If you run a solo practice or a small firm with 5 to 25 attorneys, the conversation around AI has probably been frustrating. Half of what you read assumes you want to replace your associates. The other half assumes you have a general counsel and a legal-ops team to evaluate vendors. Neither of those is your life. Your life is the intake call that ate forty minutes because the prospect had two prior conflicts to chase. Your life is the brief draft that lost a Saturday because the outline never quite turned into prose. Your life is the client who replied three weeks ago and has not heard back because the email got buried under discovery review.
This article is about the third thing that AI can actually do for a firm your size. It does not replace your judgment. It does not draft your closing argument. It clears the operational sludge that surrounds the legal work so the legal work gets more of your hours.
The structure I am going to describe is the same one I run for every small business that hires me. It is a $997 fixed-fee assessment, with a written plan in 45 hours, and a guarantee that identifies at least 5 hours per week of repeatable work to automate. I am going to walk through what that looks like inside a law firm specifically.
What an AI Assessment Actually Covers for a Law Firm
The AI Efficiency Assessment is three touchpoints over roughly one week. The first is a 45-minute discovery call. I ask you to walk through your intake process, your case-management stack, your billing cadence, and the specific tasks that consume your week. The conversation is recorded so the written plan can cite specific moments back to you.
The second touchpoint is the deliverable. Forty-eight hours after the call, a 2 to 3 page ranked plan lands in your inbox. Each item on the plan names a specific tool, the workflow it replaces, the estimated hours per week it saves, the implementation difficulty, and the order to attack it. The third touchpoint is a 30-minute walkthrough call where I hand it to you and answer the questions that come up when you actually read it.
That is the entire engagement. There is no retainer behind it. The plan is yours to keep whether you implement any of it or not.
Workflows That Actually Fit a Law Firm
Here are the workflows I see surface most often when I run the assessment for a firm. None of them touch legal judgment. All of them recover attorney or paralegal time.
Intake Triage and Conflict Screening
Most firms run intake through some combination of a contact form, a phone line, and an associate or paralegal who triages. The first thirty minutes of a new-matter call are spent collecting facts a structured form could collect first. A workflow that uses a conversational intake form, routes the response into your case-management system, and runs a preliminary conflict check against your existing client list before you ever pick up the phone can save two to four hours per week at a five-attorney firm. The tools that do this well right now include Clio Grow chained with a workflow tool like Make.com and a language model for free-text summarization.
Document Review Pre-Pass
The pre-pass is the workflow that returns the most hours in most litigation firms. Before an associate spends ten hours reviewing a production set, a language-model pre-pass can tag documents by issue, flag privilege candidates, surface key dates, and extract every name that appears more than twice. The associate still does the actual review. The pre-pass cuts the time-to-relevance by half because the associate is no longer reading the first hundred documents to figure out what the case is about. Vendors in this space include Everlaw, Relativity aiR, and increasingly Claude running against extracted text in a contained workspace.
Brief Drafting from Outlined Arguments
The verification step is the entire point of this workflow. A language model is useful for turning your outlined argument into a first-draft paragraph. It is not useful for citation. Every cite that comes out of a model gets verified against the actual source before it touches a filing. A workflow that pairs a drafting model with a citation-verification pass closes the loop. The verification side is what Veritas exists for, and it is a workflow your assessment will probably name explicitly if briefing volume is in your top three time sinks.
Client Communication Follow-Up
The unanswered-client problem is universal. A workflow that watches your inbox for client emails older than seventy-two hours and surfaces them in a daily digest catches the threads before they become complaints. The same workflow can draft acknowledgment replies for routine messages so the attorney is approving a draft rather than writing from scratch. Two hours per week per attorney is a conservative estimate.
Time Entry Capture from Email and Calendar
Time leakage on billable work is the silent cost most small firms accept. A workflow that scans your sent email and calendar entries, drafts time entries with matter codes, and presents them for one-click approval at the end of the day can recover four to six billable hours per attorney per month. Tools that do this include Smokeball native capture and increasingly Clio Duo, both supplementable with a custom language-model pass for narrative cleanup.
Discovery Review Preparation
Before a meet-and-confer or a Rule 26 conference, a workflow that summarizes the current production state, identifies the gaps your opposing counsel will probably raise, and drafts your initial position for review saves the partner the night-before scramble. This is judgment-adjacent work, not judgment work. The partner still decides the position. The workflow just stops the position from being drafted at eleven on a Tuesday.
Calendar and Deadline Reconciliation
Court deadlines drift. Local rules change. A workflow that reconciles your matter calendar against the actual court docket once a week catches the drift before it becomes a missed filing. The tool stack here is usually a docket-monitoring service plus an automation layer that surfaces drift to the responsible attorney.
The Regulated-Data Reality
Every workflow I described above carries a data-handling question, and the question is not optional. Client confidentiality, attorney-client privilege, and your bar association's technology-competence obligations all converge on the same requirement. Any AI workflow inside your firm needs an explicit data-handling story before it goes live.
The assessment treats this as a first-class constraint, not a footnote. Each recommended workflow comes with notes on whether the vendor trains on submitted data, whether the vendor offers a business-associate-grade contract, whether prompt and output logs are retained, and where the data physically resides. Workflows that cannot be run inside a defensible data perimeter do not enter the plan. That is the bar.
What the Assessment Deliverable Looks Like
The written plan is two to three pages. The opening section ranks the recommended fixes by hours saved per week. The body of the plan walks through each fix with the specific tool, the workflow it replaces, the implementation difficulty, and the order to implement. The closing section is the data-handling appendix, which lists every tool referenced and its current posture on training, retention, and contract terms.
If you want a longer treatment of the deliverable structure across industries, the post on how an audit differs from a consultant engagement walks through the structural difference. The post on why DIY AI projects stall covers what happens when a firm tries to skip the planning step and assemble tools from blog posts.
Who This Is Not For
The assessment is not built for every firm. Big-law practices with more than twenty-five attorneys already have the legal-ops apparatus to run their own evaluations. Pure-transactional firms with very low document volume do not have enough repeatable surface area to make the math work. Any attorney looking to replace judgment, draft unverified citations, or run client work through tools without a data-handling story should not buy this. The assessment is built for solo and small-firm litigators who want to reclaim hours, not outsource the practice of law.
Find the hours hiding in your practice
Forty-five minutes on a call. Forty-eight hours later, a written plan with specific tools, hours per week saved, and a defensible data-handling story for every workflow. If the plan does not identify at least 5 hours per week of repeatable work to automate, you do not pay.
Book your AI Efficiency Assessment →