How I shaped the MVP feature set, designed cross-feature connectivity, and built a Tailwind-aligned design system for an AI marketing platform entering a crowded market.
This case study is shared privately. Message me for the password, or enter it below if you already have one.
Incorrect password. Try again.
By mid-2025, the AI content tool market had become genuinely crowded. Every category — writing assistants, social media schedulers, brand kits — had multiple well-funded competitors. The harder problem wasn’t building an AI writing tool; it was giving a startup a reason to exist in a space where everything already existed.
SAVRA’s founders saw a real gap: small businesses and startup teams were stitching together multiple separate tools to manage their marketing — and losing their brand voice in the process. Every platform required re-explaining who they were. No single tool knew the brand deeply enough to produce content that actually sounded like them.
The opportunity was coherence: brand knowledge stored once, powering everything across the product.
I joined as the sole designer at zero: no existing design language, no defined feature scope, no user research backlog. I owned the full design surface — discovery, information architecture, key feature design, wireframes, high-fidelity UI, and the design system. I worked directly with the founder and a marketing strategist throughout.
The primary constraint was scope ambiguity. The team had a strong product vision and many features in discussion simultaneously, but no clear MVP definition — several concepts were running in parallel without a coherent structure connecting them. Before I could design anything meaningful, I needed to help the product find its shape.
Before any screen, the product needed a scope decision. I catalogued all the features in discussion, benchmarked leading AI content tools to understand what behavioral patterns users were developing with AI, and worked with the founder and marketing strategist to identify what SAVRA could do better — not just differently.
The target user became clear: startup operators and small business owners who create content themselves, don’t have dedicated marketing teams, and need tools that understand their brand without requiring them to explain it from scratch every time.
I proposed narrowing the first release to three features that had something competitors didn’t: they were designed to depend on each other.
Company Knowledge Hub → AI Writer → Social Media Post Generator
A business builds their brand voice and knowledge base once. The AI Writer draws on that foundation to generate on-brand long-form content. The Social Media Generator repurposes and distributes that content across channels. One source of truth — three surfaces, no context switching, no re-explaining who you are.
The most deliberate design principle I introduced was cross-feature connectivity, designed around the specific moments in a task where a user’s natural next step leads to another feature.
Finishing a draft in the AI Writer surfaces a direct path to the Social Media Generator. Updating brand information from within a content creation flow connects back to the Knowledge Hub without losing context. The features share the same knowledge layer, so the brand voice established in one place applies everywhere, automatically.
My core concern was user action flow — making sure users never hit a blocker mid-task. In a tool aimed at solo operators and small teams, friction doesn’t just slow people down; it breaks the habit of coming back. I mapped every junction between features and designed the handoffs to feel natural rather than forced.
The engineering team was building on Tailwind CSS. Rather than designing in isolation and handing off a spec for interpretation, I built the design system in direct correspondence with Tailwind’s token architecture.
I defined the color palette, typography scale, spacing system, and branding direction — then created a Figma reference table that mapped each design decision to its Tailwind equivalent. Developers could look at any component and immediately find the corresponding class without translating between two separate systems.
I also established UI layout patterns and component behaviors that applied uniformly across all three features — so that even as different parts of the product were built at different times, the experience felt like a single coherent product.
SAVRA launched as an AI marketing operating system positioned around brand voice consistency — view the live product → — “Your entire marketing team. One AI that sounds like you.” — which maps directly to the product architecture designed here: brand knowledge established once, distributed everywhere.
Early users have reported measurable results: conversion up 18%, trial signups up 22%, reply rates doubled in one quarter. The signal users cite most consistently is brand consistency — “Every post, email, and landing page sounds like us.” — which is exactly the problem the product was designed to solve.
The design system held at launch. The cross-feature connectivity model worked as intended: users move between features at natural task transitions rather than treating them as separate tools.
Designing at zero in a crowded market taught me that the most important design work often isn’t a screen — it’s a structural decision. The IA audit that narrowed many features into three, and the choice to make those three function as one connected system, shaped everything that followed. If I were doing it again, I’d push for direct user research earlier in the process; we had strong market intuition, but talking directly to small business operators sooner would have pressure-tested the feature bets faster. That said, the core insight, that coherence is the differentiator in a fragmented tool market, proved out in both the product architecture and the user response at launch.