Using AI does not mean charging less: a court ruling on legal fees and artificial intelligence

A fee claim challenged over AI use

A group of claimants lose a lawsuit and must pay the opposing lawyer’s fees. When challenging the fee claim, they argue that the firm used artificial intelligence. If the tool saved work, they contend, the bill should be lower.

In Flux Drive v. Rong, that argument reached a federal court of appeals based in Denver, which hears cases from Colorado and five other western states. On 30 October 2026, it upheld two fee awards: an initial $114,663.11 and a supplementary $68,990.50.[1]

This is the first appellate ruling to examine whether AI use alone justifies reducing recoverable legal fees. The answer was no: the appellants had not proved the excess they alleged.[1]

Reasonable hours at a reasonable rate

The court applied Colorado law to fees recoverable following a dismissal before trial. The calculation followed the lodestar method: reasonable hours multiplied by a reasonable rate. The party seeking payment must justify it. The judge must exclude unnecessary, irrelevant or duplicated work.[1]

The time records, voluntary reductions and market rates had been examined. Faced with that documentation, the appellants did not identify specific duplication or provide evidence that AI had led to excessive billing. Pointing to the tool did not show that the hours or the amount were unjustified.[1]

The ruling recognises that AI can reduce research time and, with it, fees calculated by the hour. It also recalls that professional conduct rules require lawyers to check the court decisions they cite. Getting an answer is not the same as having a legally verified result.[1]

The ruling does not authorise billing for hours that nobody worked. Nor does it calculate how much time the tool saved or declare that reviewing its output always uses up those savings. It resolves a specific challenge under Colorado law. It does not set the price of legal services or establish a rule applicable in every country.

The practical consequence is narrower: challenging a fee claim requires identifying and substantiating the excess. AI use can form part of that explanation, but cannot replace it.

In the United States, the ruling is consistent with the position of the American Bar Association, the US bar association — guidance, not law. Its Formal Opinion 512 allows billing for the actual time spent instructing the tool and reviewing its output. It distinguishes between ways of passing on the cost of tools and warns that, with a fixed fee, substantial time savings may make keeping the same fee unreasonable.[2]

Spain and Norway: identifying the excess

In Spain, the price agreed with the client must be distinguished from the amount recoverable from the party ordered to pay costs. Article 245.4 of the Civil Procedure Act (LEC) requires a challenge to identify the fee statements, the specific items and the reasons for disagreement.[3] Applied to AI, the argument would need to explain which time, duplication or task is excessive. This is a reasoned application of the provision, not a specific rule established by case law on these tools.

In the relationship with the client, article 14.1 of the Spanish Code of Conduct for Lawyers provides that the amount and basis of fees are freely agreed.[4] Bar associations’ indicative criteria are reserved for the assessment of litigation costs and the special procedure for recovering lawyers’ unpaid fees. They are not general fee scales for clients.[5]

In Norway, § 20-5 of the tvisteloven allows recovery of necessary litigation costs, taking account of whether it was reasonable to incur them given the importance of the case.[6] RGA 3.3.1 requires fees to bear a reasonable relationship to the assignment and the work performed.[7]

HR-2022-563-A recalls that the hours actually worked do not, on their own, determine whether a fee is reasonable. It concerned a contractual relationship, unrelated to AI or the opposing party’s costs.[8] Publicly funded legal assistance is governed by the salærforskriften.[9] Its amendment, applicable to new assignments from May 2026, requires billing only for time actually spent. This is not the general regime for private legal practice.[10]

The comparison raises two separate questions: how much the client should pay for the agreed service, and what costs the opposing party should bear. Technology does not remove that distinction.

AI also costs the firm money

Saving time does not mean working at no cost. Licences, tools and infrastructure are part of the firm’s costs. The question is how they are reflected in the price and what the client is told.

A general-purpose tool used by the firm is an overhead like any other. The use of an external service for a specific matter may be charged to the client if this has been agreed. In both cases, the billing method must be explained and the same cost must not be charged twice.

Simply adding an “AI” item to the bill is therefore not enough. It must be clear what it pays for and whether its cost was already included elsewhere. That transparency makes it possible to discuss the price with information about its components.

Investing in technology does not automatically make its cost an expense that can be passed on to the client.

In economic terms, net savings depend on the time saved, the cost of the tool and the review required. A faster answer alone does not reveal how much cheaper the complete service has become.

Giving instructions and checking are work too

Instructing the tool and reviewing its output are part of the lawyer’s working time. With hourly billing, the charge is for actual time spent. Time spent learning to use the application cannot simply be passed on to the client.

That work of instruction and oversight is the lawyer’s responsibility. The generated answer needs to be reviewed before it is incorporated into advice or a court submission. As Flux Drive recalls, checking the court decisions cited is a professional obligation.[1]

Review counts when it is performed. It does not justify adding the hours that research without AI would have required, or assuming that checking always consumes all the time saved. The record must reflect the work actually done. Its necessity and reasonableness remain open to scrutiny.

For the client, the question goes beyond how long the tool took. What the lawyer did with the answer, and the cost attributable to that work, also matter.

If the result is better, why should it cost less?

The paradox emerges when comparing two ways of conducting legal research. Suppose a trainee lawyer is given the task: it takes longer and produces a poorer result. The lawyer must then review it. With hourly billing, that process can cost the client more.

Now suppose that AI, under the lawyer’s instructions and oversight, allows the same task to be completed more efficiently, in less time and with a better result. At the same rate and for the same scope of work, the hourly bill falls if total actual time decreases. Technology costs must also be included in the comparison.

This scenario does not claim that any AI outperforms any trainee lawyer. It asks what happens when greater efficiency translates into fewer billable hours. The client gets a better result sooner and pays less. The firm gains capacity for other assignments, but receives less for that service.

A fixed fee distributes the savings differently. If the price stays the same and the cost falls, the firm’s margin increases. The client retains price certainty and may receive the work sooner. Advokatforeningen accepts that efficiency can benefit the lawyer under a fixed fee, subject to reasonableness. Applying that criterion to AI is an inference: the opinion did not concern these tools.[11]

There is a limit to that margin: the fee must remain reasonable. Prior agreement does not, on its own, settle how much of the savings the firm can retain.

A price linked to the value of the service shifts attention towards its scope, risk and the work undertaken. It does not remove the need to justify the amount. It is not the same as charging a percentage of the outcome, an arrangement restricted in Norway by RGA 3.3.2.[7]

Flux Drive resolved a failure to prove excessive billing. A different question remains open: if better research, completed in less time under the lawyer’s responsibility, delivers more value to the client, why should it cost less? Is the hour the right measure?