Introduction

The past decade has seen a great increase in the accessibility and widespread use of artificial intelligence. Not only are people using artificial intelligence (“AI”) in their everyday lives to draft emails, brainstorm gift ideas, or find baking recipes, but the use of artificial intelligence tools is unsurprisingly seeping into the legal profession. Research companies like LexisNexis and Westlaw have produced their own versions of legal artificial intelligence to help with research efficiency. Lawyers and law firms are integrating generative artificial intelligence to assist with drafting, research, and even discovery. This new use of artificial intelligence has its advantages, such as time-efficiency and cost-effectiveness, but the integration of artificial intelligence into the legal profession is not without its hardships. Those hardships typically include the high cost of AI models and programs, threats to client confidentiality, poor reliability, and adverse effects on hiring and training structures. This blog will analyze the integration of artificial intelligence into the legal practice and address the ethical dilemmas faced by lawyers using artificial intelligence.

Why Do Lawyers Use Artificial Intelligence?

Generative artificial intelligence’s (“Generative AI”) improvements to grammar and sentence structure have made it useful for drafting various legal documents. It can be used to summarize discovery materials and for writing complaints or motions to the court. Artificial intelligence can also be used for drafting various types of contracts or even drafting emails to clients or opposing counsel.[1] Many firms are promoting the use of artificial intelligence and approving its use for drafting inter-office materials, such as emails, memorandums, and summaries.

Specific AI programs designed for legal practice are also beginning to develop. Legal artificial intelligence has the capacity to greatly decrease the time needed for legal research, especially in cases where research questions can span multiple jurisdictions or practice areas.[2] When lawyers conduct research, they typically are conducting numerous searches to try out different search terms or Boolean instructions. Such searches produce a list of cases where each case must be individually read and analyzed to determine the rule of law and factual similarities to the client’s case. Legal artificial intelligence programs are becoming increasingly sophisticated to increase the number of cases, secondary sources, and statutory materials that are reviewed for each research prompt. These AI models can answer a research question within minutes and provide summaries of the state of the law, on-point cases, and nuanced aspects of the doctrine. Legal artificial intelligence is capable of accomplishing these tasks much faster than many attorneys could.[3] The lengthy step-by-step legal research process done by attorneys is easily reduced to the single step of inputting the research question into a legal artificial intelligence tool.[4]

Previous Ethical Violations Using Artificial Intelligence

While both generative AI and legal artificial intelligence tools can be efficient, its use raises ethical concerns to client confidentiality, reliability, attorney hiring and training, and billing. Thus far, some lawyers using artificial intelligence have already run afoul of the ethical and professional rules. A database maintained by Damien Charlotin “tracks legal decisions in cases where generative AI produced hallucinated content[,]” which has found over 700 cases in just the United States.[5] The most well-known examples are cases of lawyers using artificial intelligence that produces fake cases due to what is known as AI hallucinations. AI hallucinations are where “the AI has confidently generated information that is false or nonexistent.”[6] For lawyers, these hallucinations often take the form of “phantom cases and courts, invented quotations or attribution, and misstated holdings or law.”[7]

The case Mata v. Avianca is the best example of where reliability issues with artificial intelligence tools can get lawyers into trouble for violating ethical and professional rules.[8] In Mata v. Avianca, an attorney, Steven A. Schwartz of Levidow, Levidow & Oberman P.C., used AI to research and draft an affirmation opposing a motion to dismiss.[9] This motion was then filed with citations to cases which did not exist, because [10] ChatGPT fabricated multiple cases when helping draft  the motion.[11] As a result of the lawsuit, the court ordered [Peter LoDuca, Steven A. Schwartz, and Levidow, Levidow & Oberman P.C.] to send letters  to “each judge falsely identified as the author of the fake” cases cited to in the motion and assigned a $5,000 penalty.[12] Similar situations have occurred in other courts where lawyers face sanctions for filing motions that include fake cases due to AI hallucinations. These reliability issues are often the heart of the artificial intelligence debate. Lawyers could face sanctions including monetary fines or even disbarment or even losing their reputation.

The Billable Hour Model

The current state of most billing for law firms is billing by the hour, also known as the billable hour. Bills are determined by time rather than expensing per end product. The billable hour model was “[i]nitially introduced to ensure fairness and accountability.”[13] “This model, in theory, ensures that clients pay for the exact amount of work done, allowing for a detailed breakdown of services renders, and ensures that lawyers are compensated for every minute they work.”[14]

Tasks in legal practice often require great lengths of time and review by multiple levels of attorneys ranging from associates to partners. This process racks up costly bills for clients and generates high revenues for law firms. “Under the billable-hour system, a law firm is able to generate additional profits by increasing the hourly rates that are charged to clients, increasing the billable hours of the attorney, leveraging other attorneys and staff who are billed out at rates above the firm’s allocable fixed and variable costs, or a combination of these methods.[15] This method of billing has been the status quo since around 1975.[16]

The Effect on the Billable Hour Model

Clients are charged high rates per hour depending on both the seniority and ability of the attorney producing the work, but what happens when the work is a product of both an attorney and an artificial intelligence tool? With the increased productivity of attorney work due to the use of artificial intelligence, there are concerns that the billable hour method will be forced to change. If artificial intelligence is rapidly implemented into legal practices and time for tasks are reduced, the hours typically billed for a project would be reduced as well. Law firms generate a lot of revenue from time-consuming tasks like: document review, legal research, drafting, and writing. Typically, these tasks are done by associates. The first drafts are then circulated upwards in the chain of command for partner review and edits. This kind of automation challenges the billable-hour model.

The reason the billable-hour model has remained the status quo across the legal industry is due to its high-revenue generating nature. However, adding artificial intelligence into the picture may have the effect of reducing time but not increasing revenue. The more hours worked in a firm, the more revenue. Artificial intelligence’s widespread use in a firm will reduce the hours worked but there are no other changes to help offset a loss in revenue. Further, retaining legal artificial intelligence tools that will protect confidentiality will require costs that will not be offset either.

The implementation of AI into legal work has already begun in firms. Since the implementation of artificial intelligence, some studies “estimate[] that at large law firms only four percent (4%) of the lawyers’ time is used for document review[, but t]he rest is outsourced or done by artificial intelligence.”[17] “Much of the work performed by a new associate in many firms, large or small, was composed of these types of activit[ies] now being performed by artificial intelligence.”[18] Not only are the substantive assignments and work allocation affected by the widespread use of artificial intelligence within a firm, the attorney compensation model can also be affected. For associates and partners at a firm, end of year bonuses are often based on the billable hours produced per person. This model often incentivizes attorneys to bill more hours, which in turn produces more revenue for the firm. Attorney compensation has been increasing each year across the market. However, using artificial intelligence tools instead of paying a yearly salary to a junior associate may be more cost-effective for law firms.[19]

Aside from inside of the firm, this integration of artificial intelligence into law firms also affects the bills going outside the firms to clients. Clients expect high quality products for the costly bills they pay, however, if there is a chance to receive the same quality product at a cheaper price, then clients will obviously choose the cheaper option. Many clients are actively pressuring firms to use artificial intelligence to reduce their bills.[20] Refusing to use artificial intelligence may cause firms to lose clients, therefore reducing revenue. However, using artificial intelligence can also reduce revenue under the billable-hour model. So, the choice to integrate AI may be a lose-lose situation for law firms at first glance.

How Should Lawyers Bill When Using Artificial Intelligence?

The American Bar Association (“ABA”)’s rule on fees requires that fees are not unreasonable.[21] Reasonability of fees depends on a variety of factors, including time, labor, deadlines, and results.[22] The ABA has discussed these same issues about the inherent incompatibility between increasing efficiency of legal practice with artificial intelligence and billing by the hours.[23] Even when lawyers are working faster and more efficiently, “lawyers who bill clients an hourly rate for time spent on a matter must bill for their actual time.”[24] Lawyers are ethically bound to not bill more than the amount of time actually spent on work, but the use of artificial intelligence tools may make them too efficient to continue billing by the hour.

One alternative to the billable hour model is flat fees. Changing to more of a flat-fee structure has already begun, as “law firms are charging 34% more of their cases on a flat-fee basis compared to 2016.”[25] Charging per task, like drafting a motion or conducting due diligence, allows firms to take into account time spent on every aspect. It would turn law into an a la carte cost system.

Even when charging flat fees, the dilemma of billing when using artificial intelligence is still not solved. The ethical standard for billing is reasonableness. As the ABA described in Formal Opinion 512, charging the same fee for an assignment completed by one lawyer who used no artificial intelligence versus a lawyer who used artificial intelligence and completed it half the time may not be reasonable.[26] This is true especially if the flat fee is for an assignment done almost entirely by artificial intelligence tool, not the lawyer.[27]

However, this may lead to a less transparent cost model. When charging based on time, attorneys include descriptions for how that time was spent and keep a detailed log of .10 or .25 hour intervals. If firms change to this flat-fee model, it may be more difficult for clients to understand where the value of each task comes from, whether the value is from time spent researching, reviewing, etc. Further, time-based billing allows clients to see the time spent by associates and partners. Whereas a flat-fee system would make it more difficult for clients to understand and ensure that more experienced attorneys are spending enough on their task. Under the billable hour model, the way to increase firm revenue is increasing the time spent on legal work. However, under the flat rate system, attorneys are then incentivized to complete as many tasks as possible to increase firm revenue, rather than to spend time on ensuring perfection.

An extreme solution could be outlawing artificial intelligence categorically from the practice of law. While the ABA produces a set of model rules that most jurisdictions adopt to govern their state bar, each state bar association has the power to create rules for their attorneys. If a bar association wanted to, they could adopt a rule to ban any use of artificial intelligence in the legal practice within their jurisdiction. This solution may be seen as extreme by many, but it is possible. Some states, especially those who generate lots of tax revenue and capital from big law firms, may be incentivized to ensure the legal practices’ billable-hour method keeps law firms prosperous. Given the previous issues with reliability of work product and client confidentiality issues from artificial intelligence tools, state bar associations likely have a good faith basis to believe artificial intelligence is unsuitable for the legal professional, at least in its current state. If a jurisdiction adopts this categorical ban on AI tools, the topic may need to be revisited later in time if the reliability issues improve and artificial intelligence becomes more widely accepted. If one state bar association adopts a stance against AI, it may cause firms to leave their state. Lawyers may face pressure from clients and could choose to practice in another state where they are able to use artificial intelligence in their legal practice. Or, if one state adopts such a rule banning AI use altogether, it could also cause a domino effect of more jurisdictions to follow.

Will Artificial Intelligence Make Legal Work Too Efficient to Bill by the Hour?

The exact direction of the legal profession after the continued growth of artificial intelligence is uncertain. However, there will have to be changes. The use of artificial intelligence in the legal profession has already begun and shows no signs of stopping. Its use in legal practice however is incompatible with the status quo of billing model in most firms. A shift to a flat-fee based system is likely the next move within the legal industry. Categorically banning the use of artificial intelligence, while possible, is unlikely at this time. So, law firms will have to adapt to the changing AI landscape.

Law firms are businesses at their core, driven by goals of expansion, reputation, and revenue. Law firms generate their revenue by serving their clients. Clients, however, are pushing for artificial intelligence to be used on their cases in hopes it will decrease their bills. This pressure will not go away if firms decide not to use artificial intelligence, but the clients still champion it. To keep clients, firms will have to use artificial intelligence which will naturally decrease their hours and therefore profits. Switching to a flat-fee cost structure could allow firms to use the artificial intelligence tools their clients want while also maintaining the revenue and expansion goals. This solution, however, is not perfect, as the flat fee billing does not entirely solve the problem of how to manage the standard of reasonableness when billing for work done with AI.

 

---

[1]  Morris Wilner, Artificial Intelligence and Its Usage in the Business and Practice of Law, 50 W. St. L. Rev. 125, 132-35 (2023).

[2]  Id. at 145.

[3]   William J. Connell, Artificial Intelligence in the Legal Profession – What You Might Want to Know, R.I.B.J., May/June 2018, at 7 (“[One AI model] purports to review legal documents and reduce manual reviewing error rates, reducing the average time for legal review to less than 20 minutes, and increasing legal review accuracy by twenty percent….”).

[4]  Wilner, supra note 1 at 145 (“Other AI software programs can benefit law firms by providing automation in certain tasks, such as with Cerenade, and with client information.”).

[5]   Damien Charlotin, AI Hallucination Cases, https://www.damiencharlotin.com/hallucinations/?q=&sort_by=-date&states=USA&period_idx=0 (last visited Mar. 11, 2026).

[6]   Brent J. Hoeft, When AI ‘Lies’: The Legal Risks of Hallucinations, 98-SEP Wis. Law. 39, 40 (2025).

[7]   Id.

[8]  Mata v. Avianca, 678 F.Supp. 3d 443 (S.D.N.Y. 2023) (sanctions order).

[9]  Id. at 4.

[10]  Id. at 5.

[11]  Id. at 6.

[12]  Id. at 34.

[13]  Chuong M. Le, The Billable Hour Mandate Holding Us Back?, Colo. Law., June 2024, at 13.

[14]   Id.

[15]  Stuart L. Pardau, Bill, Baby, Bill: How the Billable Hour Emerged As the Primary Method of Attorney Fee Generation and Why Early Reports of Its Demise May Be Greatly Exaggerated, 50 Idaho L. Rev. 1, 5-6 (2013).

[16]  Id. at 3 (“After the Supreme Court’s 1975 decision in Goldfarb v. Virginia State, which held minimum fees schedules violated antitrust law, the billable hour method (which was already growing in popularity) became the predominant method of compensation.”).

[17]  Connell, supra note 3 at 5.

[18]  Id. at 6.

[19]  Id. (“The costs of A.I. can be much cheaper than hiring and training an associate.”).

[20]  Steven Lerner, Clients Are Driving the AI Revolution in Law Firms, Law 360 Pulse (Oct. 22, 2025), https://www.law360.com/pulse/articles/2402413/clients-are-driving-the-ai-revolution-in-law-firms (“The push for more widespread use of generative artificial intelligence in law firms is increasingly coming from clients.”).

[21]  Model Rules of Prof’l Conduct R. 1.5 (2020).

[22]  Id.

[23]  ABA Comm. On Ethics & Pro. Resp., Formal Op. 512 (2024).

[24]  Id.

[25]  Robinjit J. Eagleson, The Tech Age is Here. Embrace It, 104-Jun Michigan B. J. 42, 43 (2025).

[26]  ABA Comm. On Ethics & Pro. Resp., Formal Op. 512 (2024).

[27]  Id.