Vecflow (now August) helps law firms find and reuse their own past work. Idealite worked with the startup on its early product.
{ Project Overview }
About the Project
Law firms sell expertise, yet much of a firm's own experience is hard to reach. How the firm handled a similar clause or fund term sits in old files, memos and email, while public sources such as SEC EDGAR and US case law are searched separately. That hunt eats into associate time, and confidentiality rules keep many firms away from generic AI tools. Vecflow, a New York legal AI startup now called August, set out to put that experience one question away. Its early product was built to let a lawyer ask, in plain English, when the firm had given tag-along rights to an LP before, and get one answer from firm files and public records, with sources listed. Vecflow also promised that personal data would be redacted and client documents never used to train AI models. Idealite worked with Vecflow on this early product while the company was onboarding law firms through demos and a waitlist.
- Industry
- AI
- Services
Web App Development • AI Integration
{ The Challenge }
A Firm's Experience, Hard to Reach
Key Challenges
- Firm know-how buried in old files
- Slow, manual hunts for past precedent
- Private and public sources searched separately
- Routine drafting eats associate time
- Confidentiality rules out generic AI tools
- Lawyers need answers they can verify
{ Our Solution }
One Question Across a Firm's Past Work
- Plain-English firm search
- One search, every source
- Answers with cited sources
- First drafts from instructions
- Privacy-first hosting choices
{ How We Built It }
Development Process
How the Vecflow engagement came together, step by step.
- 01
Learn How Firms Find Precedent
Learned how lawyers were finding precedent: in document systems, old emails and public filings, each searched on its own.
- 02
Follow Vecflow's Vision
Followed Vecflow's one-question approach, so a lawyer asks what the firm did before and gets a sourced answer.
- 03
Respect Firms' Confidentiality
Treated Vecflow's commitments to firms as fixed requirements: personal data redacted, client files kept out of model training, and hosting where each firm chooses.
- 04
Through Demos and Onboarding
Worked alongside Vecflow during its waitlist and demo phase, while the company onboarded law firms as clients.
{ Tools We Used }
Technology Stack
React
{ Results & Impact }
Delivering Business Value
Precedent on Demand
Lawyers can reuse the firm's past deal terms without hunting through old files.
Time for Clients
Associates start routine memos from a draft, leaving more time for clients.
Ready for Demos
Vecflow could show prospective firms one question answered from firm and public sources.
Trust Built In
Cautious firms can adopt AI with client files kept out of model training.
Technical details
Requirements, architecture, development, testing and deployment notes.
Technical details
Requirements, architecture, development, testing and deployment notes.
Objectives
- Surface Firm Know-How: Help lawyers reuse how their firm handled a clause, deal or fund term before, instead of digging through old files and email.
- One Search, Every Source: Answer one question from a firm's own files and from public sources such as SEC EDGAR and US case law.
- Free Up Associate Time: Turn plain-language instructions into first drafts of memos, contracts and templates, so more lawyer time goes to clients.
- Earn Firms' Trust: Win over confidentiality-conscious firms with answers they can check and data handling their policies allow.
- Be Ready to Demo: Give Vecflow a product it could show prospective law firms as it onboarded clients from its waitlist.
Requirements
- Plain-English questions answered from a firm's own past work: case files, memos, templates and emails
- Connections to the systems firms already use, such as iManage, NetDocuments, Box, Google Drive and Outlook
- Public sources such as SEC EDGAR and US case law searchable alongside private files
- Answers that list the sources they came from
- Drafting of memos, contracts and templates from plain-language instructions
- Personal data redacted, and client data never used to train AI models
- Cloud, self-hosted or on-premise setup to match each firm's data policies
Architecture
- One retrieval layer over a firm's private documents and public legal sources
- Connectors to the document management, file storage and email systems firms already use
- A privacy layer that redacts personal data and keeps client documents out of model training
Technology stack
- React
Front end
- React web interface
- Interface designed around one plain-English question, with sources shown for each answer
- Drafting from plain-language instructions for memos, contracts and templates
Back end
- Connectors that bring in case files, memos, templates and emails from iManage, NetDocuments, Box, Google Drive and Outlook
- Search over public sources such as SEC EDGAR and US case law
- Retrieval that combines private and public results and keeps track of where each answer came from
- Personal-data redaction, with client data excluded from model training
Deployment
- Cloud, self-hosted or on-premise options, so each firm can keep its data where its policies require
Challenges & solutions
- Scattered knowledge: a firm's precedent sits across document systems, email and file shares, each with its own search.
- Two kinds of source: private firm files and public sources such as SEC EDGAR and case law had to answer the same question together.
- Trust: lawyers need to see where an answer came from before they rely on it.
- Confidentiality: firms cannot risk client documents training AI models or leaving approved hosting.
- Early stage: the product had to be ready to show in demos while the company onboarded clients from its waitlist.
Outcomes
- Firm memory on demand: lawyers can ask what the firm did before and get a sourced answer.
- One search: firm files and public legal sources answer the same question.
- Drafts on request: memos, contracts and templates start from plain-language instructions.
- Confidentiality built in: personal data redacted, no training on client data, and a choice of hosting.
Next case study
Helping Atlassian Ship Confluence and Jira Changes With Confidence
{ Let's Build Together }
Let's build what's next
We partner with ambitious businesses to design, build and scale digital products that create lasting value.
Contact Us