Skip to content
Client
Andrej Drats, Mizu AIAI automation builder, 0 to launch
Year
2025
Involvement
Product Design, Frontend, Backend, Full-Stack Build

Andrej came to us with a vision and none of the system behind it: no product, no frontend, no backend, no integration layer. We built Mizu AI's automation platform from the ground up — a Lindy-style builder where users describe a workflow in plain English and get a working automation, backed by real integrations and real billing underneath. Auth, onboarding, a conversational automation builder, a visual canvas, five integrations wired through Composio SDK, and a full subscription system — designed and engineered together, end to end, in six weeks.

Mizu AI: Shipping an AI-Native Automation Builder in 6 Weeks

Mizu AI: Shipping an AI-Native Automation Builder in 6 Weeks

The challenge

Andrej needed a product that could compete with category leaders like Lindy — where the bar isn't 'does automation work,' it's 'does it feel effortless.' Non-technical users had to go from a plain-English prompt to a deployed, integrated automation without ever seeing the complexity underneath: auth, credential management, workflow state, billing tiers, credit metering. All of it had to exist, and none of it could be visible. And it had to ship fast — six weeks, full stack, with nothing to build on top of.

Our approach

We didn't build screens in isolation — we designed the system first: how a user's intent turns into a workflow, how that workflow becomes an editable canvas, how integrations attach without friction, how credits and plan tiers gate usage without feeling punitive. Frontend and backend were built in lockstep so the product could ship as one coherent experience, not a UI wrapped around a backend built after the fact. Three things drove every decision: speed without shortcuts — a full auth flow, onboarding, dashboard, automation builder, template system, integrations, and billing, shipped in six weeks, not a stripped-down MVP; AI-native UX — the product leads with a conversational builder, not a form, where users describe intent and the AI asks clarifying questions before building the flow, with a visual canvas underneath for anyone who wants to go deeper; and integration depth via Composio — Gmail, Slack, Google Calendar, Notion, and Google Docs wired through Composio SDK, so credential handling and step configuration stay consistent across every integration rather than each one being a one-off.