Merchant AI
AI website chatbot SaaS with Shopify, WooCommerce and WhatsApp integrations
An AI assistant trained on a business's own website, files and FAQs, embedded as a chat widget.
- Year
- 2026
- Client
- Own product
- Live
- merchantai.io
The live product site. The assistant answers only from approved sources (pages, files and Q&A), can capture a lead mid-conversation, and hands over to a human by email or in the dashboard.
Overview
Merchant AI lets a business train an AI assistant on its own content, then put it on its website, Shopify store or WhatsApp. It answers only from approved sources, captures leads mid-conversation and hands over to a person when it should.
Research & service design
Designed around real support and knowledge use cases rather than AI novelty: what customers ask, where the answers already live, and when a person needs to step in.
Made setup a guided journey: add website pages, files and Q&A, review what the assistant has learned, then publish.
Kept trust visible: the assistant answers only from approved sources, and every conversation can be reviewed, escalated or handed over.
What I built
Built retrieval with OpenAI and Pinecone over each business's own content, gathered by a Playwright site crawler.
Wrote the backend in Encore.ts on Postgres, with training versions that can be rolled back.
Added lead capture mid-conversation, human handover by email or in the dashboard, and topic and sentiment analytics.
Scoped workspaces with roles and plan limits, with Stripe for plans.
Shipped a Shopify app (OAuth, webhooks and a theme extension), a WooCommerce plugin and WhatsApp through Meta's Cloud API.
Inside the work
Outcomes
Live at merchantai.io, with a Shopify app, a WooCommerce plugin and WhatsApp integration.
GitHub Actions publish the changelog and monitor uptime and TLS.
Runs as a distinct product with its own brand, blog and roadmap.
AI product strategy, retrieval and guardrails, SaaS architecture, platform integrations and commercially grounded product execution.