AI Workflow Automation Specialist for Curated B2B Platform (Claude/GPT + Make or n8n + Airtable/Noti

I'm the founder of an early-stage B2B platform serving a specialized industry vertical. The business runs on a curated, human-approved workflow, and I'm building the internal automation layer that lets a solo operation scale. I need an experienced automation builder to implement a defined set of AI-assisted workflows. The architecture and specifications are already written. This is execution work, not discovery work. Read this part before applying. The automation is Phase 1. Later, these workflows and their data will connect to a customer-facing web platform that my users log into. You are not building that website in this job. But the automation you build has to be designed to connect to it cleanly. If you build a closed system that only runs inside one tool, it has to be ripped out and rebuilt later, which is exactly what I am avoiding. Build it modular, with clean interfaces and data I own. What you'll build (five sequenced deliverables): A content review workflow that checks outbound copy and documents against a defined rule set before anything is sent (rule logic is provided; you build the pipeline and flagging output) An intake pipeline that takes structured form submissions plus uploaded documents (decks, financials) and standardizes them into a defined schema A document generation workflow that produces consistent, professionally formatted briefs from the structured intake data A second document workflow forked from the same data that produces a client-facing roadmap deliverable Three templated, merge-field email sequences triggered from workflow stages, with human approval required before send Estimated total effort: 30 to 45 hours. I'll structure this as five fixed-price milestones. Stack (flexible where noted): LLM layer: Claude API preferred, GPT acceptable Orchestration: Make, n8n, or Zapier (your recommendation, justified) Data layer: Airtable or Notion Forms: Tally, Fillout, or similar Documents: professionally formatted Word/PDF output (I use docx and WeasyPrint pipelines; if you have a better approach, propose it) Billing/email: Stripe and standard founder-grade email tooling exist but are not in scope for Phase 1 Customer-facing website: out of scope for Phase 1. The automation must connect to one later, but you are not building the web app in this engagement Hard requirements: You've built at least two LLM-integrated business workflows in production, not demos. You'll be asked to walk me through one. Strong prompt engineering for structured output (JSON schemas, consistent formatting, low-variance results) Every AI output in this system routes to human review before it touches a customer. If your instinct is full automation, this isn't the right fit. Integration-ready build. The workflows must be able to be triggered by, and return data to, an external web application later, through clean interfaces (APIs, webhooks, or equivalent). No logic locked inside one tool with no way in or out. Portable data. Submission records and all structured outputs live in a data store I own and control (Airtable or a database), not trapped inside the orchestration tool. A future website reads from that store. Clear written communication. I review async and approve milestones in writing. Nice to have: Experience in regulated or compliance-sensitive industries (financial services, legal, healthcare) Document automation experience (docx generation, PDF pipelines) Experience connecting automation workflows to web front ends or member portals Process: Shortlisted candidates sign an NDA and receive the full build brief before quoting final milestone prices. To apply, answer these three questions. Applications that don't answer them will not be reviewed: 1) Describe one LLM-powered workflow you built that a business uses today. What did it replace and what broke the first time? 2) Which orchestration tool would you default to for a five-workflow system owned by a non-technical founder, and why? 3) What's your approach to keeping LLM output format-consistent across hundreds of runs? Budget: $2,500 to $4,500 fixed across five milestones, depending on experience and final scope.