Why Your Firm's AI Strategy Is a Professional Responsibility Issue

The following is based on Session 1 of AI Ready, a two-part series from Affinity Consulting, Centerbase, and NetDocuments. View the webinar recording here.

Most firms already use AI. Far fewer can prove how the work got done. That distance between using AI and governing it is where your professional exposure is. This is not simply a technology problem, although the right technology can help. It is, however, a leadership decision about how your firm operates.

You Are Not Behind. The Game Got Harder.

Around 87 percent of large firms report active AI use. Fewer than half have a comprehensive governance framework. Most cannot produce a complete account of how a given piece of work got done when asked.

Four Places the Gap Becomes an Ethics Problem

Ethical risk does not live inside the AI model. It emerges at the operational handoffs, the moments where output moves from one person to the next without anyone owning the review. Each of the four patterns below maps to a specific Model Rule.

Pattern one: unverified output, Rule 1.1

An associate uses AI to draft a motion. It cites three real cases, all correctly formatted. One of them was overruled 18 months ago. The associate skims the draft and files it. The court issues a show-cause order.

Two people are exposed here. The associate failed to verify the output independently. The supervising partner failed to ensure a review process existed at all. Competence under Rule 1.1 includes independent verification, which means judgment cannot be handed off to a tool. The cautionary case is Mata v. Avianca, where a lawyer was sanctioned for filing AI-invented citations.

The accountability problem, both directions

Unreviewed AI output that cannot be defended is a problem whether your lawyers are sending it or your clients are. Clients now use AI to draft correspondence, summarize facts, and even propose strategy. Some of that input is useful and some of it is quietly, dangerously wrong. The standard is the same in every direction. Read it critically, make it your own, defend every word, and verify client-provided facts before you rely on them.

Pattern two: supervision gaps, Rules 5.1 and 5.3

Partners supervise associates under Rule 5.1. Lawyers supervise nonlawyers under Rule 5.3. So who supervises the AI? When no documented review process exists, both the person using the tool and the person responsible for them are exposed. The workable answer is to supervise AI the way you would supervise a law clerk or a contract attorney. Give it a defined scope, require review, and document the oversight. No one gets to say they did not know a tool was used or assumed the result was correct.

Pattern three: confidentiality exposure, Rule 1.6

The high-risk act is an associate pasting a sensitive document into an open AI system. Privilege logs, settlement strategy and authority limits, engagement letters with fee caps, and expert reports with damages calculations all qualify. Once a mediation brief goes into a consumer tool, that confidential information now sits in an outside provider's systems. Rule 1.6 requires reasonable efforts to protect client information, and reasonable now means your systems can classify documents by sensitivity and enforce boundaries. If you cannot classify a document, you cannot protect it. This is where a document management system stops being convenience and becomes an ethical control layer.

The law here is still moving. Federal courts are applying familiar principles of confidentiality and work product to a new technology, and the early results are highly fact-specific. Courts are starting to agree on a few things. An AI platform is not a lawyer. What the provider is permitted to do with your inputs matters. Counsel involvement matters but is not a magic phrase that cures everything. Safeguards matter more than labels. Still unsettled are questions like whether enterprise safeguards preserve privilege, whether client-initiated and lawyer-directed use differ, whether prompts and outputs count as work product, and how relying on AI output in a filing affects waiver.

Pattern four: billing integrity, Rule 1.5

A task that would have taken four hours by hand takes 20 minutes with AI, plus another 90 minutes of the Associate’s time reviewing, verifying, and refining. The Associate bills four hours because that is what the work is “worth”. Rule 1.5 requires fees based on actual effort, and billing a manual rate for AI-assisted work without disclosure is a problem. The firm needs billing guidelines for AI-assisted work and needs to communicate AI use to clients before they ask. If a client asks how much of this was AI-assisted and no one can answer, that is a billing integrity issue.

A Policy Is Not a Strategy

Many firms wrote an AI acceptable use policy in 2024, and then nothing changed. A policy states acceptable use, requires training, and documents intent. Useful, but a policy without enforcement is a suggestion. A strategy defines where AI is allowed to operate, enforces the review workflow, integrates with your systems, creates the audit trail, and assigns accountability. Policies express intent. Strategy enforces behavior. You cannot policy your way out of an infrastructure problem.

Human in the loop, and the honest version of it

Every failure pattern comes back to one question. Was a human applying judgment? Judgment brings experience and context to what AI produces. Documentation, in the form of an audit trail, shows that the judgment actually happened and makes it defensible. You need both. An audit trail without judgment is a rubber stamp. Judgment without documentation is hard to defend when someone asks you to prove it. The honest framing is that no audit trail existed for traditional legal work either, so this is not a new burden invented by AI. Integrated systems are simply what now make it possible to show a human was in the loop.

What comes next

Check out our next session, From Risk to Readiness, which runs on August 11, 2026. It covers the four pillars of an ethical AI strategy, how to identify your firm's maturity stage and next move, using AI as a leadership tool rather than only a drafting one, and a preview of a 90-day blueprint.

Register here: https://www.centerbase.com/webinars/ai-readiness

Missed session one? View it on-demand by registering here: https://www.centerbase.com/webinars/ai-readiness

—

Reference material: ABA Formal Opinion 512 on generative AI, ABA Model Rules 1.1, 1.5, 1.6, 5.1, and 5.3, and Mata v. Avianca, Inc. (S.D.N.Y. 2023). Prepared from Session 1 of the AI Ready series, presented by Debbie Foster, Affinity Consulting Group and sponsored by Centerbase, NetDocuments.