AI in Contract Negotiation: Why the Attorney Becomes More Important, Not Less
Updated: 4 hours ago
By Trenton Roberts and Carson McGurk
It is natural to ask whether software will soon negotiate contracts on its own. Anyone who has watched a generative AI tool produce a credible first draft of an NDA in a few seconds has wondered. It is a fair question. The drafting part of contract work is changing, and it will keep changing. But the more we think about where AI in contract negotiation is heading, the more convinced we are that the lawyer's role is moving toward the center of the deal, not away from it.
This is a forward-looking piece. It is our view of where the work is going, grounded in what the ethics authorities have already said about how lawyers may use these tools.
What the tools are getting good at
Start with what is plainly true. Generative AI tools can summarize long agreements, compare a draft against a standard form, flag missing provisions, propose alternative language, and produce a first pass at a markup. For high-volume, repeatable documents, the kind of work we describe in our post on high-volume NDA review, those capabilities will keep pulling time out of the mechanical parts of the job.
It is reasonable to expect that over the next several years, both sides of many routine negotiations will use software to produce their first turns. A buyer's tool will mark up a seller's form against the buyer's playbook, the seller's tool will respond against its own, and the issues list that reaches the lawyers will be shorter and sharper than it is today.
That future sounds like less lawyering. We think it is the opposite.
Why AI in contract negotiation does not replace judgment
A contract negotiation is not a document exercise. It is a decision process about risk, money, and relationships, conducted under uncertainty, between people who will have to live with the outcome. Drafting is the record of those decisions. When the drafting gets faster, the decisions do not get easier. They get exposed.
Consider what actually decides a hard point in a negotiation:
Risk allocation. Whether a seller should accept a fundamental representation with a longer survival period depends on what the seller knows about its own business, what the buyer's diligence found, and what insurance is available. No playbook can know which of those facts matter most in a given deal.
Client context. A company that needs this customer to close a financing round will trade different points than a company that has three competing offers. The right answer on a liability cap depends on the client's strategy, not on market statistics.
Relationships. Many commercial agreements are the start of a long relationship. A lawyer who wins every point in a supply agreement and damages the relationship has not served the client.
Timing and sequencing. Knowing when to concede, when to hold, and when to escalate to the principals is judgment that comes from reading people, not from reading text.
Accountability. When a provision fails in a dispute, someone has to answer for it. Clients want that someone to be a licensed professional who owes them duties.
As the tools take on more of the document work, these are the parts of the job that remain, and they are the parts that were always the most valuable.
The ethics rules already say this
The organized bar has not waited to address these questions. Two opinions set the frame for Texas lawyers.
The American Bar Association's Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512 on July 29, 2024. It addresses competence, confidentiality, communication with clients, supervision, candor, and fees when lawyers use generative AI. On the point that matters most here, it is direct. While generative AI may be used as a springboard for legal work, lawyers may not abdicate their responsibilities by relying solely on a tool to perform tasks that call for professional judgment. The opinion says lawyers may not leave it to AI tools alone to offer legal advice to clients, negotiate clients' claims, or perform other functions that require a lawyer's personal judgment or participation.
The Professional Ethics Committee for the State Bar of Texas issued Opinion 705 in February 2025, at the request of the State Bar's Taskforce for Responsible AI in the Law. It ties AI use to the Texas Disciplinary Rules of Professional Conduct. On competence, it notes that comment 8 to Rule 1.01 says lawyers should strive to remain proficient in the benefits and risks of relevant technology, and that a lawyer who uses a generative AI tool must have a reasonable and current understanding of it. On supervision, it says lawyers are responsible for the work product they submit regardless of who or what did the original drafting, and that relying on AI risks many of the same perils as relying on an inexperienced or overconfident nonlawyer assistant, citing Rule 5.03.
Neither opinion treats AI as a replacement for the lawyer. Both treat it as a tool the lawyer is fully responsible for.
Confidentiality will shape how negotiation tools are used
Contract negotiation runs on confidential information: a client's walk-away price, its fallback positions, the weakness it hopes the other side does not notice. That information is exactly what a negotiation tool would need to be useful, and exactly what the confidentiality rules protect.
Opinion 705 explains that many generative AI tools invite a conversation in which the lawyer's prompts reveal facts, legal theories, and mental impressions, and that with any tool the lawyer should be reasonably satisfied that the program will not reveal confidential information to others or use it to the client's disadvantage. If the lawyer is not satisfied, the lawyer should not input confidential information without consulting the client and getting consent. Formal Opinion 512 goes further for self-learning tools, concluding that informed consent is required before inputting information relating to the representation, and that boilerplate consent language in an engagement letter is not sufficient.
The practical result is that the lawyer has to decide, matter by matter, what can go into which tool and what the client needs to know. That is a judgment call about the client's interests, not a software setting.
Fees will follow the value
Both opinions also address billing. Under Opinion 705, a lawyer billing hourly may charge for the actual time spent using a generative AI tool, including refining and checking its output, but may not charge for time saved. Formal Opinion 512 says the same and adds that if a tool lets a lawyer complete a task much more quickly, charging the same flat fee may be unreasonable.
We think this will push contract work further toward pricing that reflects outcomes and responsibility rather than hours. If the tool produces the first markup, the client is paying for the lawyer's judgment about what in that markup to accept, what to fight, and how. That is a healthier conversation about value than counting minutes.
Where verification becomes the job
The most widely reported AI failures in law have come from filings with citations to cases that did not exist. In Mata v. Avianca, Inc. (S.D.N.Y. 2023), the court sanctioned lawyers who submitted a brief citing fabricated opinions generated by an AI tool and imposed a $5,000 penalty. Contract work has its own version of that risk. A tool can produce a clause that reads well and does not do what the client needs: an indemnity that fails the Texas express negligence test, a limitation of liability that leaves out the carve-outs the client relies on, or a governing law clause that undercuts a forum the client wanted.
Our post on Texas indemnity and limitation of liability clauses explains why those drafting details decide outcomes. A lawyer who reviews AI output has to understand those rules well enough to catch what the tool missed. As drafting gets faster, the review is where the value sits.
What changes for the client relationship
If we are right about the direction of the work, clients should expect a few changes from their lawyers over the next several years.
First, faster first turns on routine documents, with the lawyer's time concentrated on the points that actually move risk or money.
Second, clearer conversations about how tools are used on their matters, what information goes into them, and how the work is checked, consistent with the communication duties the ethics opinions describe.
Third, more emphasis on playbooks and decision rules written in advance, so that when software or junior reviewers handle volume, the client's positions are applied consistently. Our post on NDA negotiation points gives a sense of the kind of judgment that goes into those positions.
Fourth, and most important, more of the lawyer's time spent with the client rather than with the document. The lawyer who knows the client's business, its risk tolerance, and its counterparties is the one who can tell the client which of the hundred proposed edits matter.
What does not change
The duties do not change. A Texas lawyer owes the client competence, confidentiality, loyalty, candor, and a reasonable fee, whatever tools are on the desk. The responsibility for the advice stays with the lawyer who gives it. Tools can expand what a lawyer can review in a day. They do not decide what the client should accept.
That is why we think the lawyer becomes more central as AI in contract negotiation matures. When both sides can produce competent paper quickly, the difference between a good deal and a bad one is judgment, context, and trust, and those still come from people.
How we can help
We negotiate commercial agreements, NDAs, and transaction documents for growing companies and private equity clients, with a focus on the points that decide risk and value.
If you want to talk about how your contract process is changing, email info@robertsandwillie.com or call (469) 910-4017. Sending a message does not create an attorney-client relationship.
This article is for general information and is not legal advice. Attorney Advertising. Roberts & Willie, PLLC, Dallas, Texas; Trenton Roberts, responsible attorney.
Comments