Kirk Thompson Jordan Thompson
Hire Us Immediate start
All work

AI Copilot · 2024

04

Vanta
CoPilot

LLM Product Internal Tool Development
Vanta CoPilot interface

AI that works
where lawyers work.

A 120-lawyer firm was piloting AI tools — but none of them worked inside their existing document management workflow. Associates were copy-pasting between tools, breaking their concentration and creating security risks. The firm needed AI that lived inside their systems, not alongside them.

We built Vanta CoPilot — an AI assistant embedded directly into their document management platform. It reads context from the current document, surfaces relevant precedents, flags risks, and suggests language — all without ever leaving the workflow the team already knows.

Client type Legal / Professional Services
Engagement Scoped build · 6 weeks
Services Development + Consultancy
Delivered Q3 2024

60% of time
in document review.

Associates were spending the majority of their billable hours reading documents for risk, precedent, and inconsistency — work that required significant concentration but was largely pattern recognition. Existing AI tools required context to be manually provided, meaning associates spent as much time managing the AI as benefiting from it.

Context switching

Every AI interaction required leaving the document, opening a separate tool, providing context, and re-importing results — breaking flow and costing time.

Data security risk

Copying client document content into external AI tools created compliance risk — blocking broader AI adoption.

No institutional memory

12 years of precedent documents existed but were effectively unsearchable — junior associates couldn't leverage the firm's expertise.

Embedded AI.
Zero context switching.

Vanta CoPilot runs inside their document management platform via a secure plugin — with read-only access to the current document and a private RAG index over their precedent library. Associates get AI-powered review, risk flagging, and language suggestions in a sidebar that never requires leaving the document.

In-platform AI sidebar

Secure plugin architecture — AI accesses document context directly, no copy-paste, no external data transfer.

Precedent RAG system

12 years of case documents indexed and searchable by semantic similarity — surfacing relevant precedents in context.

Risk and language engine

Domain-fine-tuned model identifying risk clauses, inconsistencies, and suggesting firm-standard alternative language.

65%

Faster document
review

40%

Fewer review
errors

98%

User
adoption

6wks

From brief
to live