IntakeOS is an Independent Showcase Project for service businesses that receive project requests through forms, email, and referrals. It converts an incoming request into a structured opportunity that a human can review and route with confidence.

Client intake often breaks down between the first message and the first useful decision. Important details remain buried in free text, teams evaluate fit inconsistently, high-value requests wait too long, and project creation happens before risks or missing inputs are visible.

I designed and built the complete showcase: workflow definition, information architecture, responsive interface, synthetic data model, API-backed actions, human approval behavior, structured agent tools, failure handling, automated QA, and visual assets.

The system places every request in a shared operational queue. It presents a concise AI-style brief, detected requirements, fit score, qualification progress, risk, budget, timing, and routing recommendation. Operators can request more information or approve the opportunity and create its project state.

For a client-services team, this pattern reduces time spent translating free-form requests, makes qualification criteria consistent, exposes missing information earlier, and keeps responsibility for the final decision with a person. It can be adapted for agencies, consultancies, implementation teams, support operations, or internal project offices.

The application uses Next.js, React, TypeScript, Tailwind CSS, and accessible interface primitives. A local API route validates create, approve, and follow-up actions. Structured browser tools expose request listing and approval through the same application state. All demonstration companies, contacts, budgets, and requests are synthetic.

Structured request capture

AI-style classification and summary

Delivery-fit scoring and qualification checklist

Budget, timing, risk, and requirement extraction

Routing recommendations

Human approval before project creation

Search, status filters, empty state, loading state, and error recovery

Responsive operations workspace

The main product challenge was making the automation useful without implying that a score should make the final decision. The design keeps the recommendation compact, exposes the evidence behind it, and puts the irreversible-looking project action behind a clear human control.

The local application passes lint, production build, WebMCP contract checks, and automated browser acceptance. Verified behavior includes selection, search, filters, form validation, loading, simulated API failure, successful approval, project-state creation, empty state, and desktop/mobile layout. Four upload-ready 16:10 screenshots were generated from the passing build.

The acceptance flow checks the complete path from request review to a stable created-project state. It covers form validation, loading feedback, a simulated API failure and recovery, approval, filters, empty-state behavior, responsive layout, horizontal overflow, and browser console monitoring.

Each state-changing action provides explicit feedback and a clear terminal state. That makes unfinished behavior easier to identify, fixes easier to verify, and the final workflow easier to hand off with concrete acceptance criteria.

I can translate a messy operational process into a clear full-stack tool with automation, structured decision support, safe human intervention, API behavior, responsive design, and repeatable QA.

Next.js 16, React 19, TypeScript, Tailwind CSS, accessible Radix/Shadcn interface primitives, a REST-style API route, WebMCP-compatible structured tools, and Playwright browser automation.