AI Saved the Hour. The Lawyer Cannot Bill It Twice.

The court affirmed a record, not an AI billing rule

Flux Drive and related companies brought trade-secret and other claims against several defendants in Colorado federal court. The district court dismissed the claims against three Illinois defendants for lack of personal jurisdiction and awarded those defendants fees under a Colorado fee-shifting statute. The initial award was $114,663.11. A later supplemental award, reduced to $68,990.50 after an administrative percentage was excluded, covered additional fee litigation. The Tenth Circuit affirmed the dismissal and both awards in a published opinion.

The plaintiffs attacked the first fee award on several grounds. They alleged serial billing, partner time spent on ministerial work, and hours that should have been lower because defense counsel used AI and online research tools. The panel did not say that technology is irrelevant to the lodestar. It expressly acknowledged that AI and online research can reduce the time reasonably required for a task and, in some circumstances, lower the hours used to calculate a fee.

The appeal failed because possibility is not proof. The plaintiffs did not identify an inflated entry, a redundant task, or evidence connecting the firm's AI use to an unreasonable number of hours. The magistrate judge had reviewed detailed billing records and found proper billing judgment. The appellate court applied abuse-of-discretion review and concluded that the challengers had not shown why that finding was wrong.

That distinction matters. A published opinion affirming a fee award can be flattened into the claim that courts approve traditional billing even when AI is used. The opinion supports no such safe harbor. Its narrower lesson is procedural: a party challenging fees must do more than point to the existence of a tool. It must connect the tool, the task, the time entry, and the asserted excess.

New York answers the question before the fee dispute

The New York State Bar Association's House of Delegates approved its Legal Ethics Guidelines for Lawyers' Use of Artificial Intelligence on October 31. The guidelines do not amend the New York Rules of Professional Conduct or create independent disciplinary obligations. They explain how existing duties apply to AI tools and systems. On fees, the language is unusually direct.

An hourly lawyer may bill for time actually spent formulating inputs, reviewing output, verifying authorities and facts, and revising the work. The lawyer may not bill the time that would have been spent without AI or charge for the value of the work product while describing the charge as hourly time. General familiarization with an AI tool is not client work. Adapting a tool to a particular matter can be billed if the time is reasonable. Ordinary acquisition and maintenance costs are generally overhead unless a matter-specific charge is reasonably allocable, disclosed, and reasonable.

Non-hourly fees receive different treatment. The guidelines say that an AI efficiency gain does not, standing alone, require a flat or contingent fee to be reduced. Those arrangements remain subject to the overall reasonableness factors in Rule 1.5. That is not a loophole. It is a recognition that the client agreed to buy an outcome, defined service, or allocation of risk rather than a ledger of six-minute units.

The national baseline points in the same direction. ABA Formal Opinion 512 says that hourly lawyers must bill actual time spent and cannot charge a client for general training in a tool the lawyer expects to use across matters. New York's report adds detail, but the economic signal is familiar: technology may change the cost of producing legal work without changing the duty to describe the price honestly.

Verification is real work, not a surcharge

The most important sentence in the Tenth Circuit's discussion may be its reminder that lawyers must verify authorities and ensure that arguments are grounded in law. That work takes time. AI can accelerate retrieval, comparison, drafting, and organization, but it can also generate a confident answer that requires more investigation than an ordinary research trail. A fee analysis that assumes every prompt instantly replaces lawyer time misunderstands the work.

The opposite mistake is just as serious. Verification cannot become a vague surcharge added after a firm bills the time a task supposedly would have taken without AI. If a lawyer spends 18 minutes generating and checking a first-pass chronology, the hourly invoice cannot quietly report two hours because two hours was the historical estimate. Nor can the firm add another block for generic review while keeping no record of what was checked. The ethical value lies in the actual judgment exercised, not in preserving an old quantity of time.

Good time narratives will become more specific. They should identify the legal task performed and the professional review applied, without exposing privileged strategy or turning every entry into a technology log. 'Research and analyze personal-jurisdiction authorities; verify cited holdings; revise argument' is more defensible than 'AI research' or 'review AI output.' The first describes legal work. The latter descriptions invite the client or a court to guess what happened.

Courts reviewing fee petitions will need similar discipline. AI use alone is not evidence of overbilling, just as the absence of disclosed AI use is not evidence that the time was reasonable. The useful questions remain concrete: What work was necessary? What time was actually spent? Was the staffing appropriate? Did the record show billing judgment? Did the technology eliminate, duplicate, or create any part of the claimed work?

The billable hour cannot keep every efficiency gain

For decades, law firms have sold both expertise and time. Technology makes the tension between those products harder to ignore. If a better system lets a lawyer produce the same sound result in half the time, an hourly client receives the immediate savings. The firm may benefit through capacity, speed, client retention, or the ability to handle more matters. It may not recreate the eliminated hour on the invoice.

That will push some work toward flat fees, phased fees, subscriptions, success components, and other structures that price a defined service rather than the time used to manufacture it. None of those arrangements is automatically fair. A flat fee can be excessive, poorly explained, or misaligned with the client's risk. The answer is not to rename hourly work. It is to decide openly what the client is buying and test the price against professional rules and commercial reality.

Firms also need internal incentives that do not punish lawyers for using approved tools well. A lawyer who completes excellent work efficiently should not have to choose between missing a billable target and padding the record. Compensation, matter budgets, utilization measures, and partner expectations must change with the production model. Otherwise the firm will announce an AI strategy while preserving an economic system that rewards the opposite behavior.

Flux Drive protects a fee award that was supported by a reviewed record. New York's guidelines tell lawyers how to create an honest one. Together they reject the two easiest slogans: AI does not make legal work free, and it does not make saved time billable. The lawyer can charge for judgment actually exercised. The firm must find a business model for the rest.

AI can save time. It cannot convert time never spent into an hourly charge.