Shopify AI Automation Developer: Core Stack, Custom Dashboard and Pilot

We are building a long-term branded Shopify ecommerce business, initially for Australia and the USA, with additional countries later. We need an experienced Shopify + AI automation developer to implement a client-led architecture and provide focused technical assistance. The client side will prepare the architecture, workflow designs, acceptance criteria and backlog. Please review the proposed design, identify gaps and quote implementation; do not assume a separate large discovery engagement is required. We are pre-launch: product shortlist, final store size and order volume are not yet established. We want a practical, maintainable pilot and measurable delivery, followed by optional ongoing support. We are seeking proposals only, not authorising paid work or subscriptions through this posting. PROPOSE THE BEST CORE STACK Compare the following options and recommend one coherent stack, with reasons, limitations and total operating cost: - Shopify CLI, Admin GraphQL API and Shopify Dev MCP for development where appropriate. Explain each tool's purpose and distinguish development tooling from production runtime. - n8n self-hosted versus n8n Cloud, Make and Zapier. Self-hosting is a candidate, not an assumption that it is cheapest: include server, backups, monitoring, updates and maintenance labour. - Claude API for agents and Claude Code for app/theme development; evaluate suitable alternatives where materially better value. Confirm currently supported model IDs and Shopify API versions. Treat any starter snippets as illustrative only, not production-ready. - Reuse Shopify Sidekick, Shopify Magic, Flow and existing apps where they meet requirements instead of recreating them. - Propose a small branded dashboard and shared database for products, supplier evidence, quotations, tasks, approvals and operational reporting. CLIENT-LED ARCHITECTURE BASELINE Use a shared task queue and database as the source of truth. Each task needs an owner, status, deadline, evidence, audit trail and cost allowance. Use event-driven workflows, bounded retries, duplicate-action prevention and a manual recovery queue. Keep fulfilment and money calculations deterministic; use AI for interpretation, drafting and reasoned recommendations. Prefer one Shopify store with AU/US market configurations initially, subject to verified payment, shipping and product requirements. MODEL ROUTING AND APPROVALS Use an economical model for bulk product descriptions, classification and routine summaries; escalate difficult support, pricing and other reasoning-heavy cases to a stronger model. Provide a model/task/cost comparison and evaluation results rather than selecting the most expensive model for everything. Routine support can use templates or smaller models when sufficient. Require human approval before publishing/editing live product data, changing prices, shifting ad spend, placing purchases/reorders, issuing refunds or other material customer actions. Approval must cover the exact proposed change; reject stale approvals when the underlying data changes. Agents must not invent product features, delivery promises or policy statements. PHASE 1: SMALL WORKING PILOT Please separately price the smallest useful pilot: 1. Branded dashboard with product and supplier records, evidence links, task queue, approval inbox and activity/cost log. Start with user-provided product links/images and permitted data sources; do not promise to monitor every Instagram Reel or bypass platform restrictions. 2. Product listing/SEO assistant: draft descriptions, titles, metadata and alt text only from verified product facts; preview and approval before Shopify publication. 3. Customer support assistant: initially internal drafts, grounded in approved shipping/returns policies and authorised order lookups; escalate uncertain cases to a human. 4. Daily/weekly operational report: distinguish real data from estimates and explicitly report missing data, cost and exceptions. Email summary initially; Slack is optional. 5. Test Shopify integration and one supplier integration path (standard app first; custom API only if justified). Never create duplicate orders or dispatch without approved rules. LONG-TERM ROADMAP: TEN CAPABILITIES, NOT TEN REQUIRED SUBSCRIPTIONS 1. Customer support: compare Gorgias AI, Tidio Lyro and custom Claude support. 2. Product listings and SEO. 3. Inventory forecasting and draft reorder/purchase orders. 4. Order exceptions: fraud signals, bad addresses, delayed shipments. 5. Email/SMS marketing: compare native options and Klaviyo; respect consent requirements. 6. Ad optimisation: recommendations first, then approved changes within agreed caps. 7. Pricing and competitor monitoring: permitted sources, margin rules, suggestions before changes. 8. Review management: compare Judge.me/Loox, genuine review requests and draft replies. 9. Analytics and management reporting. 10. Dev/QA assistance using Claude Code and Shopify Dev MCP, with review and testing before deployment. Also retain product discovery and supplier verification as core business workflows alongside this roadmap. DELIVERY REQUIREMENTS All accounts, code repositories, credentials and infrastructure remain client-owned. Use least-privilege access, secure secrets, minimal customer data, staging/testing, backups and a documented restore process. Include failure alerts, handover documentation/video, workflow exports, tests and a clear maintenance plan. Do not use live customer data in development unless specifically authorised. PLEASE ANSWER IN YOUR PROPOSAL 1. Your recommended stack and why; buy-versus-build decisions for each pilot capability. 2. Two comparable Shopify + AI/automation projects and what you personally delivered; a demo or redacted walkthrough is preferred. 3. Who will actually implement the work, their availability and communication/reporting schedule. Disclose agency/subcontractor involvement. 4. Fixed-price milestones, hours assumptions, timeline and dependencies. Separate pilot, later phases, optional support and any architecture review fee. State currency and all exclusions. 5. Expected monthly subscriptions, API usage, hosting and maintenance at illustrative 100 / 500 / 1,000 monthly orders. These are comparison scenarios, not promised volumes. 6. How you will test approvals, model quality, retries, duplicate webhooks/orders, supplier/API outages and rollback/recovery. 7. What you can remove or simplify to lower cost without weakening reliability, and what you need from us to start. We will compare proposals and may agree a small paid technical trial before selecting one lead developer. Please do not begin paid work, connect accounts, purchase inventory, launch ads or buy subscriptions without a separately agreed scope and approval. Our goal is a reliable long-term partnership with controlled costs and clear ownership. Kind regards, Martin