SCOPEGUARD
Evidence-backed change orders for client work
ScopeGuard reads the agreement, verifies the relevant wording, and leaves the pricing and final decision with the person doing the work.
Hackathon product overview
SCOPEGUARD
01
PROBLEM STATEMENT
A small request can hide a large scope change
SCOPEGUARD
02
The moment
"One more small thing before launch"
For freelancers, a good response requires finding the relevant clause in a statement of work, separating facts from assumptions, and preserving a record of what changed.
The risk: a vague request can become an untracked commitment.
SOLUTION OVERVIEW
A clear path from request to change order
SCOPEGUARD
03
ScopeGuard links the client message to the confirmed scope before anyone discusses price.
1 Ingest the agreement and request
2 Surface verified quotes and an evidence-based label
3 Let the user confirm the decision and enter estimates
4 Render a consistent, frozen change order
TARGET USERS
Built for client work where scope matters
SCOPEGUARD
04
Freelance web developers
Need a reliable way to respond when an existing client request changes what the agreement covers.
Small agencies
Need a shared record that protects the project team from version drift and accidental commitments.
What they need
A factual answer
that points back to
the agreement
ScopeGuard provides evidence and consistency without turning the system into a legal, pricing, or sending authority.
PRODUCT FEATURES
Trust controls shape every output
SCOPEGUARD
05
Verified citations
Quotes must match the source document before the product displays them.
Evidence rules
Unsupported conclusions fall back to a clarification request instead of a stronger claim.
Human pricing gate
The user enters the hours and approves the final version. The AI has no fields for costs or dates.
TECHNICAL ARCHITECTURE
A web application with a verification layer at its core
SCOPEGUARD
06
Application layer
Next.js 16 App Router, React 19, TypeScript, and server-rendered UI.
Trust spine
Citation verification, evidence rules, a monetary-claim guard, and deterministic document rendering.
Data and integrations
Supabase Postgres with row-level security, private storage, OpenAI, and per-user Slack OAuth.
AI TECHNOLOGIES USED
Structured model output, bounded by evidence
SCOPEGUARD
07
OpenAI runs three controlled operations
Scope extraction
Builds structured scope items from the confirmed source
Request analysis
Compares each client request with supporting evidence
Document drafting
Prepares display-only language that the system checks again
Safety boundaries
Strict JSON schema
Verbatim quote validation
Evidence-rule downgrades
No AI fields for hours, cost, or dates
IMPACT AND VALUE PROPOSITION
More defensible scope decisions, with evidence preserved
SCOPEGUARD
08
Evidence over guesswork
Every displayed claim needs a verified quote or the system asks for clarification.
Consistent approved figures
One calculation produces the on-screen total, client reply, PDF, and invoice.
Current fixture-mode evaluation: 100% citation validation and 0 evidence-rule violations.
FUTURE ROADMAP
ScopeGuard extends its workflow without taking control away
SCOPEGUARD
09
01
Deployment and onboarding
Polish the deployment flow and support per-workspace setup.
02
Honest integrations
Add Gmail drafts, outbound CRM sync, and payment links tied to human-entered amounts.
03
Broader safety evaluation
Expand the labelled evaluation set across models and track evidence-rule violations as a regression gate.
The roadmap preserves the same boundary: the person stays responsible for the decision, the price, and the final send.