A high-volume law firm was spending 15—20 minutes per intake manually reading emails, downloading attachments, and typing data into Clio. We replaced the entire process with a pipeline that reads the email, extracts structured case data with AI, builds a Google Drive case folder, files all attachments, and creates a verified Clio matter — without a human touching anything.
Legal intake is supposed to be where a client relationship begins. In practice, for this firm, it was where time went to die. Every new client inquiry arrived as an unstructured email — no consistent format, no guaranteed fields, no way to process it without a human reading it first.
Staff would open the email, manually extract the relevant details, download any attached documents, check whether the client already existed in Clio, and then type everything in by hand. Twenty minutes per intake, multiplied across dozens of inquiries a week. And that's when it went right. When it went wrong — a typo in a name, a missed attachment, a duplicate contact — the errors compounded downstream.
Names misspelled, phone numbers mistyped, emails entered incorrectly. Every manual keystroke was a point of failure that created broken records further into the matter lifecycle.
Staff regularly created new Clio contacts for clients who already existed. Case history became fragmented across multiple records — no single source of truth for a returning client.
PDFs and supporting documents sat in Outlook inboxes rather than being filed against the matter. Attorneys searched for documents that had technically arrived but were never properly stored.
Intake only happened when someone was in the office to process it. A client who emailed at 9pm on Friday waited until Monday morning — a 60-hour gap before the firm even knew they existed.
The moment an email arrives in the Outlook inbox, the pipeline runs. Here's exactly what happens.
Zapier listens to the firm's Outlook inbox. The moment a new email lands, it captures the sender name, sender address, subject line, body content, and any attachments. Nothing is touched manually. The pipeline runs whether it's 2pm Tuesday or 11pm Sunday.
The full email — sender details, subject, and body — is passed to GPT-4o-mini with a strict extraction prompt. The model is instructed to return a JSON object only, with no preamble or explanation. Six fields are extracted every time, regardless of how the client phrased their message.
Two formatter steps run in sequence. First, a regex pattern extracts the Case Summary field cleanly from the AI's JSON output — ready to be written to a document and a CRM note. Second, the client's email address is normalised from a line-item format to a plain string. This normalised email becomes the consistent identifier used in every downstream step: folder names, Clio searches, document titles.
A new folder is created in the firm's Google Drive, named using the pattern [client email] — [received date]. Every case gets the same structure, every time. No naming inconsistencies, no folders in the wrong place, no manual filing.
Any attachments from the original Outlook email are uploaded directly into the newly created case folder. PDFs, Word documents, images — whatever the client sent — are moved from the inbox and filed correctly without anyone touching them. The attorney opens the folder and the supporting documents are already there.
The case summary extracted by the AI is saved as an actual Google Doc inside the case folder, titled "Case Summary - [client email]." When the attorney opens the folder, they don't read the original email thread to understand the situation — they read a clean, 2—3 sentence brief written by the AI. Context is there before they've opened Clio.
Before anything is created in Clio, the system searches by email address. If a contact already exists — a returning client, a previous inquiry — their record is returned and used. If not, a new contact is created. This search-or-create pattern is the duplicate prevention mechanism. It's not a check that happens sometimes. It happens on every single intake, every time.
Every duplicate contact in Clio means fragmented case history — two records for the same client, with matters and notes split across them. Attorneys lose context. Billing becomes complicated. The search-or-create step costs milliseconds and prevents a problem that used to take hours to clean up manually.
A new matter is created in Clio, linked to the contact, assigned to the correct attorney group and practice area, with the case summary embedded in the matter's custom fields. The attorney opens Clio and the matter is already structured — they didn't create it, they didn't assign it, and they didn't have to type a single word.
The final step writes the AI-generated case summary as a note on the matter in Clio, timestamped with the original email's received date and time — not the automation's run time. The audit trail is accurate. Open the matter, open the notes, and the intake brief is there exactly when the client first made contact.
The same intake process that once required a trained staff member, an open inbox, and 15—20 minutes of careful manual work now runs in under 30 seconds without human involvement.
| Metric | Before | After |
|---|---|---|
| Time per intake | 15—20 minutes of manual processing | ~30 seconds — fully automated end-to-end |
| Data accuracy | Human error risk on every keystroke | 100% normalised — AI extracts, formatter cleans, no typing involved |
| Duplicate client records | Regular occurrence — no dedup step | Zero — search-or-create runs on every intake without exception |
| Attachment filing | Left in Outlook — frequently lost or missed | Auto-uploaded to the correct case folder in Drive immediately |
| Case brief for attorneys | Read the full email thread to get up to speed | AI-written Google Doc summary waiting in the case folder |
| After-hours intake | No processing until the next business day | Processed immediately — 24/7, including weekends |
Each tool handles exactly one responsibility. The pipeline is linear and auditable — if any step fails, it's immediately clear where and why.
The pattern — AI extraction from unstructured email, structured file creation, CRM deduplication — applies across any professional services firm where client intake arrives via email and ends up in a case management system.
High volumes of client inquiries with complex supporting documents. AI extraction pulls visa type, country, timeline, and urgency. Documents are auto-filed by client. No intake form required from the client.
New client onboarding via email — company details, filing deadlines, tax year, service required. AI structures the data, creates a client folder in Drive, and opens a job in the practice management system.
Policy inquiry emails arrive in varying formats from different sources. AI extracts coverage type, asset details, and contact information. Client record created or updated in the CRM before a human reads the email.
If your team is reading emails, downloading attachments, and typing into a CRM by hand — this system eliminates that entirely. It works while your office is closed and makes no typos.
Book a Discovery Call// 30 minutes — No pitch — Just your situation and whether we can solve it