Step by step: how a matter folder becomes a record ready to file on your network.
DilloLex is a small, secure computer that runs your firm's document automation pipeline locally, inside the storage folders you already use. No portal, no uploads, nothing new to learn. Below is exactly how a matter flows through it, where your data goes at each step, and how your staff stays in control of every output.
Two steps. Then it runs in the background.
We install the appliance in your office
A standardized mini-computer running the DilloLex stack. It sits on your local network with no inbound access from the internet, a drive encrypted at rest, and automatic security updates. Your firm owns the hardware outright.
You connect the folders you already use
DilloLex connects to the document storage you already use: Google Drive. No separate AI account, no per-seat subscription, no usage meter. The AI runs on the appliance, so the only login your staff needs is the one for the storage they already have.
It works inside the folders you already use.
Your team keeps saving documents exactly where they do today. DilloLex watches that folder and takes it from there. No portal, no uploads, nothing new to learn.
Save to the matter folder
Your staff drops the client's files into the matter folder in the cloud storage your firm already uses. That's it. It's the step they already do.
The appliance grabs them automatically
It watches that folder over your network and pulls each new file the moment it lands. No upload, no button to press.
It reads, identifies & catalogs
OCRs every page, figures out what each document actually is, proposes a clean name, and organizes the whole set into the matter.
It stamps, flags & assembles
Bates numbers the set, flags PII/PHI for redaction, and builds the disclosure, exhibit, or demand package the matter needs.
Organized work lands back in the folder
The finished, named, ordered work product appears in a "Ready for review" folder, right where you already work. Your paralegal and attorney approve it. Nothing is final without a human.
What happens to a stack of documents.
Each step runs on the appliance. Nothing is finalized without your approval.
1 · Read every page (OCR)
Scanned PDFs, photos of screenshots, and image-only documents are converted to searchable text, in English and Spanish, so the pipeline works with what each document actually says.
2 · Classify & auto-name
DilloLex identifies what each document is (a 2020 tax return, a bank statement, a text thread between the parties, a medical bill), pulls the dates, parties, and amounts needed to name it well, and proposes a clean, consistent name with the source line your staff can verify at a glance. We tune this to your firm's own documents during onboarding.
3 · Organize the matter
Documents go into a structured matter record. Duplicates are flagged, low-quality scans are flagged for a human, and documents carrying PII/PHI are surfaced for extra care.
4 · Detect & redact PII / PHI
Social Security numbers, account numbers, and protected health information are detected and surfaced. Your staff authorizes the redactions; the appliance burns them in; a quality check flags likely misses for human review. Final verification stays with your staff. No tool guarantees complete removal.
5 · Bates stamp & assemble
Indexed exhibit sets, NRCP 16.1 / 16.2 / 16.205 disclosure compilations, and PI demand and litigation packages are assembled, numbered, ordered, and ready for review. Or let a draft drive it: hand DilloLex your motion and it reads every document the motion references, finds each one in the matter record, and assembles a Bates stamped, indexed exhibit set in the order you cite them. You confirm each match; anything it cannot find with confidence is flagged as unmatched for you to resolve.
6 · Flag deadlines for your docketing
The appliance proposes rule-based dates and publishes them to a subscribable calendar feed for your docketing process to verify. It never calculates a date silently. Every date is a draft to check against the current rule. On Nevada matters it reads the triggering date out of an order or filed document (the entered date on an order, the service date on a notice) and proposes the resulting deadlines as drafts.
A folder named "New folder" becomes a record you can file.
Every firm has the folder: cryptic names, three copies of the same draft, a 16 GB zip nobody has opened. DilloLex reads all of it, proposes a clean structure, and preserves every original untouched, for your staff to approve.
From a real Nevada firm's case folder. The client name and file names have been changed for confidentiality. Counts are illustrative.
Court documents go date-first. Every pleading renamed with its date, so the folder sorts itself in order.
Photos get described, not just dated. The on-box vision model reads each photo and names it for what it shows (a damaged bumper, a bedroom, a text-message screenshot), so DSC1234.jpg becomes something you can scan. Text exports keep their names. Anything low-confidence waits in a review queue.
Every page is read and searchable. OCR turns scans, screenshots, and that zip into plain-language search across the whole matter.
Every original is preserved untouched. The "_Originals (as received)" folder keeps your file exactly as it came in, nothing renamed or moved in place.
Grounded in Nevada law. In your firm's format.
The drafting engine draws from two sources: your firm's own document templates and Nevada law. A draft motion starts in the format your firm already uses and is grounded in Nevada authority. Every draft is AI-assisted and attorney-reviewed. The attorney approves the substance before anything goes out.
Motions
Trial and hearing motions drafted from the firm's own prior motions and templates. The engine reads the matter record, finds the supporting documents, and assembles a draft grounded in how your firm writes. Your attorney revises and approves.
Appellate briefs
Opening briefs and answering briefs, produced alongside the firm's attorneys on real Nevada appeals. The engine grounds each brief in the record, the firm's prior appellate work product, and Nevada authority. Every argument is reviewed by the attorney before the brief goes out.
Record cite checking (decision support): on a respondent-side brief, the engine ingested the record, generated accurate pinpoint citations to the appendix, and flagged several of the appellant's pinpoint citations that did not support what the appellant claimed was at those pages. The attorneys reviewed and confirmed the discrepancies. The engine flags candidate issues; the call stays with the attorney.
Pre-filing candor check
Before a brief goes out, DilloLex reads the draft for the mechanical candor problems that survive a read-through: a quotation that does not match the opinion's words, language taken from a dissent but cited as the holding, a citation it cannot verify against the record or a public source, and an authority cited in the body but missing from the Table of Authorities. It flags each one for the attorney, who fixes it or records an explicit override. It checks; it does not file, and it does not overrule your judgment. More on the pre-filing check →
Appellate appendices
NRAP 30 appendices compiled from the record: the engine reads the brief, identifies every record cite, finds the document in the matter record, and assembles a Bates stamped, indexed appendix. Unmatched cites are flagged for you to resolve.
Demand packages
PI demand letters, medical-record summaries, lien letters, and HIPAA / records-request authorizations drafted from the firm's own templates and matter record. Your attorney finalizes before anything goes to the other side.
Disclosures and exhibit sets
NRCP 16.1 / 16.2 / 16.205 disclosure compilations and Bates stamped exhibit sets assembled from the matter record. Hand it a draft motion and it builds the exhibit set from the documents the motion actually cites.
Your firm's format throughout
Every draft lands on your pleading paper, in your font, using your standard language. Onboarded from your own samples during setup. The finished product reads as yours, in your firm's own voice.
Litigation Intelligence
DilloLex builds two memos from our own corpus of appellate decisions. The judge memo shows what facts, exhibits, and arguments moved the judge, which errors led to reversal, and the judge's affirmance and reversal rates in our corpus. The counsel memo surfaces the arguments and authorities your opponent reuses across cases, so your opening motion can address them first. Every finding cites a source document and page and is flagged for the attorney to verify. You read the memo, check it, and apply it. It supports your judgment; it does not predict the outcome.
See which practice areas DilloLex supports: Practice areas →
Then ask it anything, in plain language.
Once every page is OCR'd and indexed, the whole matter becomes answerable. Instead of opening file after file looking for the one line you remember, your staff asks the way they'd ask a colleague: "didn't Jane say she'd sell the truck?" or "was there a transfer to John out of the joint account?" DilloLex reads for meaning, going beyond keyword matching, and returns the exact passage with the document and page it came from. Open the source, read it in context, verify it yourself. The search runs on your documents, on the appliance. The fact comes straight from your file. Nothing is invented.
All the AI is built to run on the box your firm owns.
Every part of the AI pipeline is designed to run on the appliance your firm owns. OCR, Bates stamping, redaction detection, document assembly, drafting, and analysis all run on the box in your office. By design, no client or confidential content is ever sent to any outside AI service. No "bring your own AI key," no outside model in the loop.
AI on the appliance
The drafting and analysis models run on the appliance itself, alongside the document pipeline. Client content stays on hardware your firm owns. Nothing is handed to an outside AI service.
No outside AI account
Because the reasoning runs on the appliance, there is no Anthropic or other outside AI bill, no API key to manage, and no per-token meter. The only accounts you keep are the storage and tools you already use.
Connects to the tools your firm already uses.
Storage
Google Drive is connected today. The appliance reads the matter folder and writes finished work product back where you keep it.
Practice & calendar
DilloLex publishes a subscribable calendar feed (.ics) you can add to Outlook, Google Calendar, or Clio today.
Research, your choice
DilloLex is a document appliance that leaves legal research to the tools you already trust. Keep using Westlaw, Lexis, or whatever you prefer. DilloLex works alongside them.
Don't see your stack?
We add connectors based on what our firms actually use. Tell us what you run and we will tell you where it fits. Get in touch →
See it on your own documents.
The fastest way to understand what DilloLex does is to watch it work on a de-identified file from your own firm.
Book a demo