Should Lawyers Bill Hourly for AI Agents?
- For lawyers using AI agents, two fee models compete: meter the agent's time separately, or stop selling time and price the work.
- ABA Opinion 512: hourly billers must bill actual time, and the same flat fee may be unreasonable when AI makes the work much faster.
- Either way, explain the basis for any AI charge before charging, preferably in writing, and check your own jurisdiction's rules.
Lawyers who hand real work to an AI agent now face a billing question with two serious answers. One says keep the billable hour and put the agent on the invoice at its own rate. The other says stop selling time altogether and price the thing the client is buying. Both have to survive the same referee: the fee section of ABA Formal Opinion 512. Here's how each model works, and what the opinion's text asks of each.
We've covered the opening problem before, in what a lawyer should charge when AI does it in minutes, and the case against the hour itself in why AI turns the billable hour into a pay cut. This post is about the mechanics: what each model puts on the bill, and what it has to do before it can.
Where the opinion starts: your actual time
ABA Formal Opinion 512, issued in July 2024, is the ABA's guidance on generative AI. Its fee section begins with the hourly biller:
"GAI tools may provide lawyers with a faster and more efficient way to render legal services to their clients, but lawyers who bill clients an hourly rate for time spent on a matter must bill for their actual time."
Its example shows what the lawyer's own hours can include:
"If a lawyer uses a GAI tool to draft a pleading and expends 15 minutes to input the relevant information into the GAI program, the lawyer may charge for the 15 minutes as well as for the time the lawyer expends to review the resulting draft for accuracy and completeness."
Input time counts, and so does review. The tool's own working time isn't covered. That gap is where the two models part ways.
Model one: put the agent on the clock
Seth Fenster, who writes about AI and law, laid out the first model in a LinkedIn post arguing law firms should bill by the hour for AI agents. Fenster frames the bind like this:
"In a law firm, billing the full hour is an ethics violation, and billing only your 10 minutes means losing 50 minutes of billable revenue."
Fenster's answer is to bill the agent's time separately, at a much lower rate than the lawyer's. In the hypothetical invoice, the lawyer's 10 minutes and the agent's 50 minutes together come to a quarter of what a full lawyer hour would have cost.
The model rests on what the agent is. Fenster doesn't mean a chatbot anyone could open. It's the firm's own build, which is also why mapping the process before you automate it matters so much. In Fenster's words:
"I'd say the agent is just an expensive computer configured by the firm, with a lawyer checking its work."
Fenster contrasts law with other businesses: "In almost every business on earth, if you deliver the same product while spending 80% less time and money, you get promoted." In this model, the client pays far less than a lawyer hour, and the firm still earns on the agent's minutes.
Model two: stop selling time
The other answer comes from an SPU course that predates AI agents by years: Fixing Your Fees, Fixing Your Practice, taught by Jay Shepherd. The course's position was that hours were never the product in the first place:
"We sell things, we should think about that a little more. You don't sell time, you sell things, whether defenses of cases or wills as I've talked about, or real estate closings or whatever, you sell a widget, you sell a thing. And the thing has to be priced."
On that view, "whose time do we bill?" never comes up, human or machine. The course put it bluntly when a consultant insisted fixed-price firms keep tracking time:
"Okay, if you're not selling your time and you're pricing your work instead, who cares how long it takes?"
The course also had a view on what tools are. Rent, insurance, computer research and payroll were costs of running the firm, not of any one matter: "This idea that we should somehow have to apportion that overhead to our clients is ridiculous." Expenses were different. In the lecture's words, "Expenses are things that you pass along, things like court reporters and deposition transcripts." Hold on to that split, because the ABA draws nearly the same line. (For how the course turned this into an actual number, see the 14 steps for setting a fixed price.)
The referee: three places an AI charge can go
Opinion 512's fee section doesn't specifically address an agent billed at an hourly rate of its own. That section covers the lawyer's time, flat and contingent fees, and the cost of AI tools. But it sets the frame both models work inside. First, Rule 1.5 "applies to representations in which a lawyer charges the client for the use of GAI," and under Rule 1.5(b):
"before charging the client for the use of the GAI tools or services, the lawyer must explain the basis for the charge, preferably in writing."
Second, the opinion sorts tool costs. A tool that works like part of the office is overhead: "To the extent a particular tool or service functions similarly to equipping and maintaining a legal practice, a lawyer should consider its cost to be overhead and not charge the client for its cost absent a contrary disclosure to the client in advance." A third-party service that charges per use, such as one used to review thousands of contracts for one client, is different: "it would ordinarily be reasonable for the lawyer to bill the client as an expense for the actual out-of-pocket expense incurred for using that tool." And the opinion says no surcharge on the actual cost of an expense without disclosure.
Where an AI cost can go under Opinion 512
What the agent-hour model has to answer
Read against that text, a firm billing agent time faces three questions. None of them is a verdict on the model. They're the things its engagement letter would need to handle.
Which bucket is the agent in? An agent billed at an hourly rate, like Fenster's hypothetical $100, isn't an out-of-pocket cost passed through at actual cost, so the opinion's expense path doesn't describe it. If the agent is closer to equipment the firm built and maintains, the opinion's overhead language applies, and that language turns on "a contrary disclosure to the client in advance." Either way, the model depends on telling the client up front.
Is the lawyer's time still actual time? The 15-minute example supports billing the lawyer's input and review as they happened. It doesn't let the lawyer's line absorb the agent's minutes.
Is the total reasonable? Rule 1.5(a) still governs the whole bill. And Fenster reports that one response in a group chat was that "at least one state has already banned it." Fenster doesn't name the state, and neither do we. That alone is reason to check your own jurisdiction's rules before trying it.
What the fixed-price model has to answer
The course's model gets no free pass either. The opinion's flat-fee passage is aimed squarely at it: "if using a GAI tool enables a lawyer to complete tasks much more quickly than without the tool, it may be unreasonable under Rule 1.5 for the lawyer to charge the same flat fee when using the GAI tool as when not using it." It adds, quoting a Maryland case: "A fee charged for which little or no work was performed is an unreasonable fee."
So "price the thing" means price it on the Rule 1.5(a) factors, not on last year's number. The course made the same point about reasonableness, years before AI, in its ethics lecture:
"Your fees have to be reasonable, and that is really important, obviously. That is the escape hatch. You cannot just rely on, oh, hey, my client and I agreed that this would be our fee because the client could agree to an unreasonable fee."
The course read the factors as a list, not a clock: time and labor are "only one of the factors. In fact, it is only part of one of the factors." The others include the amount involved and the results obtained, the experience and ability of the lawyer, and the fee customarily charged in the area.
On tools, the two sources line up. The course called computer research overhead. The opinion says to treat a tool that works like equipping the practice as overhead and not charge for it absent advance disclosure. A fixed-price firm that absorbs its AI costs is already there. If it passes a per-use tool through, actual cost, explained in advance, is the line. The scope and the price belong in a written agreement, which our post on fixed-price agreements that hold up covers, along with this archive piece on writing a better fee agreement.
Two models, one referee
Both Explain the basis before charging, preferably in writing. Check your jurisdiction's rules.
Picking a model
We're not going to pick for you, and the opinion doesn't either. Its fee section is about honesty and reasonableness, not which structure wins. Both models can be explained to a client in a sentence, and both need that sentence in writing before the bill goes out. State rules vary, and this isn't legal advice. Check your own jurisdiction's rules and any opinions your bar has issued since July 2024.
Whichever you choose, say it the same way everywhere a client looks. If your engagement letter bills agent time and your website promises flat fees, a client will notice. SPU offers a free law firm website review that can catch that kind of mismatch. For more on the tools behind these decisions, see our technology coverage.
Work through your fee model with other lawyers
The full Fixing Your Fees, Fixing Your Practice course is in the SPU community, included with membership. Beyond the ethics lecture quoted here, it covers pricing litigation by phase, risk-shifting fees and fixing pricing mistakes, and ends with an exercise where you price one of your own services. SPU membership gets you the community, monthly strategy sessions and the course library.
Get one practical lesson for your firm every week.
Free, from practicing lawyers. Every issue includes a lesson from the archive. Unsubscribe anytime.
The SPU Team writes and edits Solo Practice University's articles. Susan Cartier Liebel founded SPU in 2009 to teach lawyers the practical side of building a practice, the part law school leaves out, with courses taught by practicing lawyers. Today SPU is a community for solo and small-firm lawyers, with an AI Fluency program for firms putting AI to work. Nothing here is legal advice: check your own jurisdiction's rules.
About Solo Practice University →