AI Inside a High-Volume Plaintiff Firm

Collin McKee

Welcome back to the Endeavors AI podcast. I'm Collin McKee at Endeavors AI — we focus on integrating AI into operations to create measurable leverage inside firms. Today I'm joined by Tim Semelroth, managing partner of the largest injury and disability firm in the state of Iowa. Tim leads a full-service plaintiffs' firm handling personal injury, workers' compensation, employment law, and disability cases — and he's teaching AI at the University of Iowa College of Law. So Tim, you're not only talking about AI, you're implementing it at scale inside a high-volume plaintiff firm. Let's start there: what does AI actually look like inside your practice today?

Tim Semelroth

It's used for all sorts of things — whether it's summarizing medical records, helping organize medical-expense summaries, doing first drafts of answering written discovery, or analyzing expert reports. I could go on and on. It really does touch every step of the process in our firm.

Collin McKee

You've said you're less interested in abstract ethics debates and more interested in what lawyers can actually do next week. So what are two or three practical applications lawyers should be implementing immediately?

Tim Semelroth

It's not that I'm poo-pooing ethics — those are non-negotiable: keeping client information confidential, understanding how the tools work, using them competently. But that's what everyone's talking about right now; that's what law professors are writing articles about. My biggest fear is lawyers who say, "Wow, that seems complicated, I'm going to stick my head in the sand." That's a huge mistake. So when I speak to practicing lawyers, my focus is: here's something you can go do next week.

It can be simple. If you're in litigation, we all have the opposing counsel who's setting you up for a motion to compel by sending a three-page letter explaining all the deficiencies in your discovery responses. When I get those I think, "What a drag — I have to parse this out line by line." So I love taking a really long letter like that, dropping it into an AI tool, and saying, "Identify all the things he says he's entitled to." Now I've got my checklist, and that's how I'll respond to each item. That's one.

Another — and you'd asked beforehand what's made the biggest difference in the past 12 months — for me it's using a Plaud device, an AI-powered voice recorder. I'm old enough that when I started practicing, they'd send the associates into the library with a dictaphone to dictate deposition summaries and case memos. That didn't last long for me, because I came up typing — I can type faster than I can dictate with punctuation. But now things are coming back around, because a tool like Plaud is a voice recorder that's AI-powered. You get a much cleaner transcript — punctuation, paragraphs — and then you can chat against it and ask it questions. Plaud has its own summary feature, but what I love to do is take a Plaud transcript into another tool like Claude or ChatGPT and use it as the basis for a document for a case, or even a marketing project — I'll dictate my thoughts on a frequently asked question, throw it into Claude, and have it turned into a script for a web video.

Collin McKee

I've got one myself. Neither of us is sponsored by them, but maybe we should look into it, because we can both sing its praises. It does make a huge difference. What about intake — has AI improved intake quality or speed in your firm?

Tim Semelroth

Yes. There are things we all know we should do — it's like exercise. We've known for years we should record our intake calls and listen to them, both for quality control and to make sure the people doing that first touch with potential clients are following the scripts and asking the right questions. That advice is 20 years old. But busy lawyers never really had the time to listen in. Until AI. Now you can drop transcripts of all your intake calls into AI and have them graded against criteria you think are important, and get a report: here are this many calls, and here's how they graded. So now you can supervise and coach your intake staff at a level you just couldn't before.

Collin McKee

You also mentioned that podcast transcripts and online content are increasingly being digested by AI tools. Are you thinking about visibility differently in the age of AI research?

Tim Semelroth

Absolutely — that's why we're talking. If you do a search on Google now, the AI Overviews are something Google is pushing to stay relevant. Ask a question like "what are the most important things to do after a car accident in Iowa?" and click into the AI Overview, and you'll see web videos pop up — it's pulling from YouTube because of the relationship between Google and YouTube. So a podcast interview like this, or a video on your YouTube channel, is being digested by Google and by the large language models. When people ask very specific questions, if you've got the resources out there, it pulls from them. If you ever need help convincing someone to be a guest on your show, tell them this is the easiest way to make sure you're feeding the beast — especially for the questions potential clients and referral partners actually want answered.

Collin McKee

You've got to get out there to be discoverable. Now that you're teaching, do you think law firms broadly understand how AI will reshape attorney discovery and practice?

Tim Semelroth

Not on a big scale yet — not compared to other industries. Insurance and medicine are further along in their AI journey than lawyers are, partly because we're a lot of small-business owners. We have the largest injury and disability firm in Iowa — 19 lawyers across three offices — but we're still a small business by most metrics. So someone at each firm has to say, "We're going to commit to this; we're going to figure out how to incorporate AI into each aspect of what we do." And the biggest issue — it's not a problem for us because we do almost exclusively contingency-fee work — but the scary thing for the profession is that a significant number of lawyers still bill by the hour. If you can buy a tool that lets you do in one hour what used to take ten, you're costing yourself money. So there are unique challenges to bringing AI into a law firm that don't exist for the tire shop down the street.

Collin McKee

And doing it without a structured plan can be very costly in your field — there aren't really "small mistakes" in law practice.

Tim Semelroth

The rules of ethics are not optional. We have hurdles to overcome in using AI that a lot of industries don't.

Collin McKee

You mentioned you spend a good amount to make the phone ring. What did your intake audit reveal that surprised you?

Tim Semelroth

Our frontline staff is pretty good, but there are always curveballs when you're talking with the public — cases people don't realize we can help with, or ones we can't. One big thing I realized: you constantly have to educate your staff about what you do and the different ways you can help, and do it so they're enthusiastic about recommending you. When someone says, "This horrible thing just happened to my family," you want someone on the phone who says, "Oh my gosh, that's terrible — but you're in the right place, because we have just the lawyer you need to talk to." We probably don't do a good enough job keeping everyone in the loop about how talented our lawyers are and all the ways we can help.

Collin McKee

You spoke a little about Smith AI. Can you tell our audience about your experience with them?

Tim Semelroth

We didn't hire them initially as an "AI service." We hired them because, particularly post-COVID, people are ready to make a buying decision faster than ever. If someone isn't answering your phone 24/7, you miss potential cases — because when someone has finally built up the courage to call a lawyer (which is scary), and they get your voicemail, they're not going to wait for a call back. They move on to the next person. So we hired Smith AI for after-hours and overflow answering, and they did a fine job. At the time they didn't offer an option to actually sign people up if they met criteria — they may now, so I don't want to talk down their capabilities.

Collin McKee

That's not the goal at all.

Tim Semelroth

What became important for us — and why we made a move — is that some people are ready to sign with a lawyer right now, and you have to meet them where they are. As long as you've given clear criteria about what type of cases you want, you want someone empowered to say, "I'm going to text you a contract right now that you can e-sign." I think that's essential in 2026, and that's why we moved to a different answering service.

Collin McKee

That may be hard to quantify, but I'll ask anyway — compared to just a voicemail after hours, what does removing that friction mean financially? Did you track it?

Tim Semelroth

We're always doing five things at once, so it's not a true A/B test — we're always trying new ways to get the phone to ring. But I can tell you that 24/7 answering, combined with a service that can sign people up when they meet your criteria, is one factor in why we keep growing. We're a very mature firm — over 30 years old — and we're not in a huge market, but we're doing 15% growth year after year. The after-hours and overflow answering service is one reason we're able to keep that up.

Collin McKee

Either way you slice it, removing barriers to entry is a good thing. You also made it a rule that consumer-grade AI can't be used for casework unless the firm provides a professional account. Why is that so important?

Tim Semelroth

Because of a lawyer's duty to keep client information confidential — that's a tenet of the practice of law. With a free AI account, whatever the LLM is, they're going to train on whatever data you upload, and the only way to opt out is to pay. There are also concerns about where that information goes. And in a business where we traffic in words, the idea that your paralegal or associate is not going to try to use AI is delusional — it's going to happen. So one of our hard-and-fast rules is: no client information goes into an AI tool we're not paying for. If you decide there's a task AI can help with, it's non-negotiable — you go to our office manager and get a paid account, so we're doing what we can to keep clients' information confidential.

Collin McKee

Very good information — I bet some people still aren't aware of that. Now, you testified as an expert and used AI to simulate cross-examination. Walk us through how that went.

Tim Semelroth

I've practiced law for 28 years and this was a first. I was hired as an expert witness in a case involving thousands of pages of medical records I had to know. As is common, you write your report months before trial; often the other side then hires an expert, issues a contrary report, depositions are taken, and you get a clear idea of how you'll be attacked. In this case they didn't really hire an expert responding to my opinions, and they chose to take my deposition. I asked the lawyer who hired me where he thought they'd come from, and he wasn't sure. So I took my report, put it into an LLM tool with sufficient confidentiality protections, and asked it to simulate cross-examination — "How would you cross-examine me? What are the weak points? What parts of my opinion are grounds for attacking my credibility?" Those weren't the exact questions I got, but it forced me back into records I hadn't looked at in months.

Then I took my report and all of those simulated cross-examinations and put them into my secure version of NotebookLM, and had it create podcasts. I had to be in court about 45 minutes away on two different days, so each day I drove to the courthouse listening to a podcast based on my report and the cross-examinations. It was a great refresher — even though the heavy reading was almost a year earlier, it was all fresh when I was on the stand.

Collin McKee

It goes back to getting your reps in and keeping it fresh — and using the technology to spar with you. That's brilliant. Looking ahead: what separates plaintiff firms that will thrive over the next five years from those that fall behind?

Tim Semelroth

AI can help you figure out what lane a plaintiff's personal-injury case is in — does it need to be settled, or litigated and potentially tried? — and you can figure that out a lot faster. It gives you the ability to communicate that with your clients sooner. Once you're in that lane, your time is used much more efficiently. Some cases need to be tried, which requires a certain level of commitment and staffing. Others just need to be settled — process the medical records, get the documentation the insurance company needs, and get the demand out as fast as possible. What brings it together is client education: the sooner you figure out each case and have that human conversation — "here's where I think this should go, and we've got a team to get you there" — the better, and that translates into better client reviews, because you're giving them a clear path to their goals.

Collin McKee

Time is a precious resource, especially for those billing by the hour. One final question: if a mid-size firm wants to modernize intelligently, like yours has, where should they start?

Tim Semelroth

Decide whether you'll realistically have the expertise in-house. When you use AI tools, you have to read the user agreements and understand how things are changing. To teach this class at the University of Iowa, I'm listening to podcasts every day and subscribe to four different AI-specific newsletters, because everything changes so fast. So either you have someone in your office who'll do that, or you consult with someone who'll stay on top of it for you. That's the key choice — develop the in-house expertise or hire the expert. What you can't do is ignore it. Whether you like it or not, AI is here, and people are getting in trouble with courts every day because they don't understand how AI tools work.

Collin McKee

Tim, I really appreciate the clarity and the practicality today. For those listening — if you're interested in practical AI implementation inside law firms, not just theory, subscribe and follow. We'll keep bringing on operators who are actually building systems. Until next time, Tim, thank you very much for your time.

Tim Semelroth

Thank you.