CBA Record January-February 2026

THE LEGAL PROFESSION AND THE JUDICIARY IN THE AGE OF ARTIFICIAL INTELLIGENCE

The cure is simple: Constrain the tool to the clause in front of you and require that edits fit the existing structure. That one discipline eliminates most of the chaos. Ethical Guardrails Still Apply Letting an AI system revise on its own is not compatible with ethical duties. Illinois Rule 1.1 on competence doesn’t require lawyers to be technologists, but Comment 8 does require us to understand where a tool helps and where it fails and to integrate it in ways that preserve legal judgment. Competence in a digital practice means setting a narrow scope, supplying reliable inputs, and reviewing the output with the same care we would apply to a junior lawyer’s draft. Rules 5.1 and 5.3 on supervision apply to how we deploy and oversee technology: Partners must establish reasonable policies for its use, train their teams, and ensure that oversight is real rather than ceremonial. A tool that redlines unsupervised, drawing from unknown sources and making changes we didn’t direct, turns supervision into cleanup. By contrast, a closed set of approved language, prompts tied to the specific clause and business point, and docu mented attorney review restore the proper order: The AI pro poses; the lawyer decides. Designing a Disciplined Workflow The path forward is neither blind trust nor blanket refusal—it’s disciplined design. That workflow begins before the first prompt is ever typed. A firm should assemble a closed set of approved language: a curated library of forms or clause variants vetted by practice lead ers that reflect the firm’s risk posture and drafting conventions. This isn’t code. It’s organized language the firm already uses with confidence. The AI tool must be limited to that library and to the deal materials supplied by the lawyer. By excluding open-internet sources, the firm avoids stylistic “creep” and unvetted risk allocations. For associates, especially those learning a document type, the library becomes a quiet superpower. It shows what “good” looks like and keeps proposals anchored to the document at hand. One of the hardest skills to develop is spotting what’s missing, not just fixing what’s there. An AI tool, guided by the library and checklists, can help flag likely omissions so reviewers know where to look first. It’s triage, not gospel. The lawyer still verifies the text and decides what belongs. But psychologically, it shifts the work from blank-page anxiety to focused analysis. Instead of facing a sprawling redline, the reviewer starts with a short list of pressure points, like a physician who works more pre cisely when the pain is localized, not described as “everywhere.” A Clause-Level Loop Once the foundation is in place, the attorney-led loop is straight forward. The partner captures intent in plain English—what to change, where to change it, and why.

The goal isn’t to offload judgment; it’s to spare the partner from line editing while anchoring the material points. The AI tool executes that instruction using only the approved language library as its source. The reviewing attorney then reads the clause, checks the com ment against the edit, verifies defined terms and numbering, and confirms that each change was necessary and material. Any hal lucinated or stylistic changes are deleted. That final pass is nondelegable. It’s also a training moment. Editing the AI’s draft teaches newer lawyers how to translate direction into precise, bounded edits that respect the negotiated framework.

Example: Calibrating a Diligence Indemnity Consider a purchase and sale agreement where the diligence indemnity has drifted and become broad. The partner’s instruc tion is high level, but specific: “This indemnity is too broad. Match the definition of ‘Losses’ to our precedent. Buyer should indemnify only Seller, not ‘Seller Parties.’ Losses must be actual and out of-pocket. Tie indemnity to Buyer’s inspections and access. Add a carve-out for Seller’s negligence, willful misconduct, or fraud, and no indemnity for discovery of pre-existing defects. Limit it to one year after termination.” A well-configured AI tool confined to the firm’s approved language executes those directions inside the clause that already exists. The associate reviews for substance and restraint. Substance means ensuring each directive is implemented and harmonized with defined terms. Restraint means policing needless churn, for example, trimming synonyms where the AI rephrased without adding value.

CBA RECORD 23