AI Ethics
How to Bill Clients for AI-Assisted Legal Work
80% of corporate legal executives expect lower bills because of AI, but only 9% of law firms have heard the ask. That gap won’t close gently.
Computational scientist with a Ph.D. from UC Irvine and peer-reviewed research in NLP, deep learning, and large-scale data modeling. Over a decade of experience building systems that process complex document sets at scale. Founded Hintyr to bring defensible AI workflows to litigation teams navigating document review, redaction, and production.
The Gap Nobody’s Talking About
A pricing disconnect hiding in plain sight
Here’s a number that should keep managing partners awake. According to the LexisNexis 2024 Investing in Legal Innovation Survey, 80% of corporate legal executives expect their outside counsel bills to drop because of AI. On the other side of that relationship, only 9% of law firm leaders report hearing this expectation from clients. The question of how to bill for AI legal work has no consensus answer yet, but the ethics rules and client expectations are converging fast.
That is not a communication gap. It’s a pricing collision that hasn’t happened yet.
The clients aren’t wrong to expect savings. AI tools cut routine legal tasks by 30-70% in controlled studies. A contract review that took six hours now takes two. A research memo that billed at four hours comes back in forty minutes. If you’re a general counsel watching your department budget, the math writes itself.
But the firms aren’t deaf, either. They’re stuck. The billable hour remains the primary billing model at 67% of law firms, according to the ABA’s 2024 TechReport, and the entire economic model of BigLaw, from associate compensation to partner draws, runs on hours billed. AI doesn’t just make lawyers faster. It makes the dominant pricing model of the last fifty years structurally unstable.
And the pressure is coming whether firms are ready or not. The ACC/Everlaw 2025 report found that 61% of in-house counsel plan to push for pricing changes tied to AI efficiency. Mark Smolik, General Counsel at DHL, put it bluntly at a recent industry summit: “We are done waiting.” Meanwhile, the Georgetown/Thomson Reuters 2025 Report on the State of the Legal Market concluded that “continued reliance on an inputs-driven model is simply not viable in the long term,” even as roughly 90% of legal fee revenue still flows through hourly billing, per the same report’s analysis of Thomson Reuters Legal Tracker data.
If you’re already using AI in your practice, or even considering it, this tension will find you. The ethics rules are already written. The client expectations are already set. The only question is whether your billing model will catch up before your clients force the issue.
This post covers the ethics, the models, and the engagement letter language you’ll need. No jargon. No hedge. Just what seven jurisdictions, the ABA, and the latest industry data actually say about billing for AI-assisted work.
What the Ethics Opinions Actually Agree On
The emerging consensus, and where it fractures
Start with what’s settled.
Seven jurisdictions and the ABA have now issued formal ethics guidance on AI billing: ABA Formal Opinion 512 (2024), Texas Opinion 705 (2025), Oregon Formal Ethics Opinion 2025-205, Fla. Bar Ethics Opinion 24-1 (2024), Virginia Legal Ethics Opinion 1901 (2025), D.C. Bar Legal Ethics Opinion 388 (2024), and North Carolina 2024 Formal Ethics Opinion 1. Washington State Advisory Opinion 202505 and Vermont (both 2025) have added their voices since.
On hourly billing, they all say the same thing. You bill for actual time worked. Period. If AI turns a four-hour task into a forty-minute task, you bill for forty minutes. ABA Opinion 512 quotes ABA Formal Opinion 93-379, the foundational billing ethics authority from 1993: “The lawyer who has agreed to bill on the basis of hours expended does not fulfill her ethical duty if she bills the client for more time than was actually expended on the matter.” Texas Opinion 705 echoes it. So does D.C. Opinion 388, with an analogy that cuts deep: a lawyer who reuses a research memo for a second client can’t bill the second client for the full original research time. “The same is true when the use of GAI reduces billable time and the lawyer’s fee agreement with the client is based exclusively on the time the lawyer spends working on the matter.”
You also can’t bill clients for learning how to use AI. ABA Opinion 512 is direct: “in most circumstances, the lawyer cannot charge a client for learning how to work a GAI tool.” Oregon agrees. So does every other jurisdiction that has addressed the question. (For more on the 300+ court disclosure rules that layer on top of these ethics opinions, we’ve covered that separately.)
But here’s where the consensus fractures. On non-hourly billing, specifically whether firms can charge the same flat fee for AI-assisted work as for manual work, Virginia breaks from the pack. Virginia LEO 1901, adopted by the Supreme Court of Virginia, holds that “it is not per se unreasonable for a lawyer to charge the same non-hourly fee for work done with the assistance of AI as work done without the use of AI.” The opinion goes further: “Rule 1.5 does not equate reduced time with proportionally reduced fees.” It argues that effective AI use may actually increase the skill factor, because it takes specialized knowledge to prompt, verify, and integrate AI outputs.
The ABA reads it differently. Opinion 512 warns that “it may be unreasonable under Rule 1.5 for a lawyer to charge the same flat fee when using a GAI tool as when not using it if the GAI tool enables the lawyer to complete tasks much more quickly.”
That split matters. If you’re pricing non-hourly work and your jurisdiction hasn’t weighed in yet, you’re choosing between two plausible readings of Rule 1.5. Virginia says the fee can reflect the value of the result. The ABA says even a flat fee may be unreasonable if AI has drastically reduced the effort required. Neither position requires you to track hours on a flat-fee engagement. But your engagement letter needs to reflect whichever reading you adopt, and why.
Who Pays for the Software?
Cost recovery and the overhead question
Even where the billing model is settled, firms still trip on a narrower question: can you charge clients for the AI tool itself?
The answer depends on how you pay for it.
If your firm uses a per-use AI service, where the vendor charges by query, by document, or by matter, most jurisdictions let you pass that cost through. Texas Opinion 705 draws the analogy directly: per-use AI fees can be billed “in much the same way some clients agree to pay for the use of traditional online research tools like Westlaw and LexisNexis.” ABA Opinion 512 agrees, as long as the cost is disclosed and reasonable.
If your firm pays a flat subscription, the analysis flips. Oregon Opinion 2025-205 is the clearest: “Lawyers who are unable to determine the actual cost associated with a particular client’s matter may not ethically prorate the periodic charges of a specialized AI, including GAI, and instead must account for those charges as overhead.” Translation: your $500/month CoCounsel subscription or your $1,000/month Harvey AI license? That’s overhead. You absorb it, the same way you absorb Westlaw flat-rate contracts. And remember Waggoner v. Chadbourne: a client successfully sued a firm for billing Westlaw at retail transactional rates while holding a flat-rate subscription. The same logic applies to AI tools.
ABA Opinion 512 sorts it neatly. AI embedded in your word processor (grammar checking, formatting) is overhead. A third-party AI service used to review thousands of contracts for a specific client, charged per use, is a billable expense. Everything in between requires a judgment call and, ideally, a line in your engagement letter.
The Harvard Center for the Legal Profession found that none of the ten AmLaw 100 firms they studied plan to pursue direct AI cost recovery from clients. The prevailing strategy at the top of the market is to absorb AI costs as an investment in efficiency and competitiveness. Smaller firms will need to make their own calculation.
Five Models for Billing AI Legal Work
Matching the model to the work
So if the billable hour penalizes efficiency and the ethics rules say you can’t pad time, what do you actually do? Five models, each with a different risk profile.
| Model | Ethics | Revenue impact | Client preference | Who captures savings | Best for |
|---|---|---|---|---|---|
| Hourly + disclosure | Universal (ABA 512, all states) | Negative: shrinks as AI accelerates tasks | Low: clients see legacy pricing | Neither: savings reduce billable hours | Novel, bespoke litigation |
| Flat-fee | Strong (VA LEO 1901, FL 24-1, D.C. 388) | Positive: speed becomes margin | High: 71% prefer flat fees (Clio 2025) | Firm retains savings as profit | High-volume: contracts, compliance |
| Direct cost recovery | Weak-moderate (OR 2025-205) | Neutral: marginal revenue, high overhead | Low: clients view AI as overhead | Client pays; firm breaks even | Large-scale e-discovery |
| Value-based | Strong (Rule 1.5 factors; VA LEO 1901) | Strongly positive: better tools, higher fees | Medium: complex to negotiate | Shared or negotiable | High-stakes M&A, regulatory, patent |
| Hybrid (flat + hourly) | Strong: each component supported | Positive: efficiency on routine, hourly on complex | High: top satisfaction (ACC 2024) | Progressive: migrates to flat-fee over time | Full-lifecycle litigation |
1. Hourly billing with AI disclosure. The path of least resistance. You keep billing hourly, but you disclose AI use and bill only for actual time. Your time entry says “0.3 hrs, review and revise AI-generated draft contract” instead of the three hours it would have taken by hand. Every ethics opinion supports this model. But it’s also the model that shrinks your revenue as your tools improve. The Clio Legal Trends Report 2025 estimates that 74% of hourly billable tasks could be automated with AI, putting roughly $27,000 of annual revenue per lawyer at risk.
The inputs-driven model is not viable long-term. If you’re picking this model, do it as a transition, not a destination.
2. Flat-fee billing. Price the outcome, not the hours. AI makes you faster, and under a flat fee, that speed becomes profit instead of a write-down. Clio data shows that 71% of legal consumers prefer flat fees, flat-fee matters close 2.6 times faster, and firms collect payments nearly twice as fast. Virginia LEO 1901 explicitly supports charging the same flat fee for AI-assisted work as for manual work; Florida encourages the model outright. The risk? ABA Opinion 512’s warning that flat fees can still be “unreasonable” if AI has drastically reduced the effort. Set your flat fees using historical data that reflects AI-driven efficiency, and document the value delivered.
3. Direct cost recovery. Treat the AI tool like a research database charge. Per-use costs get passed through as line items, with disclosure in the engagement letter. This works for per-query or per-document AI services. It does not work for flat-rate subscriptions (Oregon, 2025-205). And it won’t generate meaningful revenue. It’s a cost-neutral model, not a growth model.
4. Value-based billing. This is where Virginia’s opinion matters most. If “Rule 1.5 does not equate reduced time with proportionally reduced fees,” then the fee can reflect expertise, outcomes, and the specialized skill of working with AI effectively. The model works best for high-stakes, outcome-dependent work where clients care about results, not timesheets. It requires more sophisticated scoping. But it’s the only model where better tools lead to higher, not lower, fees.
The practical challenge with value-based billing is documentation: you need to show clients what the AI contributed and how your judgment refined it. AI review tools that produce outputs with traceable citations back to source documents make this easier. (Hintyr’s document review agents are built around this principle, though it applies regardless of which platform you use.) That documentation is what value-based billing demands.
5. Hybrid models. The realistic option for most firms. Flat fees for routine, AI-automatable tasks: document drafting, compliance filings, due diligence. Hourly billing for complex work: negotiations, novel strategy, court appearances. BigHand’s 2025 data found that 100% of surveyed firms report AI is affecting their pricing, yet only 34% have updated their models. The hybrid lets you start the transition without a full overhaul. Most firms will start here.
If you’re looking for a recommendation: flat-fee or hybrid billing, paired with value-based pricing for high-stakes work, is where the economics and the ethics point. Value-based billing is the only model where better tools lead to higher, not lower, fees. Virginia LEO 1901 gives you the legal cover. The client data gives you the business case.
What Clients Already Know (and What They’re Planning)
The demand side of the equation
Firms can debate billing models. Clients have already made up their minds.
The ACC/Everlaw 2025 report, surveying 657 in-house legal professionals across 30 countries, found that 61% plan to push for pricing changes tied to AI. Of those whose firms use generative AI, 59% have seen “no noticeable savings yet,” with 58% blaming law firms for not adjusting pricing. And 59% of in-house professionals don’t know whether their outside counsel uses AI at all.
That last number is the one that should worry you most. Your biggest clients may already be evaluating your firm’s AI posture without telling you. So what happens when they find out you’ve been using AI for six months and never mentioned it? Deloitte’s 2025 report captured the shift: clients are no longer asking whether AI is being used; they are expecting it.
The Clio 2025 data adds texture. Among firms with heavy AI adoption, 45% have adjusted their pricing. Of those, 25% increased fees (presumably through value-based models), 11% reduced fees, and 8% added AI-specific fee line items. Firms with wide AI adoption are nearly three times more likely to report revenue growth than those without it. The data doesn’t support the fear that AI necessarily means lower revenue. It supports the conclusion that the right billing model matters more than whether you use AI.
And the market is voting with real money. Eudia, a $105 million AI-native law firm built entirely on fixed-fee billing, has attracted validation from the general counsels of Duracell and DHL. The model works at scale and for sophisticated clients because it aligns firm incentives with client outcomes.
None of this means the billable hour disappears overnight. It won’t. But the firms that treat AI billing as a pricing strategy rather than an accounting adjustment will capture the value. The firms that don’t will watch their most sophisticated clients leave for firms, or AI-native competitors, that are already pricing differently. The disconnect between 80% and 9% will close. The only question is whether your firm is on the right side when it does.
The Engagement Letter
Put it in writing before the client asks
Every billing model discussion eventually comes down to one document: the engagement letter.
ABA Opinion 512 warns that “boilerplate provisions in engagement letters” won’t satisfy the informed consent requirement for AI use. Florida Opinion 24-1 recommends obtaining consent before using any third-party generative AI with confidential information. Texas Opinion 705 requires billing transparency when AI reduces time on a task. Oregon mandates that lawyers inform clients, preferably in writing, of any intent to charge for AI costs.
So what should your engagement letter actually say?
At a minimum, disclose that your firm uses AI tools in its practice and specify how you handle confidentiality. Name the categories of AI use (research, drafting, document review) without overcommitting to specific products that may change. If client data enters any AI system, disclose the data handling practices and obtain informed consent. “We use secure, enterprise AI tools with no model training on client data” is a sentence worth including. Boilerplate won’t do. ABA Rule 1.4 requires you to consult with the client about the means used to accomplish their objectives. AI counts.
The harder provisions are the billing-specific ones. Specify your model for AI-assisted work. If you’re billing hourly, confirm that time entries reflect actual time spent. If you’re billing a flat fee, explain that the fee reflects the value and complexity of the work, consistent with Rule 1.5’s eight factors. If you’re passing through per-use AI costs, state the basis and provide an estimate. A working clause: “Fees for this engagement are based on the value, complexity, and risk of the matter. Our firm uses AI-assisted tools as part of its workflow; all AI outputs are reviewed and verified by a licensed attorney. Per-use AI processing costs, if any, will be itemized separately at actual cost.”
Then address the supervision obligation under Rules 5.1 and 5.3. Confirm that all AI outputs are reviewed by a licensed attorney before use and that every citation is independently verified. This isn’t just good ethics. It’s your best defense if a billing dispute ever reaches a disciplinary committee, and it aligns your engagement letter with the verification standard that ABA Opinion 512 and every state opinion demand.
Draft these provisions now. Not after a client raises the issue. Not after a billing dispute. The firms that embed AI billing transparency into their engagement letters today will be the ones that avoid the painful retroactive conversations tomorrow.
For more on the ethical risks of AI in practice and the emerging disclosure requirements, we’ve covered both topics in depth.
Where This Goes
The profession's pricing reckoning
The billable hour isn’t dying. But its monopoly is over.
For fifty years, time-based billing survived every challenge: alternative fee arrangements, client pushback, consultant critiques. It survived because nothing forced the issue. AI forces the issue. When a tool can replicate hours of associate work in minutes, billing by the hour stops being a pricing model and starts being a subsidy for inefficiency. Clients know this. The data says so.
What’s new is the strategic dimension. Firms that move to value-based or hybrid models will capture AI’s efficiency gains as profit. Firms that stay on hourly billing by default will watch those gains flow to clients, or to competitors who moved to new models sooner.
The 80/9 gap will close. The engagement letter conversation is coming. The only real question is whether you write the terms or your clients do.
Frequently Asked Questions
Can lawyers bill clients for hours “saved” by AI?
No. Every jurisdiction that has addressed the question agrees: if your fee arrangement is hourly, you bill for actual time worked. ABA Formal Opinion 512, Texas Opinion 705, D.C. Opinion 388, Oregon Opinion 2025-205, Florida Opinion 24-1, North Carolina FEO 2024-1, and Virginia LEO 1901 all prohibit billing for time not actually spent, regardless of how productive the AI output was.
Can law firms charge clients for AI software costs?
It depends on the pricing model. Per-use AI charges (billed per query or per document) can generally be passed through to clients with advance disclosure and consent. Flat-rate AI subscriptions cannot be prorated across clients and must be treated as firm overhead, per Oregon Opinion 2025-205 and ABA guidance. Firms should specify their cost recovery approach in the engagement letter.
Is it ethical to charge the same flat fee for AI-assisted work?
Jurisdictions disagree. Virginia LEO 1901, adopted by the Supreme Court of Virginia, holds that “it is not per se unreasonable for a lawyer to charge the same non-hourly fee for work done with the assistance of AI as work done without the use of AI.” ABA Opinion 512 takes a more cautious position, warning that flat fees may be unreasonable if AI has drastically reduced the effort required. Consult your jurisdiction’s guidance.
Do lawyers have to tell clients they use AI?
In most circumstances, yes. ABA Rule 1.4 requires lawyers to consult with clients about the means used to accomplish their objectives. ABA Opinion 512 recommends disclosing AI use in engagement letters, particularly when AI is relevant to the basis of fees or processes confidential client data. Florida Opinion 24-1 and Texas Opinion 705 reinforce this. Several jurisdictions now also require AI disclosure in court filings under standing orders.
Which billing model is best for firms using AI?
There is no single best model. Flat-fee and value-based billing let firms capture AI efficiency gains as profit rather than write-downs, and Virginia LEO 1901 provides the strongest ethical support for value-based pricing. Hybrid models, combining flat fees for routine AI-automatable tasks with hourly billing for complex work, offer the most practical transition path for most firms. The Clio 2025 report found that firms with wide AI adoption are nearly three times more likely to report revenue growth, regardless of billing model.
Disclaimer: This blog post is published by Hintyr for informational purposes only and does not constitute legal advice. The discussion of ethics rules, bar opinions, and billing practices is general in nature and may not reflect the rules applicable in your jurisdiction. Ethics opinions cited here are advisory and non-binding unless otherwise noted; Virginia LEO 1901 is an exception, having been adopted by the Supreme Court of Virginia. Attorneys should consult their state bar’s ethics opinions and qualified legal counsel before making billing and pricing decisions. No attorney-client relationship is created by reading this post.