A personal injury firm automated client intake with AI, cutting 12 steps to 4 and reducing follow-up from 48 hours to under 4.
Article text
How a 28-Person Law Firm Cut Client Intake from 12 Steps to 4 (And Stopped Losing Cases to Slow Follow-Up)
The bottleneck wasn't software. It wasn't staffing. It was a filing cabinet in the back hallway with a handwritten label that said "NEW CLIENTS."
When a partner at a 28-person personal injury firm asked us to look at their intake process, I expected to find an outdated CRM or a clunky form. Instead, I watched a paralegal pull a yellow legal pad from under her keyboard, read back scribbled notes from a phone call that morning, then walk those notes to a filing cabinet. From there, another staff member would type the same information into a spreadsheet. A third person would copy it into the case management system. A fourth would run a conflict check. Then someone would route it to the right attorney.
Six handoffs. Three re-entries of the same data. Twelve steps from first phone call to attorney assignment.
And the whole time, the prospective client was waiting.
The Problem Nobody Had Measured
Here's the thing about law firm client intake: most firms don't measure how long it takes because it doesn't feel like a revenue problem. It feels like an admin problem.
This firm hadn't measured it either. When we mapped the full process on a whiteboard, the partners were surprised it took 12 steps. They were shocked when we timed it: a new client who called Monday morning wasn't hearing from an attorney until Wednesday afternoon. Sometimes Thursday.
If you're running a similar process and wondering whether the friction is costing you cases, we do free 30-minute discovery calls where we walk through exactly this kind of diagnostic.
The managing partner's guess was that they were losing "a few" clients to slow response. The actual number was worse. When we pulled three months of intake records and cross-referenced them against signed retainers, the pattern was stark: prospects who received attorney contact within 24 hours signed at 72%. Prospects who waited longer than 36 hours signed at 31%.
Half the firm's lost cases weren't going to better lawyers. They were going to faster ones.
What We Found on the Whiteboard
We use the same process-first diagnostic methodology with every client. The approach is always the same: before we talk about AI, we map the process as it actually exists. Not how the owner describes it. How it actually runs, step by step, with real people doing real work.
Here's what the law firm's intake process looked like:
Receptionist takes phone call, writes notes on legal pad
Notes placed in filing cabinet for intake coordinator
Intake coordinator retrieves notes (sometimes same day, sometimes next morning)
Coordinator enters data into shared spreadsheet
Coordinator copies data into case management system (Clio)
Coordinator emails conflict check request to senior paralegal
Senior paralegal runs manual conflict check across three systems
Paralegal emails conflict check result back to coordinator
Coordinator updates Clio with conflict status
Coordinator emails case summary to practice area lead
Practice area lead reviews and assigns to attorney
Assigned attorney contacts prospect
Twelve steps. Six different people touching the file. Three separate data entry events where the same name, phone number, and incident details get typed in from scratch.
Then we asked the question that changes every engagement:
which of these steps require a human judgment call?
Four. Steps 7 (conflict check review for edge cases), 10 (case evaluation), 11 (attorney assignment based on caseload and expertise), and 12 (the actual attorney-client conversation). Everything else was data entry, routing, and copy-paste.
What We Built
The build took three weeks. Here's what replaced eight of those twelve steps:
One AI intake agent
that does the following:
Captures call details in real time (integrated with the firm's phone system via a structured intake form that auto-populates during the call)
Runs automated conflict checks across all three systems the firm uses
Populates Clio with the complete case record, eliminating all manual data entry
Flags genuine conflict check edge cases for human review
Routes the case to the appropriate practice area lead based on case type, attorney availability, and current caseload
The new process:
Receptionist takes phone call; intake agent captures details in real time
Agent runs conflict check, populates Clio, routes case
Practice area lead reviews and assigns
Attorney contacts prospect
Four steps. One person reviews instead of six touching the file.
We didn't replace anyone's job. The intake coordinator and the paralegals who used to spend hours on data entry now spend their time on substantive case preparation work. The firm actually increased their capacity without adding headcount.
The Numbers
The results showed up fast.
Follow-up time:
Dropped from an average of 48 hours to under 4 hours. Most prospects now hear from an attorney the same day they call.
Intake hours recovered:
19 hours per week across the intake team. That's nearly half a full-time position redirected to higher-value work.
Client retention:
The firm retains 3-4 additional clients per month that they would have previously lost to slow response. At their average case value, that's meaningful revenue.
Error rate:
Data entry errors (wrong phone numbers, misspelled names, incorrect incident dates) dropped by roughly 80%. When you eliminate three rounds of manual re-entry, the math is straightforward.
Process steps:
12 to 4. The filing cabinet is still in the hallway, but nobody uses it for intake anymore.
Why This Matters Beyond One Firm
Personal injury is one of the most competitive practice areas in legal. Response time directly correlates to signed retainers. But the pattern we found here shows up in nearly every professional services firm we've worked with.
The symptoms look different on the surface. An accounting firm has client onboarding spread across email, a shared drive, and QuickBooks. A staffing agency has candidate processing bouncing between an ATS, a spreadsheet, and three people's inboxes. A medical practice has patient intake forms that get faxed, then typed, then typed again.
The underlying problem is the same: information enters the business once and then gets manually re-entered across multiple systems by multiple people, with nobody measuring how long the whole chain takes or what it costs when it's slow.
What This Cost to Build and Run
The three-week build covered process mapping (two days), system integration (Clio, phone system, conflict check databases), agent development and testing, and a week of parallel running where both the old and new processes operated side by side.
Monthly running cost: approximately $280 for the AI agent infrastructure and API calls. For context, the 19 hours per week of staff time this freed up would cost roughly $2,400/month at the firm's loaded labor rate. The ROI math doesn't require a spreadsheet.
The Takeaway
The managing partner told us something after the first month that stuck with me: "We thought we had a marketing problem. We were spending money on ads to get more calls. Turns out we were converting the calls we already had at half the rate we should have been."
That's the pattern. Businesses assume the problem is at the top of the funnel (not enough leads, not enough calls, not enough prospects) when the real problem is in the middle: the process between first contact and meaningful response is too slow, too manual, and too leaky.
You don't always need AI to fix it. Sometimes you just need a whiteboard and an honest look at how work actually flows. But when the bottleneck is data entry, routing, and copy-paste across systems, an AI agent can compress the timeline from days to hours.
If your firm's intake process involves people copying information between systems, or if prospects are waiting more than 24 hours to hear from someone, book a free workflow call . We'll tell you whether automation makes sense for your situation, even if the answer is that it doesn't.