TWO Design is a boutique branding studio. Eight people, no account managers, strong opinions about how work should flow. We tried Asana. We tried Monday.com. We tried Notion. None of them felt quite right.
So I built our own task management system. Named it Twotion because naming things is hard and puns are easy.
The Context
Creative agencies have specific workflows. Tasks aren't generic items; they're copywriting, graphic design, motion graphics, product design. Each task type requires different skills. Different team members specialise in different things.
The existing tools treated everything the same. A branding task looked identical to a motion task. No context about who's good at what. No understanding that our freelancers work differently from core team members.
Also, and this is petty, none of them matched our brand colours. Purple accent, teal-green for success states. It's a design agency. These things matter.
The Features
Kanban Board: Drag-and-drop columns. Backlog, In Progress, In Review, Done. Nothing revolutionary. The basics need to work before you add clever features.
Client-Project Hierarchy: Tasks belong to projects, projects belong to clients. Dynamic filtering lets you see everything, or just what's relevant. Common pattern, but surprisingly rare in tools optimised for tech teams rather than agencies.
Workload Visualisation: Real-time indicators showing who's overloaded and who has capacity. Colour-coded: light (green), medium (yellow), heavy (red). Prevents the classic agency problem of three people drowning while two people are idle.
Team Member Management: Role badges distinguish leadership, core team, and freelance consultants. Freelancers get visual distinction (dashed borders, "Ext" badges, separate sections) because their availability and commitment level differ.
Multiple Assignees: Junction table architecture for tasks that need more than one person. Design plus copywriting collaboration. Review chains. The real world is messy.
The AI Bits
Smart Task Assignment: When a new task comes in, the AI analyses the content and suggests who should handle it. Understands that "logo design" goes to different people than "video editing." Learns from our team's actual specialisations.
Task Type Categorisation: Automatically tags tasks as Copywriting, Graphic Design, Product Design, Motion, etc. Saves the tedious manual categorisation that nobody actually does consistently.
Weekly Summaries: Monday mornings, the AI generates a week-ahead planning summary. What's coming up, who's working on what, potential bottlenecks. Friday afternoons, a week review: what got done, what didn't, why.
Team member names get highlighted automatically in summaries. Small touch, but makes scanning easier.
Uses BYOK (bring your own key) with Anthropic's Claude. The agency pays directly for AI usage with full cost visibility.
The Stack
Runtime: Bun instead of Node.js. Faster, native SQLite support, less overhead. No Express or Fastify. Manual routing because the app is simple enough not to need framework bloat.
Database: SQLite. Lightweight, portable, no server to maintain. Junction tables for many-to-many relationships. Good enough for an eight-person agency.
Frontend: Standard web stack. Dark/light mode with proper brand colours. Touch-friendly for the team members who prefer tablets.
The whole thing is deliberately simple. No microservices. No containerisation. Just code that runs and does its job.
Honest Reflection
Is building custom task management software for an eight-person studio overkill? Objectively, yes. We could make Asana work with enough discipline and customisation.
But Twotion fits how we actually work. The AI suggestions save time on task routing. The weekly summaries catch things that would otherwise fall through cracks. The workload visualisation prevents burnout before it happens.
More importantly, the team actually uses it. That's the bar custom software needs to clear. If people avoid using it and default to Slack messages and spreadsheets, you've failed. Twotion is where our work lives.
The maintenance burden is real. When something breaks, I'm the one who fixes it. When someone wants a new feature, I'm the one who builds it. That's the trade-off for software that fits perfectly.
For TWO Design, it's worth it. For most teams, probably not. Know your context before building custom tools.