How To Use Claude To Audit Your Criminal Defense Intake Calls (at scale)
We analyzed 11,261 intake calls for McConathy Law, a criminal defense firm we work with. Eighteen months of data. When we ran the numbers at a $1,500 average case value, the total at-risk revenue came out to $1.4 million.
Half of all their calls were leaking. Missed calls, voicemail drop-offs, and people who were practically ready to hire and still didn't sign. The firm had zero visibility into any of it.
Most criminal defense firms spend serious money getting the phone to ring. SEO, PPC, local ads. But once that call comes in, there's often no way to see what's actually happening at scale. No data, no patterns, just a vague sense that some calls aren't converting and nobody knows why.
AI can fix that. Not by replacing your front desk, but by giving you the ability to crunch thousands of calls quickly and show you exactly where cases are falling through the cracks. This is the process we used, and you can run your own version in about 10 minutes once the system is built.
Why Most Firms Are Flying Blind on Intake
If you asked most firm owners right now what percentage of their calls are going to voicemail and never getting called back, most couldn't answer. They've never had a way to measure it cleanly across a large dataset. And the problem is wider than most people realize — Clio's 2024 Legal Trends Report, built on a secret shopper audit of 500 law firms, found that only 40% of firms answer incoming phone calls and 48% are essentially unreachable by phone.
The same goes for callers who expressed real interest, described a legitimate legal matter you handle, and still didn't book. Were those calls lost because of price? Because the front desk fumbled the close? Because someone called back too late? Most firms are guessing at all of it.
The data exists in your call tracking software right now. The problem has always been pulling it together in a way that's actually usable, and that's the gap AI closes. If you want to see how criminal defense intake lead loss typically breaks down before you build this system, that piece covers the patterns we see most often.
Step 1: Build the Dataset
AI can't analyze what it can't see, so the first step is building a dataset that contains every call your front desk receives with the right fields attached.
At minimum you need these four columns:
Landing page. Where did this lead come from? This tells the AI which sources are producing the best leads and which practice areas they're coming from. If you're spending $5,000 a month on DWI PPC and the intake data shows those calls are dropping at a higher rate than organic calls, that's a marketing budget conversation that needs to happen.
Transcript. This is the most valuable field in the dataset. We use a call tracking service called WhatConverts, which provides a full transcript for every call logged. AI can read every word of every transcript and identify patterns across thousands of calls at once. Language that consistently shows up right before a drop-off, questions that never get answered, moments where someone was close to signing and the front desk lost the thread. A human listening to calls one at a time couldn't cover this volume. AI can.
Summary. Before you export anything to Claude, you can use a short AI prompt directly in Google Sheets to pre-process each transcript into a one-line summary. The prompt is straightforward: analyze the transcript, and if the recipient doesn't appear at least five times in the conversation, mark it as a missed call, otherwise provide a brief summary. This gives you a readable overview of every interaction without having to open a single transcript manually.
Practice area. Have the AI read each transcript and categorize which practice area the caller was inquiring about. Criminal defense, DWI, traffic tickets, whatever your firm handles. Once that's in the sheet, you can filter by case type and see if intake performance varies across practice areas. Often it does.
These four fields give you a complete, structured picture of your intake that's actually ready to analyze.
Step 2: Automate It So You Never Have to Touch It Again
The workflow is WhatConverts feeding into Google Sheets through Zapier. Anytime a new call comes in through your call tracking software, Zapier automatically passes it into the sheet as a new row with the landing page and transcript fields populated. The AI summary and practice area columns run automatically in the sheet using the prompts you configure once.
Set this up once and every call that comes into the firm gets logged, summarized, and categorized from that point forward with no manual work involved.
After 30 to 90 days you have a clean dataset ready to run through Claude. Thirty days gives you enough to start spotting patterns. Ninety days is where the bigger trends and seasonal shifts become clear.
Step 3: Run the Analysis in Claude
Once you have 30 to 90 days of data, download the sheet as a CSV (File, Download, Download as CSV) and take it to Claude.
The prompt we use is intentionally narrow. You don't want a report that surfaces 20 problems and leaves you not knowing where to start. The prompt focuses on three specific leaks that show up repeatedly across criminal defense firms, tells Claude to analyze as an intake specialist across the full transcript and summary data, and asks for total count, cost per month, and average case value impact for each leak.
Before you run it, plug in your actual average case value. The example uses $1,500 but if your average signed case is $4,000 or $6,000, replace that number first. The dollar figures in the output are only useful if they reflect your firm's real economics. Attach the CSV, run it, and plan for about 20 minutes for a dataset of a few thousand calls.
The Three Leaks and What They Mean
The report breaks your intake down into three categories.
Leak 1: Missed calls. Calls where the front desk didn't pick up and no meaningful follow-up happened. In the firm we analyzed, this category alone represented $200,000 in at-risk revenue over 18 months. For context on how widespread this is, law firm intake statistics across the industry put unanswered calls at roughly 195 million per year, with over half of all firms reporting they have lost business directly because of missed calls. After-hours answering services, auto-response SMS when a call goes unanswered, callback workflows triggered within minutes — these are solvable problems with systems that already exist.
Leak 2: Voicemail drop-offs. Callers who hit voicemail and hung up without leaving a message. Another $148,000 in the analysis we ran. If someone calls and gets voicemail, your system should be triggering an outbound SMS or callback attempt within a few minutes, not hours later when the caller has already moved on to the next firm.
Leak 3: Lost closes.Callers with a real legal matter in your practice area who expressed genuine interest and still didn't sign. In the firm we analyzed, this was the single largest leak and the one with the most room to move. Unlike the first two, this is not an automation problem. It's a sales training problem. Your front desk is getting close but not closing, and the AI can go through every one of those calls and pull out what language patterns show up before the drop-off, where the conversation loses momentum, and what questions go unanswered. You use that output to build your coaching and scripting around what's actually happening on your calls, not what you assume is happening.
The reason it matters to separate leaks 1 and 2 from leak 3 is that the approaches to fixing them are completely different. Automation solves missed calls and voicemail drop-offs. Coaching and scripting work solves lost closes. Applying the wrong fix to the wrong problem wastes time and money.
What This Tells You About Your Marketing
Most people run this analysis and focus entirely on the intake side. The landing page field does something more useful than that.
If you're running PPC campaigns across three practice areas and one of them is consistently generating calls that drop at the voicemail stage, that's a marketing budget question as much as an intake question. You might be driving high-intent traffic into a broken process and continuing to fund it because nobody connected the two datasets.
The firms that get the most out of this analysis use it to look at marketing and intake together rather than treating them as separate problems. Your cost per call means nothing if the process on the other end is losing half of them.
A Realistic Expectation
The firm we analyzed was an extreme case. Half of all calls leaking and $1.4 million at risk is not typical. Some firms will come in at a 20 percent leak rate, some higher. Most firms genuinely don't know where they sit, which is the point of running this.
Even if your numbers look better than the example, there will still be patterns worth acting on. A 15 percent improvement in close rate on the lost closes category alone, at whatever your actual case value is, adds up considerably over a year. You can do the math on your own numbers once you see them.
Key Takeaways
Most criminal defense firms have no real visibility into what happens after the phone rings. The data exists in call tracking software but it has never been pulled together and analyzed.
Four fields make this analysis work: landing page, call transcript, AI-generated summary, and practice area. Everything else in the workflow builds on those.
Set up the WhatConverts to Google Sheets automation through Zapier once. After that, every call gets logged and processed automatically.
Use Claude with a focused prompt targeting three leaks: missed calls, voicemail drop-offs, and lost closes. Keeping the scope narrow produces more actionable output than trying to surface everything at once.
Missed calls and voicemail drop-offs are automation problems. Lost closes are a sales training problem. Fixing the wrong thing for the wrong leak wastes time.
The landing page field connects intake performance back to your marketing spend. A PPC campaign generating calls that consistently leak is a budget allocation problem, not just an intake problem.
Replace the $1,500 average case value in the prompt with your actual number before running it. The output is only useful if the revenue figures reflect your firm's real economics.
Want to Go Deeper?
This is the foundational version of the analysis. Even at this level, most firms will find at least one thing worth fixing immediately. The next step is running more specific prompts to dig into call quality, front desk language patterns, and what specifically is being said in the lost close calls.
Watch the next video in the series for that layer of analysis. Or if you want us to run this through our criminal defense intake service directly, book a call and we can walk through what your numbers actually look like.
