Case Study | Consumer Bankruptcy, Everett, WA

How a consumer bankruptcy firm in Everett, Washington went from 14 paid-in-full clients per month to 62

Six months. From the firm's all-time best of 10–14 paid-in-full clients per month up to 62. From 20–25 retained up to 50. No new paralegals or attorneys. No replaced software. Just Chapter 7 volume the existing team could finally handle.

The Challenge

A Chapter 7 volume practice running on pen, paper, and the paralegal's memory

The supervising attorney had built the firm on a stack that worked when the practice was small: a decade of OneDrive folder habits, Clio as system of record, emails filed by hand, everything important kept in the paralegal's head. Tracking what was missing for which client was, for a long time, literally pen and paper. Clio tasks didn't fit how the team actually worked. The supervising attorney kept a spreadsheet. That was the system.

Chapter 7s were the profitable part of the practice, so profitable that toward the end of the engagement the firm stopped accepting new Chapter 13s entirely. But Chapter 7 volume has a structural ceiling, and the ceiling is one paralegal's day. Every new client lands a half-complete document dump in the first 48 hours: rotated phone photos of pay stubs propped on a leg, screenshots of bank statements, files named IMG_4027.pdf. Auditing one client's folder against the checklist, flagging what's missing, and chasing down the rest takes 60 to 90 minutes. Six clients a day is a hard ceiling, and there are always more than six clients.

The firm needed more capacity without more headcount, and without anyone giving up the tools they already trusted. The real goal wasn't filing faster. It was being able to audit every paid-in-full client's file at once, and reach out proactively, which is impossible when the data is scattered across three systems and one person's memory.

Our Approach

We didn't build from the outside

Joseph joined the firm as a working paralegal, not as a consultant watching from the corner. Forty hours a week prepping petitions in Jubilee, chasing pay stubs, doing the audit-then-reaudit loop, staying past 5:30 to clear the 4 PM email pile. Felt the pain. Then automated it, layer by layer. Most automation vendors have never sat in the seat.

The Infrastructure

Six layers. Infrastructure first, agent last.

AI is the cherry on top after a lot of unglamorous hoops are jumped through. Most of what we shipped at this firm was infrastructure work: real intake forms, a real file home, sync, data centralization. Only after all of that does an AI agent on top actually have something to work with.

IT Infrastructure & M365 Migration

From personal accounts to a real file home

The team had been running on personal OneDrive accounts. Fine when the firm was the supervising attorney plus a few clients, structurally a ceiling on anything we wanted to build later. We migrated to Microsoft 365 Business with proper OneDrive, set up identity management, locked down sharing, and gave the firm an actual file home that automation could attach to.

Clio ↔ OneDrive Sync

One source of truth instead of two

The lead attorney had used the same OneDrive folder layout for over a decade. The case manager lived in Clio. We built a real-time, bidirectional sync so both stay in lockstep. That ended the "two systems, two truths" problem and gave every downstream tool one reliable place to read from.

Automated Email & Attachment Logging

Every email in the right matter

Inbound and outbound email plus every attachment now routes to the correct client matter automatically. That fixed the team-A-team-B problem on its own: a client emails one paralegal Monday and another Tuesday, and the second one can actually see the first conversation, instead of relying on what was kept in the first paralegal's head.

Data Centralization (DynamoDB)

One queryable store across all systems

Data was scattered across Clio, OneDrive, and email. We consolidated the operational state into a single AWS DynamoDB table with caching for fast retrieval, so every downstream agent and pipeline reasons against the same picture of each client.

Smart Document Management

Cryptic filenames, named and sorted automatically

Client uploads land as IMG_4027.pdf and scan_003.jpg: phone-camera saves of pay stubs propped on a leg, screenshots of bank statements. The system reads each file, classifies it (pay stub, tax return, bank statement, ID), renames it predictably, drops it into the right folder, and (importantly) flags what's still missing instead of waiting for a deadline to surface it.

Automated Client Re-engagement

Drip campaigns that bring people back

Clients who missed meetings and clients who received discharge get automated follow-up sequences: the same multi-step chase the firm had been running by hand for clients who went silent (paralegal email, attorney email, case manager call, supervising attorney letter), now triggered automatically from the data layer instead of someone remembering to start it.

One file, end to end

What happens when a client emails a phone photo of a pay stub at 7:42 PM

This is the actual path a single attachment takes through the system at this firm. No paralegal touches it until the very last step. And even then, only to click send.

  1. 1. The photo hits the inbox. The cloud system gets notified the instant it lands.
  2. 2. The text of the email gets logged with Clio under the right matter (the system knows which client, because the sender's email is on a matter). The attachment goes into the Documents tab.
  3. 3. From the Documents tab the file syncs to OneDrive into a folder called client documents inbox.
  4. 4. If the file is a PNG, it gets converted to PDF. Then a model reads it, deduces what it is, and renames it: pay stub, tax return, bank statement, ID.
  5. 5. An agent reviews the client's folder against what's tracked in the DynamoDB table. If new pay stubs were added but there's a one-month gap in the six-month window, it drafts a follow-up email asking the client whether the gap is correct.
  6. 6. It sets an all-day Google Calendar task for the paralegal and fires a desktop notification, so the paralegal can finish whatever they're doing and come back to it.
  7. 7. The drafted email was trained on 1,000+ real emails the paralegal had sent to clients, so it sounds like them, and got better than them at noticing the kind of detail clients tend to ask follow-ups about.
  8. 8. The paralegal reads the email for thirty seconds and clicks send. Or doesn't. The agent never sends on its own. That is a line we don't cross.

Two areas where the system is intentionally not autonomous: it never sends official communications to clients, trustees, or courts on its own. A human always clicks send. And it never permanently deletes a file. Files can be soft-deleted to a trash location with a recovery window, but the irreversible step is a human one. We sacrifice a little throughput so a model error never embarrasses the firm in front of a client or a court.

Parallel processing

Multiple client files audited at the same time. What used to take 60–90 minutes per client takes about ten, and doesn't multiply with volume.

Invisible to clients

No client ever noticed. The system learned from a thousand real emails and sounded like the paralegal who used to write them. Clients thanked us for the detailed responses.

Human in the loop

The system sets tasks for humans, waits for confirmation before official sends, and improves with every correction. Full control, none of the busywork.

Before & After

From cryptic filenames to a folder that knows what's missing

Before Lexera

478329748932.jpg

scan_003.pdf

IMG_4027.pdf

document(1).pdf

“Is the tax return in here somewhere? I'll check after the deadline.”

After Lexera

Pay Advice | Nov–Dec

Bank Statement | Q4 2025

Petition | Ch. 7 Filed

⚠ Missing: Tax Return 2024

“Hi Jane, we're missing your 2024 tax return. You can upload it at…” (drafted automatically, sent with one click)

Timeline

Six months, from sitting in the seat to peak Chapter 7 volume

November

Joseph embeds, workflow diagnosis begins

Joseph joins the firm as a working paralegal, not as a consultant, as one of five people on the team. Forty hours a week prepping petitions in Jubilee, chasing pay stubs, doing the audit-then-reaudit loop. Maps where time is actually being lost.

December

Email logging + M365 migration

Automated email and attachment logging into the right matter goes live. M365 migration begins; the firm moves off personal OneDrive accounts onto a business tenant.

January

Sync + data centralization

Clio ↔ OneDrive sync deployed. The two-truths problem is solved. Operational data centralized into DynamoDB with caching. Drip campaigns launch for missed meetings and discharged clients.

February

Document agent goes live

With clean data underneath, the agent on top is finally worth running. It watches the inbox and the documents folder, names files, notices missing items, and drafts client follow-up emails in the paralegal’s voice for human review.

March

Peak month

62 paid-in-full clients. 50 retained. The system handles volume that would not be possible for any single paralegal manually. Not faster work, parallel work.

Results

Chapter 7 volume without a single new hire

10–1462

Paid-in-full clients per month

20–2550

Retained per month

~5x

Paid-in-full growth vs. all-time peak

0

New paralegals or attorneys

~10min

Per-file audit (was 60–90 min)

“Before” numbers are the firm's all-time best months, cherry-picked and averaged. Not a bad baseline picked to look good.

Metrics reflect Lexera's account of this engagement with a consumer bankruptcy firm in Everett, Washington and are not independently audited. Individual results vary.

Running a Chapter 7 practice that's hitting the same wall?

Fifteen minutes. We'll walk through how documents, email, and case files actually move through your firm and whether the paid one-week diagnostic sprint is the right next step.

Book a 15-minute fit call