AI Due Diligence Machine
Functional DemoA first-pass AI diligence analyst that builds an acquisition risk profile from public data, APIs, and uploaded documents.
In Development
RFI summary generator
AI command center that turns construction RFIs into summaries, risk scores, and executive reports.
What it solves
Construction RFIs are scattered across Procore, Autodesk Construction Cloud, PDFs, email threads, shared drives, drawings, specs, and subcontractor notes. As a result project managers waste time reading long RFIs, teams miss important issues, RFIs sit open too long, duplicates get created, and change-order and schedule risks are caught too late. Executives lack clean project visibility and past decisions are hard to search.
Built for Construction teams — general contractors, subcontractors, project managers, and construction management firms, from small contractors to enterprise.
Overview
The app takes messy RFIs — PDFs, emails, documents, images, or pasted text — and produces clean summaries, extracted RFI fields, risk flags, action items, suggested responses, and executive-ready reports. RFIs live in a project workspace with a dashboard, an inbox with search and filters, and detail pages showing plain-English summaries, key issues, required decisions, and cost/schedule impact. Each RFI receives an explainable 0–100 risk score, and answered RFIs become a searchable project decision log. Reports can be exported to PDF, and the architecture is organized around organizations, projects, and provider-based integrations.
What this demonstrates
AI summarization grounded in uploaded document content with an offline deterministic mock provider so the app runs without a key, plus editable AI output. It builds explainable risk scoring, PDF report generation, file uploads to disk or S3, and multi-tenant organization isolation with custom JWT-based auth.
What makes it different
Not a basic document summarizer — an AI RFI command center that turns messy construction RFIs into clear decisions, risks, and next steps.
- Useful immediatelyThe RFI summary must be useful the moment it appears.
- AI stays groundedAI must not invent unsupported facts.
- Every output editableUsers can edit every AI-generated summary and response.
- Explainable riskRisk scoring must be explainable, not a black box.
- No AI expertise requiredUsers never need to understand AI to use the app.
- Isolated org dataOrganization data must be isolated and uploads secure.
Core capabilities · 10
Project Dashboard
Totals for open, closed, overdue, high-risk RFIs and average response time.
RFI Inbox
The main work area with cards showing status, due date, risk, trade, and next action.
AI Summary View
Plain-English summary, cleaned question, key issue, required decision, and impacts per RFI.
Risk Scoring
Scores every RFI 0-100 from days open, impacts, blocked work, and change-order language.
Executive Summary Generator
Weekly, high-risk, overdue, cost, and schedule reports plus a project health summary.
Duplicate RFI Detector
Detects similar or duplicate RFIs across a project.
Change Order Risk DetectorChange Order Watchlist
Flags RFIs that may become change orders based on risk language.
Smart Response Drafting
Drafts professional responses, clarification requests, and escalation messages.
Spec and Drawing Reference Assistant
Extracts and links drawing sheets, detail references, spec sections, and CSI divisions.
Project Decision Log
Turns answered RFIs into a searchable decision history.
How it works
- Log in and create or select a project
- Upload an RFI PDF, email, document, image, or pasted text
- System extracts the content
- System identifies RFI number, subject, question, due date, party, trade, and references
- AI generates summary, key issue, required decision, cost and schedule impact, and risk score
- AI drafts a suggested next action and response
- Review, edit, export, or share the summary
- System stores the RFI in the project decision log
How the risk score works
Risk score (0-100)
- Days open
How long the RFI has been outstanding.
- Due date proximity
How close or past the response deadline is.
- Trade affected
The trade impacted by the RFI.
- Cost impact
Potential cost exposure of the issue.
- Schedule impact
Potential delay to the schedule.
- Missing drawings/specs
Whether referenced documents are missing.
- People involved
Number of parties in the RFI.
- Blocked work
Whether work is stopped pending an answer.
- Vague response
Whether the response is unclear.
- Related RFIs
Links to similar or connected RFIs.
- Change order language
Presence of language signaling a cost claim.
Key screens
- Dashboard
- RFI Inbox
- High Risk
- Reports
- Projects
- Integrations
- Admin
Data model
- User
- Organization
- Project
- RFI
- RFIAttachment
- RFISummary
- RFIComment
- Report
- IntegrationConnection
Screens · 9
How it is built
A Next.js application with a Prisma data layer over SQLite in development and PostgreSQL in production. Uploaded files are stored on local disk or in AWS S3-compatible object storage, authentication is custom (jose JWTs plus bcrypt), and Anthropic Claude powers summaries with a deterministic offline fallback.
APIs and services
- AWS
- Anthropic
- AWS S3
Classification
- Industries
- Construction
Related work
AI City Expansion Engine
Functional DemoLocation intelligence that scores ZIP codes by industry-specific expansion opportunity.
AI Acquisition Hunter
In DevelopmentAn AI acquisition-intelligence platform that surfaces and scores off-market acquisition targets from public business signals.
Building something with similar AI Automation, Document Generation, Reporting functionality?
Start a project and reference this one — the conversation begins with what you have already seen.
Start a project