ITN Ops Dashboard

An AI-assisted operations dashboard built for PT Internasional Teknik Nusantara, bringing revenue, invoices, project progress, tenders, certificates, tax deadlines, and contracts into one mobile-friendly workspace. Google Sheets remains the system of record; Google sign-in preserves viewer-scoped access, changes are audited, and AI summaries or document drafts stay subject to human review. The production system and records are company-internal; a public stakeholder demo runs the same business rules on a fictional company with a Kimi-powered advisor.

Proof points

  • 8 Operational Areas
  • 89 Automated Tests
  • 2-Day Delivery

Technologies

React 19, TypeScript, Node.js, Google Sheets, Google Drive, Google Cloud Run, Cloudflare Workers, Claude Code, Apps Script, Vercel, Kimi API

Links

Problem

Operational information was spread across spreadsheets, shared-drive documents, and email. Leadership needed a clear view of projects, revenue, deadlines, and company knowledge from a phone or laptop.

Approach

Kept Google Sheets as the system of record and built a React and TypeScript dashboard with a Node.js service, viewer-scoped Google Workspace access, audited edits, project logs, and a no-tools AI advisor. AI-generated summaries and document drafts stay subject to human review.

Results

Delivered an internal dashboard spanning eight operational areas in two days. A public stakeholder demo at itnops.shiverion.com runs the same business rules on a fictional company; production records and the private pipeline source are not exposed.

ITN Ops Dashboard: AI-Assisted Operations for an EPC Contractor

Overview

ITN Ops Dashboard brings eight operational areas into one desktop- and phone-friendly workspace: revenue, invoices, projects, tenders, certificates, taxes, contracts, and company knowledge. It was built around the company's existing Google Workspace workflows so the team could keep its familiar source of record.

Product Design

  • One operating view: financial summaries, project status, tender deadlines, certificate renewals, tax tasks, and contracts are reachable from one dashboard.
  • Project continuity: project pages combine current status, next milestones, activity logs, and shared-drive files.
  • AI with human review: the knowledge advisor answers from permission-scoped records; file-based summaries and drafts are confirmed by a person before they are saved.

Trust and Access

  • Google sign-in is checked on server requests, and sheet reads and writes run as the signed-in viewer so existing Google sharing rules continue to control access.
  • Edits are validated and recorded in an audit log; spreadsheet formulas are not accepted as user-entered text.
  • The AI runs without tools or direct write access. Uploaded documents are treated as content, and business-critical outputs require human review.

Stakeholder Demo

A self-contained demo runs at itnops.shiverion.com with a fictional company (demo login required):

  • No database: six projects with activity logs, about two years of invoices, taxes in every state, tenders, certificates, and contracts are generated in the browser relative to today, so deadlines and alerts always look current.
  • Same business rules as production: edits, deletions, and log entries go through the real shared planner and validators, and the screens are built by the same payload builder. Each visitor's changes stay in their own browser, with a reset option.
  • AI advisor on the Kimi API: answers stream back in the production format; the key stays on the server, and answers are capped per session. Uploads and drafting from files are simulated, so no real document is sent to an AI.

The linked repository contains the dashboard and its server; the knowledge pipeline, Apps Script admin project, and deployment configuration remain private. All eight portfolio captures retain the original screen layouts with invented records, dates, and amounts.