Most attorneys who hear “AI document generation” picture a chatbot. You paste in some case facts, the AI writes something, you copy it into Word, fix the formatting, add your letterhead, adjust the signature block, double-check the case number, and save it with the right filename. That is not document generation. That is document assembly with an AI draft step in the middle.
Here is what it looks like when the AI is connected to your practice management system.
What goes in
The AI reads your open matter in Clio. It pulls the case name, case number, court, parties with addresses, opposing counsel, incident date, and assigned attorney. If you have uploaded medical records, billing statements, or correspondence to the matter, the AI has already extracted structured facts from those documents: treatment dates, providers, diagnoses, lien amounts, insurance carriers.
When you ask the AI to generate a demand letter, it does not ask you for the client’s name. It already has the client’s name. It does not ask you for the opposing counsel’s address. It already has the opposing counsel’s address. It does not ask you for the settlement amount. It has calculated a demand range based on your medical specials and the case facts.
What comes out
A finished DOCX file. Not a draft in a chat window. A Word document with:
- Your firm’s letterhead (firm name, address, phone) right-aligned at the top
- The date
- The recipient block (adjuster name, insurance carrier, address)
- A RE line with the case reference
- Body paragraphs with the incident narrative, treatment summary, and demand amount
- The attorney’s name, bar number, and signature
- A filename formatted as
YYYY.MM.DD Case Name - Demand Letter.docx
The attorney opens it, reads it, makes any adjustments, and sends it. The document assembly, the data gathering, the formatting, the letterhead, the signature, the filename convention: all of that is done.
The nine document types
Not every document uses the same pipeline. KrisLegal generates nine document types across three categories.
Settlement documents: demand letters, mediation letters, reports of mediation, and release agreements. Demand letters use a programmatic DOCX builder with full Markdown support in the body text (headings, bold, bullet lists, numbered lists). The other three merge data into your firm’s DOCX template.
Court filings: notices of appearance, notices of deposition, orders of dismissal, and summons. These pull the case number, court, parties, and filing attorney from the matter. Summons uses a separate pipeline that fills an official court PDF form (one per defendant) rather than a DOCX template.
Client documents: engagement letters. These pull the client’s name and email, case details, hourly rates (primary, secondary, associate), and retainer amount from the attorney profile and matter data.
How the template system works
Most documents merge data into a DOCX template using a template engine. The process:
- Validate required fields (the system will not generate with missing data)
- Load the attorney’s profile: name, bar number, signature image, billing rates
- Load the firm’s context: firm name, address, phone, fax, letterhead
- Merge all fields into the DOCX template
- Replace the signature placeholder with the attorney’s actual signature image (200x60px, falls back gracefully if none is on file)
- Format dates in legal style (“17th day of April, 2024”), times in 12-hour format (“9:00 a.m.”), and currency in USD (“$25,000.00”)
- Save with the standard filename format and return a download link
Every document is pre-filled from the matter data. The attorney does not type the case number, the client’s address, or the court name. That information is already in Clio. The AI reads it.
Custom templates
The built-in templates are starting points. Your firm can upload any DOCX template with merge fields, and it becomes a workflow available to the whole team.
Upload a template through Settings or ask the AI assistant to do it. Once uploaded, the AI can generate documents from that template using the same pipeline: validate fields, load attorney and firm context, merge, format, return.
If your firm has a demand letter template that your managing partner spent ten years perfecting, you upload it. The AI uses it. Your output looks like your output, not like a generic template.
Upload, extract, pre-fill
Sometimes you start with a document, not a blank form. Upload an existing complaint or filing (PDF or DOCX), and the AI extracts:
- Case name, case number, court name
- Plaintiff and defendant names
- Attorney names, firms, addresses, emails
- Incident date, service information
Those extracted fields pre-fill the generation form. You review the extracted data, make any corrections, and generate. The AI went from a 30-page complaint to a structured data set in seconds. You went from a stack of paper to a pre-filled form.
What this means for the firm
A paralegal who used to spend 90 minutes building a demand letter now spends 15. The information was always in Clio. The template was always in Word. The attorney’s bar number was always on file. The AI just connects those pieces and produces the document.
Multiply that across every demand letter, every notice of appearance, every engagement letter, every deposition notice your firm produces in a month. That is the capacity gain. Not from working faster. From not doing the assembly work at all.
KrisLegal connects to your practice management system and generates finished documents from your open matters. Schedule a call to see what the output looks like for your practice areas.