The proposal system Axera Flow uses to quote its own deals

The problem

For Axera Flow's first year, every proposal was a manual Google Doc — copy a template, swap names, rewrite the summary, tune scope, fight the formatting, export to PDF. Two to four hours per quote. The bottleneck wasn't the writing; it was the editorial work of sounding on-brand and specific without sounding like a template. For an AI consultancy, the proposal *is* the work sample.

The solution

A Next.js app where Claude drafts six narrative sections against a structured brief — brand-voice rules enforced in the prompt, no "in today's fast-paced world" filler. The web preview and the PDF are pixel-for-pixel identical (Puppeteer over the same Tailwind tokens). Multi-tenant data model, append-only audit log, per-tenant AI cost tracking. Brief to brand-correct PDF in under a minute.

Stack

• Next.js — React framework
• Prisma — Type-safe ORM for Postgres
• Postgres — Relational database
• Auth.js — Authentication for web apps
• Tailwind CSS — Utility-first CSS
• shadcn/ui — Composable React UI components
• Puppeteer — Headless Chromium automation
• Claude API — Modelos da Anthropic via API (Opus, Sonnet, Haiku)
• Playwright — End-to-end test automation
• Docker — Containerized deployment
• GitHub Actions — CI/CD pipelines

Measured impact