Witness prep eats time. You’re juggling prior depositions, discovery documents, case memos, and your own notes from client meetings. Then you sit down to build a question list, cross-reference statements, and flag the gaps that matter. For a single witness, that’s 6 to 10 hours of billable work that often gets compressed into 3 hours of actual invoicing because the client sees it as overhead.

Most partners I talk to describe the same pattern. An associate spends two days pulling together a witness brief. The partner reviews it, rewrites half the questions, and still walks into the deposition feeling like they missed something. The work is necessary, but it’s slow and expensive. When you’re running a firm that bills $400 an hour for associate time and $650 for partner time, those inefficiencies compound fast.

AI agents built for witness interview preparation can handle the first-pass work in minutes. They read every deposition transcript, flag inconsistencies across statements, and generate a structured question list based on the issues you care about. The result is a draft brief that would take a junior associate 8 hours to produce, delivered in 20 minutes. You still review it. You still add the nuance and strategy. But you’re not starting from a blank page, and you’re not paying someone $400 an hour to highlight PDFs.

The Manual Workflow Nobody Invoices For

Witness prep starts weeks before the interview. You pull the complaint, the answer, any prior depositions, and the relevant discovery documents. Then you read through everything, taking notes on dates, claims, and contradictions. If the witness has given a statement before, you compare it line by line to what the other side said in their deposition. If they haven’t, you’re building questions from scratch based on what you need to establish or impeach.

This is high-value work, but it’s hard to bill transparently. Clients see “document review” on an invoice and assume it’s paralegal work. They don’t understand that a partner spent 4 hours cross-referencing three depositions to find one inconsistency that changes the case. So firms eat the time or bundle it into a flat fee that doesn’t reflect the actual cost.

The second problem is consistency. One partner might spend 10 hours on witness prep. Another spends 3 and wings the rest. The quality of the brief depends entirely on who’s doing it and how much time they have that week. When you’re managing a team of five attorneys across 40 active matters, that variance adds up. You can’t scale a process that lives entirely in someone’s head.

The third issue is retrieval. You know the opposing counsel said something contradictory in a prior case, but you can’t remember which deposition or which page. So you search your document management system, scan three transcripts, and eventually give up because it’s faster to move on. That’s not a knowledge problem, it’s a retrieval problem. The information exists, but finding it costs more time than it’s worth.

AI agents solve all three. They read every document in the case file, extract the relevant facts, and cross-reference statements automatically. They don’t forget. They don’t skip pages. And they don’t bill $400 an hour to do it.

What a Document Review Agent Does for Witness Prep

A Document Review Agent is an Omni ops agent that ingests case files, depositions, and discovery documents, then produces a structured output based on the task you assign. For witness prep, that task is simple: read everything related to this witness, identify the key facts, flag any inconsistencies, and generate a question list organized by topic.

The agent starts with the deposition transcripts. It pulls every statement the witness made, tags it by subject matter, and compares it to statements from other witnesses. If the plaintiff said the accident happened at 3pm and the witness said 4pm, the agent flags it. If the witness described the vehicle as blue in one answer and black in another, it flags that too. These aren’t subjective judgment calls, they’re pattern matches across structured text.

Next, the agent reviews the discovery documents. If the witness submitted an affidavit, the agent compares it to their deposition testimony. If there’s an email chain where the witness discussed the incident, the agent pulls the relevant excerpts and links them to the corresponding testimony. The output is a timeline of statements with source citations, so you can see exactly where each fact came from and whether it’s consistent across documents.

Then the agent generates questions. It doesn’t write them the way a partner would, but it gives you a scaffold. For each topic, it suggests 3 to 5 questions based on the facts in the file and the inconsistencies it flagged. You review the list, rewrite the questions to match your strategy, and add the follow-ups that require judgment. What used to take 6 hours now takes 90 minutes, and the output is more thorough because the agent didn’t skip anything.

One litigation partner I work with describes it this way: “I used to spend half a day building a witness brief. Now I spend 90 minutes editing one. The agent finds things I would have missed because I don’t have time to read every email thread twice. It’s not replacing my judgment, it’s giving me better raw material to work with.”

The Economics of 8 Hours Per Witness

If you prepare 4 witnesses per case and handle 15 cases per year, that’s 60 witness preps. At 8 hours per prep, you’re spending 480 hours on this work annually. If half of that is partner time at $650 an hour and half is associate time at $400 an hour, the fully loaded cost is $252,000. Most firms invoice less than half of that because clients push back on document review line items.

An AI agent doing the first-pass work cuts that 8 hours to 2. You’re still reviewing, refining, and adding strategy, but the mechanical work is done. That’s 360 hours saved per year. If you redeploy that time to billable client work, you’re looking at $189,000 in recovered capacity. If you simply reduce the associate hours you’re eating, you’re saving $120,000 in cost.

The second-order benefit is consistency. Every witness brief follows the same structure. Every inconsistency gets flagged. Every question list covers the same topics in the same order. That makes it easier to train junior attorneys, easier to hand off cases, and easier to deliver predictable quality to clients. You’re not dependent on one partner’s memory or work style.

The third benefit is speed. When a deposition gets moved up by two weeks, you don’t scramble to prep. The agent runs the analysis overnight, and you review the brief the next morning. That flexibility matters when you’re managing a caseload that doesn’t respect your calendar.

If you want to see where else AI can recover capacity in your practice, we built a worksheet that walks through the highest-impact workflows. Grab the AI Client Intake Checklist for Law Firms and use it to map your own intake and matter-management process. It’s a 15-minute exercise that shows you exactly where the leakage is.

How We Build This for Law Firms

When we build a Document Review Agent for a firm, we start with an Omni Audit. That’s a 60-minute working session where we map your current witness-prep workflow, identify the documents you’re working with, and define the output format you need. We don’t pitch. We don’t present a deck. We produce three things: a process map, a cost model, and a prototype spec.

The process map shows every step in your current workflow, from pulling the case file to delivering the final brief. We time each step and tag it by role. That gives you a clear picture of where the hours go and which steps are candidates for automation.

The cost model translates those hours into dollars. We calculate the fully loaded cost of the current process, the cost of the AI-assisted process, and the net capacity gain. Most firms see a 60 to 75 percent reduction in prep time, which translates to $120K to $180K in annual savings for a practice handling 15 cases per year.

The prototype spec defines what the agent will do. For witness prep, that usually means ingesting deposition transcripts and discovery documents, extracting key facts, flagging inconsistencies, and generating a question list. We specify the input format, the output structure, and the review workflow. If you want the agent to compare statements across multiple witnesses, we add that. If you want it to flag specific legal standards or elements of proof, we build that in.

After the audit, you decide whether to move forward. If you do, we build the agent in 3 to 4 weeks. We integrate it with your document management system, train it on your case files, and test it on 2 or 3 real matters. Then we hand it off with a runbook and a 30-day check-in to make sure it’s working the way you need.

Book a 60-min Omni Audit and we’ll map your witness-prep process in the first half hour. By the end of the call, you’ll know exactly what an agent can do for your practice and what the ROI looks like.

Other Agents That Fit the Same Workflow

Witness prep isn’t the only place where document review eats time. Most litigation practices have 3 or 4 workflows that follow the same pattern: read a stack of documents, extract the relevant facts, and produce a structured output. An AI agent can handle all of them.

A Matter Triage Agent reviews intake forms and consultation notes, classifies the practice area, scores case fit, and routes the matter to the right partner with a one-paragraph brief. That cuts intake processing time from 30 minutes per lead to 3 minutes, and it ensures high-value matters get routed immediately instead of sitting in a queue.

An Intake Voice Agent answers every call, even after hours. It conflict-checks the caller, captures the basic facts of the matter, and books a consultation directly into the firm’s calendar. One firm we work with went from converting 28% of after-hours calls to converting 64% because the agent answered every call within two rings and booked the appointment before the caller hung up.

A Contract Review Agent reads engagement letters, settlement agreements, and vendor contracts, then flags non-standard clauses and produces a redline with suggested edits. It doesn’t replace attorney review, but it gives you a clean first draft in 10 minutes instead of 2 hours.

All of these agents plug into the same infrastructure. Once you have one Omni ops agent running, adding a second or third is faster because the integration work is done. Most firms start with one high-volume workflow, prove the ROI, then expand to two or three others over the next quarter.

You can see the full agent library and read more about how Omni for law firms works on the audit page. We’ve built agents for discovery review, brief drafting, client intake, and matter triage. The workflow is the same: map the process, build the agent, test it on real cases, and hand it off with a runbook.

What Changes When You Deploy This

The first thing you notice is speed. A witness brief that used to take two days now takes two hours. That means you can prep more witnesses per case, or you can take on more cases without hiring another associate. Either way, you’re increasing throughput without increasing headcount.

The second thing is consistency. Every brief follows the same structure. Every inconsistency gets flagged. Every question list covers the same ground. That makes it easier to delegate, easier to review, and easier to train new attorneys. You’re not relying on institutional knowledge that lives in one partner’s head.

The third thing is leverage. When a case heats up and you need to prep three witnesses in one week, you don’t panic. You run the agent on all three case files overnight, review the briefs the next morning, and spend your time on strategy instead of document review. That flexibility changes how you manage your caseload.

One managing partner told me this: “We used to turn down cases because we didn’t have the associate capacity to handle discovery and witness prep. Now we take the case, run the agent, and bill the client for partner time instead of associate time. The margin is better, the work is faster, and the client gets a better result because a senior attorney is doing the strategy work instead of the document review.”

That’s the shift. You’re not replacing attorneys, you’re redeploying them to higher-value work. The agent does the mechanical first pass. The attorney does the judgment, the strategy, and the client interaction. The client gets a better outcome, the firm makes more margin, and the attorney spends their time on work that actually requires a law degree.

The Build Process

We don’t sell software. We build custom agents for your practice. That means the agent is trained on your case files, integrated with your document management system, and configured to produce output in the format you already use. It’s not a SaaS tool you have to adapt to, it’s an agent that adapts to you.

The build starts with the audit. We spend 60 minutes mapping your workflow, identifying the documents you work with, and defining the output you need. We don’t pitch, we don’t present slides, we just work through the process with you and produce a spec.

If you decide to move forward, the build takes 3 to 4 weeks. We integrate the agent with your systems, train it on a sample of your case files, and test it on 2 or 3 real matters. You review the output, we refine the prompts, and we iterate until it’s producing briefs that meet your standard.

Then we hand it off. You get a runbook that explains how to use the agent, a support contact for questions, and a 30-day check-in to make sure everything is working. After that, the agent runs on its own. You feed it case files, it produces briefs, and you review them. No ongoing fees, no per-use charges, no SaaS subscription.

Most firms start with one workflow, prove the ROI, then expand. Witness prep is a good starting point because the input and output are well-defined, the time savings are measurable, and the risk is low. Once you see the agent working on witness prep, it’s easy to imagine it working on discovery review, contract analysis, or matter triage.

If you want to see what this looks like for your practice, book my Omni Audit. We’ll map your witness-prep process, calculate the time savings, and show you exactly what an agent would do. No deck, no pitch, just a working session that produces a clear answer.

Why This Matters Now

The firms that adopt AI agents in 2025 will have a 12 to 18 month head start on the firms that wait. That’s not hype, it’s arithmetic. If you save 360 hours per year on witness prep alone, and you redeploy that time to billable work, you’re adding $180K to $240K in annual capacity. Do that across 3 or 4 workflows and you’re looking at $500K in recovered capacity without hiring anyone.

The firms that move first will also have better-trained agents. An agent that’s been running for 12 months on your case files will produce better output than an agent you deploy next year. It will know your practice area, your document formats, and your preferred output structure. That institutional knowledge compounds over time.

The third reason to move now is client expectation. Clients are starting to ask how you’re using AI to reduce costs and improve turnaround time. If your answer is “we’re exploring it,” you’re behind. If your answer is “we’ve been running AI agents on discovery and witness prep for six months, and we’ve cut our turnaround time by 40 percent,” you win the pitch.

You can read more about how we’re helping law firms deploy AI across intake, matter triage, and document review on the EDNA blog and in our guides section. We publish new case studies and workflow breakdowns every month, and we’re building a library of agent specs that you can adapt to your own practice.

This isn’t about replacing attorneys. It’s about giving them better tools so they can spend their time on the work that actually requires judgment. The document review, the first-pass analysis, the cross-referencing, that’s all mechanical. An agent can do it faster, cheaper, and more consistently than a human. The strategy, the client interaction, the courtroom work, that’s still yours. The question is whether you want to spend your time on both, or just the part that matters.