A Kentucky personal injury firm running intake the way most PI firms still do, manually and during business hours.
Grossman Green is a personal injury firm handling inbound leads across multiple PI case types. Like most plaintiff-side firms, the front door of the practice is its intake process. A missed call at 7pm. A poorly qualified lead that wastes a legal assistant's morning. A case-type misroute that delays response. Each one has a real cost, both in lost revenue and in time the firm doesn't get back.
The firm runs Filevine as its case management system, so any AI intake layer had to push clean data into Filevine without creating a parallel system the team would have to reconcile later.
Manual intake couldn't scale, but generic chatbots couldn't qualify.
The intake bottleneck at Grossman Green was structural, not effort-related.
After-hours leads were getting lost. Personal injury prospects don't wait for business hours to start searching for representation, and the firm had no way to capture, qualify, or even acknowledge leads that came in overnight or on weekends.
Qualification was inconsistent. Different intake staff applied threshold questions differently. Some leads made it through that shouldn't have, eating attorney review time. Others got dropped that should have been routed forward.
Generic AI chatbots weren't going to solve it either. Off-the-shelf legal chatbots either ask the same flat list of questions to every prospect regardless of case type, or hand off to a human after a two-line greeting. Neither approach actually qualifies a lead the way a trained intake specialist would.
PI intake isn't one workflow. It's four or five different workflows wearing the same coat. A slip-and-fall qualifies differently than an auto accident, which qualifies differently than a premises case. Anything that doesn't account for that is just a fancy contact form.
A six-agent system that greets, qualifies, routes by case type, and hands off clean data.
CustomAI Studio built a multi-agent AI intake system designed around the actual shape of a PI firm's intake workflow, not a one-size chatbot pattern.
The system runs six agents working in coordination. A main routing agent handles the front of the conversation: it greets the prospect, gathers a plain-language description of what happened, and runs the threshold qualification questions covering injury, fault, timing, and location. As the conversation develops, the routing agent detects which type of personal injury case the prospect is describing and hands the conversation off to one of four specialized case-type agents. Each specialized agent has its own qualification logic tuned to the specifics of that case type. A prospect describing a rear-end collision gets a different conversation than one describing a fall in a retail store. Once a lead clears qualification, a contact collection agent takes over to gather the information the firm needs to follow up.
Qualified lead data is pushed into Filevine, which keeps the chatbot inside the firm's existing operating system instead of creating a separate intake silo. Microsoft 365 integration was delivered alongside the chatbot to bring the system into the firm's broader workflow environment.
Intake that runs while the firm sleeps, with qualification logic that doesn't drift.
Grossman Green now has a 24/7 front door that does the work a trained intake specialist would do. It greets the prospect, listens to the situation, asks the threshold questions, routes based on case type, and captures the contact information for qualified leads. The firm's legal assistants stop being the bottleneck on after-hours inquiries, and the qualification logic stays consistent across every lead regardless of when it comes in or which staff member would have otherwise handled it.
The system runs on a per-case operating cost of $10 to $15 in API consumption, with a monthly maintenance model that keeps the firm's total cost of ownership well below the cost of the headcount it offloads.
Case-type routing is the part that's hardest to replicate with generic tools. By separating the qualification logic into four specialized agents rather than one general one, the system can ask the questions a partner would actually want asked, not a flat checklist that misses the specifics of each case type.
Hear from the firm: an AI legal assistant in production
Partner Abby Green shares her first reaction to Jeannie — CustomAI Studio's AI legal assistant — and where she'd put it to work across a personal injury practice, from intake and demand letters to medical-record analysis and deposition prep.
I was very impressed.
Read the transcript
For the most part, I'm still pretty skeptical about platforms that claim to do AI. One thing I was very interested in about your work is that it's highly specific, highly customized. We were working with one company — they were like, 'We write demands, we do medical chronologies, we do discovery responses' — and then you look at it and it's all very poor.
Maybe a little introduction of who you are and what type of law you practice.
Obviously, I'm Abby Green. I'm a civil litigator in Louisville, Kentucky. Our firm exclusively represents individuals who have been injured, so most of my practice is in litigation — but we really have the full gamut of injuries and types of claims. We love our work and we love our clients.
I want to talk a little bit about Jeannie — with the understanding that you saw it for maybe two minutes, a very limited version. What I showed you briefly was the plaintiff side sending the defense the package, the complaint, and within seconds it was in their case management software, Clio; within about a minute there was a response to the pleading that addressed hundreds of pages. What did you like or dislike about what you saw in that brief presentation?
I was very impressed with it. I certainly understood the appeal for that defense firm — because what are defense lawyers worried about? Getting an answer to a complaint filed. To have that right off the bat is just so fantastic, because the lawyer can do it that day; they don't even think about it anymore. And if we put in our intake information — which I think you could help us refine and then put into our system — and it automatically logged, it would make my legal assistant's job so easy, because it would just be ten documents and nothing would fall through the cracks.
My law partner wisely has a phrase: perfect is the enemy of good. We really try to live that here. There are certain things AI can do that are good enough to keep things moving for the benefit of our clients.
So if Jeannie were available today, what part of your caseload would you delegate to her first?
I think I'd start her out doing the pre-litigation work — which we have an attorney mostly handling — to free that attorney up for higher-level work. The intake process, letter generating, and even analysis of the claims is where I'd see her being very effective. And especially working with you, Ross, there's definitely room in our litigation space to have her focus on projects that require analysis of medical records and synthesizing information — making deposition outlines based on the evidence in the case, that sort of thing AI could do a really nice job with.