Requset
AI-driven request and approval workflow platform — built solo, launching soon.
Context
Every team I’ve worked at has the same broken pattern: requests live in seven places. PTO in HR, hardware in IT, budget in Slack DMs, expense in a forwarded email chain. When a manager has to approve fifteen things a week, the friction is the loss — not the decision itself.
Requset is the answer to “what if all of those went through one inbox, with AI shaping the workflow per request?” I’m building it solo as a SaaS, owning everything from architecture to GTM.
What I built
- Multi-tenant SaaS on Next.js + Supabase with Cloudflare R2 for blob storage and a Python backend on Azure for AI workloads.
- AI workflow generation via GPT-4o and GPT-5 Nano on Azure AI Foundry — teams describe their approval flow in plain English (“PTO over 5 days needs grandboss; otherwise direct manager”) and Requset generates the form, routing logic, and reminders.
- Three-tier seat-based pricing with Polar including per-seat storage scaling, AI action limits, and overage billing logic.
- Zero-infra-cost architecture through deliberate use of free tiers and Azure credits — runway-friendly for a solo founder pre-revenue.
How it works
Workflow generation runs server-side: the user prompt becomes a structured JSON form definition (validated against a Zod schema), then renders client-side. The same AI loop handles routing inference: the LLM proposes routing logic, a deterministic validator checks for cycles and orphan branches, and the user gets a previewable diagram before activating.
Per-seat pricing math runs through Polar’s metering API, with overage logic written in Python because the billing reconciliation lives next to the AI usage tracking.
Outcome
Pre-launch. Building toward public launch in 2026.