Will AI Replace Lawyers? Tasks, Review, and Accountability
AI can draft documents, research case law, and summarize depositions, but professional judgment and ethical responsibility stay with the attorney. Here is where the line sits and how to use AI responsibly in legal work.
Updated 6 min read
AI can assist lawyers with research, drafting, document review, and organizing case materials faster than was previously possible. What it cannot do is exercise professional judgment, bear ethical duties, or maintain the client relationship that define legal practice.
For most legal work, the question is not replacement but supervised use: how to apply AI assistance without breaching duty to the client. Tools like fenbs keep approved working rules and review decisions outside individual chats, so every assistant on a matter starts from the same agreed approach rather than guessing at the firm's standards.
What AI can assist with in legal work
AI tools are genuinely useful for the preparatory and organizational side of legal work. A well-prompted assistant can surface relevant case law, summarize lengthy depositions, draft an initial contract section, or flag inconsistencies across a set of documents.
- Legal research: finding on-point cases, statutes, and secondary sources across a jurisdiction.
- Document drafting: producing a first version of contracts, briefs, motions, or demand letters for attorney review.
- Document review: reading large sets of files and flagging clauses, dates, or provisions that match defined criteria.
- Summarizing: condensing depositions, discovery responses, or exhibit lists into working notes.
- Intake checklists: drafting the questions a client interview should cover for a given matter type.
- Correspondence: preparing routine letters or emails that an attorney reviews before sending.
Each of these tasks still requires a lawyer to check the result. A research memo that cites a superseded case or a contract draft that omits a material term is worse than nothing because it can create false confidence. AI output is a starting point, not a finished work product.
Where professional judgment remains with the lawyer
Legal practice involves obligations that cannot be delegated to a tool. Under the ABA Model Rules of Professional Conduct (opens in a new tab), a lawyer must competently supervise any non-lawyer assistance, including AI-generated work. An attorney signs the filings and carries the bar license.
Client counseling is the clearest example. Advising a client on which course of action to take, weighing litigation risk against settlement, or explaining how a regulatory change affects their business requires judgment about the specific client, their situation, and their goals. An assistant that has not met the client and has no duty of loyalty to them cannot make that call.
Professional responsibility also includes maintaining confidentiality, recognizing conflicts of interest, and deciding what to communicate to opposing counsel. These duties are personal to the attorney and are not transferable. State bar associations have issued guidance on AI use in legal practice; check your jurisdiction before relying on any AI tool with client materials.
Keep approved review rules where every assistant reads them
fenbs stores Decisions and rules for each matter or practice area: which sources to cite, which clauses require partner review, which communications must not go out without sign-off. Every connected assistant reads those rules first. When a partner updates the standard for a clause, that update reaches every assistant on the next session, not just the one that was open at the time.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
What bar associations and courts have said
Several state bar associations have issued formal guidance on AI use in legal practice. Most share a common thread: attorneys remain responsible for the accuracy and completeness of AI-assisted work and must exercise independent judgment before relying on any output. The duty of competence now includes understanding AI tools sufficiently to supervise them.
Courts have sanctioned attorneys for submitting briefs that cited cases generated by an AI tool that did not exist. Those incidents did not end careers, but they did cause significant harm. The lesson is that AI research requires the same verification step as any other source: if you cannot find the primary document, do not cite the claim.
The picture across professions is similar: accountants, writers, software developers, coders, and UX designers face the same pattern of assisted tasks and retained responsibilities.
Handling confidential information with AI tools
Attorney-client privilege and client confidentiality apply to how a lawyer uses every tool, including AI. Before sending client materials to any AI service, check whether that service uses inputs to train its models, stores conversation history, and what its data handling terms say. Some firms use locally hosted models or enterprise agreements to manage this risk; others restrict AI use to general research that does not include client facts.
fenbs does not store privileged case materials. What it holds is the working instructions: which approach the team has approved for this matter type, which sources have been verified, and where the task stands. Keeping those outside the AI app separates the operational record from the confidential file.
Building a supervised AI workflow for a law practice
A practical workflow treats AI output as a draft that requires attorney review before it is relied on. The attorney sets the task, reviews the result, corrects it, and keeps a record of what was checked. That record is what makes AI use defensible if a client or regulator later asks how work was produced.
A shared record of approved steps also helps when different people use different AI apps. A junior associate working in Claude Code and a partner using ChatGPT can both read the same approved checklist for a standard contract type and add their review notes to the same task. AI agents for lawyers covers the setup in more detail.
When fenbs does not fit
fenbs does not hold privileged client files, full case documents, or evidence. It is a work-management record: tasks, rules, approved approaches, and progress notes. A full document management system or a case management platform handles the documents themselves.
How to do this with fenbs
- Create a free account. Add a project for your practice area or a specific matter type, not for a live client matter.
- Write the review rules as Decisions and rules. For example: "All AI-drafted contract clauses require attorney review before use" or "Cite only cases verified in Westlaw or Lexis." Record who decided each rule.
- Add a task for each document type you produce with AI assistance. Note the approved prompt or outline in the task plan.
- Connect your AI app over MCP from Integrations. When you start a drafting session, ask the assistant to read the project context and its rules before it drafts.
- After each session, ask the assistant to note what it drafted, what the attorney reviewed, and what remains. That note becomes the next session's starting point and the practice's quality record.
For more on AI agents in a legal context, see AI agents for lawyers. For the broader question of how AI handles professional tasks, AI employees explains the delegation model.