AI and Attorney Fees: Can You Bill Clients for AI-Assisted Work?

Yes, a law firm can bill clients for AI-assisted work, but the bill must be reasonable and honest. On hourly billing, charge for the time you actually spend, including reviewing the AI's output, not the time AI saved. On fixed fees, the price is for the outcome. Describe the work plainly on the invoice and tell clients how AI is used.

The main reference point is ABA Formal Opinion 512, issued on 29 July 2024. It is guidance from a professional body interpreting its model rules, not a statute, and ethics rules are set jurisdiction by jurisdiction, so treat this tutorial as practical orientation and take it to your own regulator or ethics adviser before you change your terms.

Follow me on Instagram@sagnikteaches

What Opinion 512 says about fees, in its own terms

The fee section applies Model Rule 1.5, which requires fees and expenses to be reasonable and the basis of a charge to be explained to the client. Before charging a client for the use of AI tools or services, the opinion says, the lawyer must explain the basis for the charge, preferably in writing. On hourly billing it is blunt: lawyers who bill an hourly rate must bill for their actual time. Its own example is a pleading drafted with AI, where the lawyer may charge for the 15 minutes spent entering the relevant information and for the time spent reviewing the draft for accuracy and completeness. Being efficient does not entitle you to bill the hours the task used to take.

Connect on LinkedInSagnik Bhattacharya

Three further points are often assumed to be missing from the opinion, and are not.

Subscribe on YouTube@sagnikteaches

• Flat fees. The same reasonableness factors apply to flat and contingent fees. The opinion says that if an AI tool lets a lawyer complete tasks much more quickly, it may be unreasonable to charge the same flat fee as when not using the tool.
• Tool costs. A tool that works like equipping and maintaining a practice is overhead and is not charged to clients absent a contrary disclosure in advance; the opinion's example is an AI grammar checker built into word-processing software. A third-party service that charges per use, such as one that reviews thousands of contracts for a particular client, can ordinarily be billed as an expense at the actual out-of-pocket cost. Absent disclosure to the contrary, no surcharge is added and any discount is passed on.
• Learning time. A lawyer may not charge a client for learning a tool the lawyer will use regularly for clients, because staying competent with your tools is your own cost. If a client explicitly asks for a specific tool the lawyer does not know, billing the time to learn it may be appropriate, ideally under terms agreed and written down first.

The opinion also covers competence, confidentiality, communication and supervision, and two of those points land squarely on billing conversations. Informed client consent is required before information relating to the representation goes into a self-learning AI tool, and merely adding general, boiler-plate provisions to an engagement letter is not sufficient to obtain it. And under the communication duty, lawyers must consult clients when the use of an AI tool is relevant to the basis or reasonableness of the fee. That is why billing and consent conversations so often arrive together.

Hourly billing: how to keep it honest

The easiest way to get hourly billing wrong with AI is to keep the old time estimate in your head. The right habit is to record what you actually did. Three kinds of time are billable in principle, and you should track them separately in the first months.

• Prompting and preparation: time spent gathering the facts, writing the instruction and loading the right documents.
• Review and verification: reading the output, checking citations, comparing against sources and correcting errors. This is often the longest part, and it is the part clients are paying for in a professional service.
• Judgement and drafting on top: applying the law to the facts, advising, and finalising the document.

The time the tool spent generating text is not your time and is not billable as yours. If a clause comparison that took two hours now takes 25 minutes of your attention, bill 25 minutes. Your rate card can reflect that your hour is worth what it is worth; it should not charge for an hour that did not happen.

Illustration: a lease review in a six-lawyer firm

Say a six-lawyer firm reviews commercial leases for small-business tenants. Historically a lease review took 3.5 hours at $300 an hour, a bill of $1,050. With an AI tool that summarises key terms and flags unusual clauses, the lawyer spends 20 minutes preparing the document set, 40 minutes verifying the summary against the lease, and 50 minutes advising and drafting a note: 110 minutes in all, or about 1.83 hours.

On an hourly basis the honest bill is about 1.83 hours, which is about $550 at the same rate. The firm's revenue per lease falls by roughly $500. The firm has three routes. It can accept lower bills and more volume. It can move to a fixed fee for lease reviews, priced on the value and risk to the client, and say what is included. Or it can use the time saved on work that is not AI-reducible, such as negotiation. What it should not do is bill 3.5 hours for 1.83 hours of work.

A fixed fee below the old hourly bill, say $750, would be reasonable if the review is consistently good, the scope is stated and the client agreed in advance. The firm now earns about $410 an hour on those matters and the client pays less than before. Keeping the fixed price at $1,050 while the work now takes under two hours is the situation Opinion 512 flags as potentially unreasonable. Both outcomes depend on the review being genuinely done: the efficiency comes from the tool, and the value from the lawyer's checking.

Flat fees and other models

AI pushes firms towards fixed prices for repeatable work because the link between time and value weakens. Three structures work for small firms.

Fixed fees do not remove the duty that the fee be reasonable and that the client understands it. They change the question from "how many hours" to "is this price fair for what was delivered?". If a task becomes routine and the price stays frozen at the old hourly equivalent, clients will notice, and the opinion's flat-fee warning applies, so review the price list each year. For help thinking about tool costs alongside fees, see per-seat or pay-as-you-go legal AI pricing.

Tool costs: overhead, expense or client charge

Three questions decide how to treat a tool's cost.

1. Is it used across the practice? A general assistant, a research subscription or a drafting add-in the whole firm uses is normally overhead, recovered through your rates. Billing it to clients as a separate line is hard to justify.
2. Was it bought for one matter? A one-off analysis tool, a large-volume document review service or extra capacity bought for one case can ordinarily be passed on as an expense at the actual out-of-pocket cost, described honestly, and on the basis you explained to the client before charging it.
3. Did the engagement letter say so? If you plan to pass any tool cost on, say it before the work, with how it will be calculated. Surprising a client with a technology charge is the fastest way to a fee dispute.

A practical default for small firms: absorb ordinary tool costs into rates, pass on extraordinary ones at cost with advance agreement, and never add a mark-up on a software subscription unless your rules and engagement letter clearly allow it. Treat the hours spent learning a tool the same way. Say the firm adopts a contract-review assistant and each lawyer spends three hours in the first month getting to grips with it: those 18 hours across six lawyers are an investment in the practice, not time to spread across client invoices. The exception is the client who insists on a particular platform the firm has never used, where the learning time can be agreed with that client in advance.

Invoice wording that is accurate and clear

Clients read narratives. Write them to describe the work in terms of what you did, not which tool did it, and be accurate about review. Compare these illustrative entries for the same task.

Weak:   "Lease review - 3.5 hrs"
Weak:   "AI-assisted lease analysis - 1.8 hrs"
Better: "Reviewed commercial lease (42 pages); verified AI-generated
         summary of key terms against the lease text; identified three
         clauses requiring negotiation; advised client by call and
         wrote summary note - 1.8 hrs"
Fixed fee: "Commercial lease review (fixed fee): summary of key terms,
         identification of unusual or risky clauses, advice call and
         written note. Includes one round of follow-up questions."

The better entry says what happened, shows verification and gives the client something to understand. Whether you mention AI by name in the narrative is a transparency choice unless your rules or the engagement terms require it; if your engagement letter says AI may be used, the narrative need not repeat it on each line. Pick a policy and apply it consistently, so that entries do not look evasive on some invoices and chatty on others.

Engagement letter wording and consent

Put the arrangement in the engagement letter, not in a later email. Illustrative wording, for your own adviser to review:

Use of technology. We may use AI tools to assist with tasks such as
summarising documents and preparing first drafts. A lawyer reviews and
is responsible for all work product. We will not enter your confidential
information into any tool that has not been assessed for confidentiality
and security. [If applicable: We will ask for your agreement before
using a tool that retains or shares your information outside the firm.]
Our fees reflect the time we actually spend, including review, or the
fixed fee agreed; we do not charge for time saved by technology. We do
not charge separately for general software tools. If we need to buy a
specialist service for your matter, we will agree the cost with you
first.

That paragraph does four jobs: it discloses use, promises review, states the fee principle and handles pass-through costs. What it does not do is obtain informed consent. Opinion 512 says a general clause of this kind is not sufficient where consent is required, which is whenever information relating to the representation goes into a self-learning tool. Consent there means a specific conversation, recorded on the file: why the tool is being used, what kinds of client information will go in, the risk that others could access it, including later users of the tool, and the benefit to the matter. Whether a given tool is self-learning in that sense depends on its terms, which the opinion says lawyers should read and understand, so ask your regulator or ethics counsel if the answer is unclear. For the meeting-recording side of that question, see AI note takers in client meetings.

What to say when a client asks "why isn't this cheaper?"

Clients who have heard that AI is cheap will ask. Prepare an honest answer that is neither defensive nor evasive. A usable pattern: acknowledge that tools have shortened some steps and that the bill reflects that; explain that the lawyer's review, judgement and responsibility remain, and that is what protects them; and show the narrative so they see what was done. If the firm has moved to a fixed fee, say what the fee includes and offer to explain how it compares with the old estimate. If the firm has not changed its price and the work is now routine, consider that the question is a sign that the price list needs review.

Another case: a letter drafted from a template

Say a firm drafts a standard demand letter. The AI tool produces a first draft in two minutes from the facts. The lawyer spends 15 minutes checking the facts, the legal basis and the tone, and 10 minutes on edits. The honest time entry is about 25 minutes, plus a couple of minutes assembling the instruction. If the firm bills in 0.1-hour increments, that is 0.5 hours. Billing one hour "for a demand letter, as it used to take" would be the phantom-hours problem. A fixed fee of, say, $175 for demand letters, stated in the engagement letter, is cleaner for everyone and does not require anyone to argue about minutes. For more routine drafting uses, see how small law firms use AI to draft letters.

A mistake to avoid: the invisible verification

A lawyer uses a tool to summarise a deposition transcript, skims the summary and bills 0.3 hours. Later the summary is found to have attributed a statement to the wrong witness. Nobody had checked it against the transcript. The firm absorbed the cost of correcting the error, and the client's trust suffered. The billing lesson is the same as the quality lesson: the review time is not a cost to minimise, it is the thing that makes the work reasonable to charge for. Keep a record of what was verified and how, for example against the source pages, so that the hours on the invoice and the work on the file match. The checking routine in stopping AI inventing case law applies to citations, and the same discipline works for summaries.

Time codes that make the honest bill easy

If your time system has one box labelled "drafting", people will put the AI's two minutes and their own thirty into it without thinking. A few extra codes turn honesty into a habit. Illustrative codes and the lease example from above:

Three months of this data answers questions that guesswork cannot: how long verification takes per kind of task, which tasks AI does not shorten at all, and where a fixed fee would be safe. It also gives you an answer if a client queries a bill, because the codes show where the time went.

Junior lawyers, trainees and supervision

A firm that bills a trainee's time at a lower rate faces a sharper version of the question. If a trainee uses AI to produce a first draft in 20 minutes that used to take three hours, the trainee's billable time is 20 minutes plus review, and the supervising lawyer's review time is billable at their own rate. Do not move the saving into the supervisor's line by inflating review time. Supervision matters here for a separate reason: the guidance discussed above treats lawyers as responsible for supervising how AI and staff are used, so a firm should write down who reviews what before the first AI-assisted matter. The cost of that review belongs in the fixed fee or the hourly bill, but it is real work, and a firm that skips it saves money on the invoice and pays for it in errors.

Illustration: a fixed-fee package for small-business clients

Say a firm wants to offer small-business clients a "contract health check" for $600. Behind the package: an intake form (10 minutes of the client's time), an AI-assisted review of up to three contracts of 20 pages each (about 30 minutes of lawyer attention per contract for verification), and a 30-minute call with a written summary. The firm's own time comes to about two hours; at a target of $250 an hour that is $500, so the package carries a small margin for the tail risk of a difficult contract. The engagement letter states: three contracts, maximum 20 pages each, one call, one follow-up round, extra contracts charged at a stated price. If a client sends a 90-page agreement, the scope clause lets the firm quote separately, instead of eating the overrun.

Notice what the package does not do: it does not mention hours at all, it does not say AI saves the client money, and it does not pretend the review is automatic. It sells an outcome, a described service and a limit on scope.

Answering a challenged bill

Suppose a client queries a $1,500 hourly bill for a research memo, saying "I thought AI did this in minutes". The firm's answer should be a short walk through the file: PREP 30 minutes, VERIFY 2 hours (checking each authority against its source, because tools can state authorities that do not say what is claimed), ADVISE 1.5 hours. That is 4 hours at a stated rate, with the verification showing exactly why the bill is not "minutes". If, looking at the file honestly, the VERIFY time was 20 minutes and the rest was padding, the right response is to reduce the bill. A firm that adjusts promptly when the record does not support the entry keeps clients, and one that defends a bill it cannot document does not. The codes make that conversation factual rather than defensive.

Checking your terms against your own rules

Before you change anything, put three questions to your regulator, bar association or professional indemnity insurer, depending on your jurisdiction:

• What do our fee rules say about the reasonableness of fees where technology reduces the time taken?
• Do we need informed consent, in writing, before client information goes into a particular type of AI tool?
• May we pass on a specialist tool's cost as a disbursement, and on what conditions?

Keep the answers on file, with the date. They matter more than any commentary, including this tutorial, because they come from the body that will judge a complaint. An insurer may also have views on AI use that affect your cover, so ask it as well.

A checklist for changing your fee model

1. List the ten tasks where AI changes the time most, and measure real time before and after on three matters each.
2. Decide, per task, whether it stays hourly, goes to a fixed fee or joins a retainer.
3. Set the fixed price from value, risk and the real work involved now; keeping the old flat fee for work AI has made much quicker is what Opinion 512 warns may be unreasonable.
4. Update the time-recording codes so prompting, review and judgement are visible.
5. Rewrite the engagement letter paragraph above and have your ethics adviser read it.
6. Decide the tool-cost policy: overhead by default, pass-through only by prior agreement.
7. Train everyone on the invoice-narrative style.
8. Review prices and codes after six months, with real data.

If you would like help planning the rollout across a small firm, that is what my AI implementation consultation covers, and the contact page is the way in. The first-90-days view of introducing AI in a firm is in the implementation plan for a small law firm, and the wider topic collects at the AI hub.

Billing questions law firms ask about AI

Can I bill the hours a task would have taken without AI?

No. ABA Formal Opinion 512 says lawyers who bill by the hour must bill their actual time. If AI cuts a four-hour task to one, the bill reflects the hour you spent, including entering the instructions and reviewing the output for accuracy and completeness, not the four it used to take.

Can I charge clients for the cost of an AI tool?

It depends on the tool. Opinion 512 treats tools that work like equipping the practice, such as an AI grammar checker in your word processor, as overhead unless you disclosed otherwise in advance. A third-party service charged per use for one client's matter can ordinarily be billed as an expense at actual cost. Check your own rules.

Is a flat fee safer for AI-assisted work?

It avoids arguments about minutes, but it is not a way round the reasonableness rule. Opinion 512 says that if AI lets you finish much faster, charging the same flat fee as without the tool may be unreasonable. Set the price from value, risk and the real work, agree it in advance and review it.

Do I need client consent to use AI at all?

Opinion 512 says informed consent is required before you put information relating to the representation into a self-learning AI tool, and that boiler-plate engagement-letter wording is not enough. Using a tool without client information, for idea generation say, needs no consent. Check your own jurisdiction's rules.

Should the invoice say AI was used?

It is good practice to describe the work done in words a client understands, and to record review time honestly. Mentioning AI in the narrative is a transparency choice unless your rules or the engagement terms require it. Be consistent.

Further reads

• What an AI Consultant Does for a Law Firm, and What It Costs — What an AI consultant does for a law firm.
• ChatGPT or a Legal AI Tool: Which Should a Small Firm Use? — A general chatbot or a legal AI tool?
• AI Client Intake for Law Firms: Qualify Enquiries Out of Hours — Qualify enquiries out of hours with AI.
• AI Readiness Checklist for Accountants, Solicitors, Consultants — A readiness check before rolling AI out.
• AI Engagement Letters for Accountants: Faster Drafts, Clear Scope — Engagement-letter scope wording, from the accountancy side.
• How Freelancers Use AI for Invoices, Proposals and Chasers — Invoice and proposal habits from a related profession.
• AI Legal Bill Review: Fix Time Narratives Before Invoices Go Out — A pre-bill routine for small firms: tidy narratives, screen for privileged detail, test client billing guidelines, then a partner signs off exceptions only.
• AI for Attorney Research: A Citation-Checking Workflow — A repeatable way to turn AI research suggestions into checked authorities before a draft reaches the court.
• AI for Bankruptcy Lawyers: Intake, Schedules and Document Chasing — A document-to-draft route for consumer bankruptcy matters, with missing-evidence checks before any client signs.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ABA Formal Opinion 512 on generative AI tools (29 July 2024), read in full from an archived copy of the ABA's PDF in November 2026, in particular its sections on confidentiality, communication, supervision and fees. Check your own regulator's rules before relying on any point here. This is not legal advice.