DIZON.LAW

Insights · Fees

Billing for AI Under Rule 1.5: What Reasonable-Fee Rules Allow

When AI cuts the hours a task takes, the reasonable-fee rules still govern; here is what the ABA, North Carolina, and Virginia say you may charge.

When a generative AI tool turns a three-hour drafting task into a forty-minute one, the firm faces a fee question older than the technology: what may you charge, and on what basis. Rule 1.5 has not changed, but AI puts pressure on the gap between the value of legal work and the time it consumes, and the bar authorities now addressing that gap deserve a careful read.

The rule has not moved, even if the tools have

Model Rule of Professional Conduct 1.5(a) prohibits an unreasonable fee and lists factors that bear on reasonableness, including the time and labor required, the novelty and difficulty of the questions involved, the skill requisite to perform the service properly, the fee customarily charged in the locality, the amount involved and results obtained, and the experience, reputation, and ability of the lawyer. Rule 1.5(b) requires that the scope of the representation and the basis or rate of the fee be communicated to the client, preferably in writing, before or within a reasonable time after commencing the representation.

Nothing in that framework singles out artificial intelligence. The point of reading the recent guidance is to see how settled principles apply when a tool compresses the hours a matter takes. Three opinions now anchor the discussion: ABA Formal Opinion 512 (July 29, 2024), North Carolina 2024 Formal Ethics Opinion 1 (adopted November 1, 2024), and Virginia Legal Ethics Opinion 1901, approved by the Supreme Court of Virginia on November 24, 2025. They agree on the hourly core and diverge, instructively, on flat fees.

Bill the time you actually spend, not the time you saved

The first principle is the least controversial and the most consequential for hourly billers. If you charge by the hour, you may bill only the time you actually spend. ABA Formal Opinion 512 is direct on this point: a lawyer who bills hourly must charge for the time actually expended, and efficiencies gained from a generative AI tool must be reflected in the bill rather than recaptured as phantom hours. North Carolina's 2024 Formal Ethics Opinion 1 states the same conclusion in plain terms, answering that a lawyer may not bill a client for three hours of work when only one hour was actually performed.

The temptation this addresses is specific. Suppose a tool drafts a serviceable first version of a brief section in minutes, and the lawyer then spends thirty minutes reviewing, correcting, and integrating it. The billable event is the time the lawyer actually worked, which here is the review and revision, not the three hours the task would once have required. Billing the historical time rather than the real time is the precise practice these opinions forbid. The efficiency belongs to the matter, and under an hourly arrangement that means it belongs to the client.

This does not mean AI-assisted hourly work shrinks to nothing. Verifying AI output is real legal work, and where the lawyer reads authorities the tool cited, confirms they exist and stand for what the draft claims, and exercises judgment about what to keep, that time is properly billed. The line is honesty about what occurred, not suspicion of the tool.

Do not bill the learning curve

A second principle follows from competence. Acquiring the general proficiency to use a tool you will deploy across many matters is overhead, not client work. ABA Formal Opinion 512 treats time spent learning to use a generative AI tool for the lawyer's general practice as a cost the lawyer absorbs, not one passed to whichever client happens to be the first matter where the tool is used. The reasoning tracks Rule 1.1 competence: keeping abreast of the benefits and risks of relevant technology is a professional obligation, and clients do not subsidize a lawyer becoming competent.

There is a narrow and sensible exception. Where a client specifically asks the lawyer to use a particular tool that requires the lawyer to come up to speed on it for that engagement, time learning that tool can be billable, provided the arrangement is disclosed and agreed in advance. The distinction is between general professional development, which is yours to bear, and client-directed work specific to the matter, which is theirs to pay for.

A short checklist for the learning-curve question:

  • General skill you will reuse across clients: not billable.
  • Tool the client specifically directed you to use for their matter: potentially billable if agreed in advance.
  • Time fixing AI errors that competent review should have caught: not separately billable as if it were productive work.

The third item matters. Time spent cleaning up a hallucinated citation or a garbled passage is not a value-adding service you bill on top; it is part of doing the work correctly, and padding the bill with it is the same problem under a different label.

Tool cost: overhead or disclosed expense

The cost of the software itself raises a separate question, and the answer turns on how the cost is structured. A general subscription or license fee that supports the firm's practice broadly looks like overhead, the same category as legal research databases, word processing, and office systems. ABA Formal Opinion 512 indicates that this kind of general access cost should ordinarily be treated as overhead and built into the fee rather than charged to a client as a line item.

Costs that are genuinely incurred for a specific matter are different. Where a tool charges on a usage or per-matter basis and that cost is directly attributable to one client's representation, it can be passed through as an expense, but only on the terms Rule 1.5 and Rule 1.4 require: the basis for the charge should be disclosed, the amount should be reasonable, and a client should not be charged more than the actual cost plus any reasonable allocation the firm has explained. North Carolina's 2024 Formal Ethics Opinion 1 recognizes both routes, allowing direct costs tied to a specific representation and a general administrative or technology charge that covers tools benefiting clients, so long as the charges are accurate, reasonable, and disclosed.

Two practical guardrails follow. First, decide whether a given AI cost is overhead or a pass-through expense before the bill goes out, and apply that classification consistently. Second, if you pass a cost through, make sure your engagement letter actually describes that category of charge. A client surprised by an AI line item they were never told about is a Rule 1.5(b) communication problem regardless of whether the dollar figure is reasonable.

Flat fees: the question the jurisdictions split on

Here the authorities part company, and the split is the most important thing for a firm to understand before it sets a flat fee for AI-assisted work. The core question: if a flat fee was historically priced against the hours a task took, and AI now cuts those hours, must the flat fee come down.

ABA Formal Opinion 512 leans toward the view that a fee can become unreasonable if it no longer corresponds to the work and value delivered, and it cautions lawyers against simply collecting the old number while doing far less. North Carolina's 2024 Formal Ethics Opinion 1 permits flat-fee arrangements for AI-assisted work but conditions them on the fee not being clearly excessive and on client consent to the structure, which keeps the reasonableness inquiry alive even outside hourly billing.

Virginia Legal Ethics Opinion 1901 takes the most permissive position of the three. It concludes that it is not per se unreasonable to charge the same flat or non-hourly fee for work done with the assistance of generative AI, and that a lawyer is not ethically required to discount a flat fee solely because AI accelerated the work. The opinion grounds this in Rule 1.5(a)'s other factors: the value delivered, the results obtained, and the skill required do not vanish because the clock ran shorter. Effective AI use can itself demand specialized skill to prompt, verify, supplement, and integrate the output, and that skill is part of what the client pays for. The opinion also flags a disclosure dimension under Rule 1.5(b), noting that a lawyer may need to explain why experience and technical skill justify the fee even though time spent has fallen.

The practical reading for a small or mid-sized firm is this. A flat fee is permissible and need not automatically shrink because you adopted a tool, but the fee still has to be reasonable on the full set of Rule 1.5(a) factors, not on time alone. The defensible posture is to price flat fees against the value and the skill the engagement actually requires, document that basis, and be candid with clients about how you work. Where you practice matters: Virginia's framing gives more room than the ABA's, and a firm operating across jurisdictions should price to the most demanding rule it is subject to rather than the most forgiving.

What this means for how you set up billing

The through-line across all three opinions is that AI changes the inputs to a fee, not the standard the fee must meet. Reasonableness, honesty about time, and clear communication still govern.

  • For hourly matters, bill real time, let the client keep the efficiency, and bill verification and judgment as the genuine work they are.
  • Treat general proficiency with your tools as overhead; bill learning only when a client directed the tool and agreed in advance.
  • Classify each AI cost as overhead or a disclosed pass-through, and make the engagement letter match.
  • For flat fees, anchor the price to value and required skill, confirm it is not excessive, and document the basis, watching the jurisdictional split between the ABA's caution and Virginia LEO 1901's latitude.
  • Tell clients how you bill for AI-assisted work before the first invoice, not after a dispute.

Billing for AI legal work is less a new body of law than a stress test of an old one. Firms that write down their approach to Rule 1.5 AI fees now, and that can explain a flat fee AI arrangement in terms of value rather than a stopwatch, will be the ones whose bills hold up.

This is general information for lawyers and law-firm leaders, not legal advice, and it does not create an attorney-client relationship. The authorities are cited so you can read them yourself.

The longer argument continues in AI in the Defender’s Office, a national field guide now in production.

Read about the book