Workflow Audit App for Small Businesses

AI Prompts to Build This

Copy these build prompts into Claude, Cursor, or your AI coding tool. Create a free account to unlock the full research behind them.

1. 1. Project Setup

   Build the weekend MVP of "Firstflow": a small-business owner describes their tools, hand-offs and the hours each chore takes, and gets a visual map and a ranked list of what to automate first. The hours saved come from the owner's own numbers.
   
   Stack: Next.js (App Router, TypeScript), Tailwind, Supabase (Postgres, Row Level Security, Auth with email magic link), the Anthropic API to read the description and score each idea, React Flow to draw the map. Deploy on Vercel.
   
   Tables (Row Level Security on, each user reads only their own rows):
   - audits(id, user_id, business_name, vertical, raw_input, status, created_at)
   - nodes(id, audit_id, label, tool_name, data_sensitivity) data_sensitivity is none, pii or payments
   - edges(id, audit_id, from_node, to_node, handoff_type, is_manual boolean, hours_per_week)
   - opportunities(id, audit_id, title, roi_score, risk_score, complexity_score, rationale, hours_saved_est)
   
   Screens: /login, /audit/new (the guided interview), /audit/[id] (the map and the ranked list), /audit/[id]/brief (a printable page).
   Env vars (names only): NEXT_PUBLIC_SUPABASE_URL, NEXT_PUBLIC_SUPABASE_ANON_KEY, SUPABASE_SERVICE_ROLE_KEY (server only), ANTHROPIC_API_KEY.
   
   Do not build: billing, plans or an agency tier, a PDF library, Zapier, Make or n8n blueprint export, account connections, a screen-share reader.
   Done when: npm run dev starts, you can sign in, and the four tables exist with Row Level Security on.

2. 2. Core Feature

   Build the one feature that proves Firstflow: a map the owner recognizes and a shortlist whose numbers they can trace.
   1. The interview asks which tools the business uses, where information passes from one tool to another, who does it (a person or an automation) and how many hours a week each manual hand-off takes. Save the free text and the hours the owner typed.
   2. POST /api/audit sends the description to the Anthropic API with a strict JSON schema { nodes: [{ label, tool_name, data_sensitivity }], edges: [{ from, to, handoff_type, is_manual }] }. Check in code that every edge refers to a node that exists and that every sensitivity is from the fixed list. Drop anything that fails.
   3. Join each manual edge to the owner's hours for it. Ask the model for { opportunities: [{ title, roi_score, risk_score, complexity_score, rationale }] } with each score from 1 to 10, one per manual edge or group of edges.
   4. Rank in code by roi minus risk minus complexity. Compute hours_saved_est only from the owner's typed hours, as the hours of the hand-offs the idea covers, and never from a model guess. Say that it is the owner's own estimate.
   5. /audit/[id] draws the map with React Flow: tools as nodes, manual hand-offs in a warning style and sensitive flows badged. Under it, show the ranked cards with the score breakdown, the rationale and the hours.
   6. /audit/[id]/brief is a printable page with the map, the ranked list, the hours and a data-sensitivity summary.
   Rules: never state a saving the owner did not give. Say that scores are a model's judgment to discuss, not a measurement.
   Empty state: with no audits, show the interview and a sample description.
   Done when: a sample description produces a map whose edges all join real nodes, the ranking equals roi minus risk minus complexity, an hours figure traces to the owner's input and the brief prints on one page.

3. 3. Landing Page

   Build a one-page landing site for Firstflow, a workflow audit for small businesses before they hire an automation consultant.
   Hero: "See how your business really runs, then decide what to automate." Sub: "Answer a short interview. Get a map of your tools and hand-offs and a ranked list of what is worth automating first." One button: Join the waitlist.
   Sections: a sample map with two manual hand-offs highlighted, how it works in four steps (describe, map, rank, export a brief), and an FAQ on where the hours come from (your own numbers), how scores are judged and what data is stored.
   Waitlist: store the email in a waitlist table in Supabase. No other service.
   Style: Geist, an off-white background, near-black type, one orange accent used only for manual steps.
   Done when: the page renders on a phone and a submitted email appears in the waitlist table.

4. 4. Branding Package

   Use a design or image tool for this one. A coding agent cannot draw a logo.
   Brand for Firstflow: a wordmark and an icon that suggest three nodes joined by one line that bends.
   Colors: off-white, near-black and one orange used only for manual steps. Type: Geist, with tabular numbers for hours and scores.
   Deliverables: wordmark, icon, a map node style with a data-sensitivity badge, three sensitivity badges (none, personal data, payments) that differ by shape as well as color, and one launch graphic.
   Done when: each deliverable is saved in one folder and the sensitivity badges are distinguishable without color.