could previously not afford. North Carolina’s opinion expressly recognizes increased throughput as a legitimate benefit of AI efficiency. www.ncbar.gov
Lower drafting costs can make a sound contract, a meritorious complaint, or a carefully checked expert report affordable. That opportunity deserves a place in every serious discussion of access to justice.
The client’s money belongs to the client. Professional ethics should protect it from avoidable waste.
No comments:
Post a Comment