Directam
- Role
- Design Technologist
- Timeline
- Aug 2025 – May 2026
- Platform
- SaaS Social Media Management Platform
Overview
One language — built into the product.
Directam is a SaaS social media management platform. The marketing site and the dashboard needed to feel like the same product. I defined that shared design language, built the Vue-ready system behind it, and extended the work into DesignOps, ResearchOps, the sales funnel, and AI features I designed and implemented — including consultation booking with Cursor.
The lasting output is not a deck. It is a system teams can reuse: tokens, components, handoff rules, and AI flows where the agent proposes while a person remains the decision-maker.
The Challenge
Two surfaces. Two vocabularies. One product.
Brand and product were solving the same interface problems with different answers. Each new feature reopened decisions that should already have been settled — density, hierarchy, state, handoff. Engineering absorbed the cost as review churn; users absorbed it as inconsistency.
Research existed, but not as a shared operating layer. Without one place to ask a question, prioritisation stayed opinion-led instead of evidence-led.
Constraints
- Marketing site and SaaS dashboard shipping on different cadences.
- Vue.js front end that needed tokens and components, not Figma-only specs.
- Research tools present but not wired into one workflow.
- AI features that had to stay human-in-the-loop — assist, never decide alone.
Research & Discovery
Make evidence cheap to reach.
I stood up a ResearchOps stack on Dovetail, Mixpanel, Clarity and CRM so behavioural data, qualitative synthesis and commercial context lived in one workflow. On that base I ran JTBD across three user groups and defined a segmentation model for product prioritisation.
Methods
- ResearchOps stack: Dovetail, Mixpanel, Clarity, CRM — one place to ask.
- JTBD interviews across three user groups.
- Segmentation model for product prioritisation.
- Audit of brand vs product language for density, hierarchy and state.
A design system without ResearchOps is still opinion with better components.
What the evidence changed
- Brand and product did not need two systems — they needed one language with two densities.
- Review churn was a handoff problem, not a taste problem.
- Funnel friction showed up in Mixpanel before anyone named it in a critique.
Process
Language, system, ops, funnel, AI.
-
01
Shared brand & product language
One visual and interaction language for the marketing site and SaaS dashboard, aligned with marketing so neither surface drifted alone.
Shared language across brand and product. -
02
Design system for Vue.js
Architected, built and maintained 45 components with tokens and variables for the Vue.js front end — reusable, not redrawn per screen.
45-component Vue-ready system. -
03
DesignOps & handoff
Standardised DesignOps and the delivery process to engineering so review cycles stopped reopening settled decisions.
Handoff model and review cadence. -
04
Sales funnel redesign
Redesigned the sales funnel to improve yearly subscription conversion — evidence first, then the path.
Funnel rebuilt for yearly conversion. -
05
AI Comment
Designed around human-in-the-loop Agent Experience — the agent assists; a person confirms.
AI Comment — assist, then confirm. -
06
AI consultation booking
Designed and implemented the booking flow with Cursor — from interaction model to something that runs.
Consultation booking built with Cursor.
The Solution
A system teams can reuse — and AI that stays under human control.
Vue-ready system
Forty-five components with tokens and variables — engineering builds from the same source design decides from.
Evidence in the loop
ResearchOps makes behavioural and qualitative evidence cheap to reach when priorities are argued.
Human-in-the-loop AI
Agents propose; people confirm. Comment and consultation booking stay under human control.
Outcomes
What moved.
- 45-component design system with tokens and variables for Vue.js.
- ~40% shorter review cycles after DesignOps standardisation.
- More than 25% growth in yearly subscription conversion after the funnel redesign.
- ResearchOps stack on Dovetail, Mixpanel, Clarity and CRM.
- JTBD + segmentation across three user groups for prioritisation.
- AI Comment and consultation booking designed and built human-in-the-loop.
What held.
- One language across brand and product removed more rework than any single screen redesign. Settled decisions stay settled.
- A Vue-ready system changed the handoff. Specs became something engineering could consume, not interpret.
- AI that proposes without deciding kept trust intact. Human confirmation is not a nice-to-have when the product sells subscriptions.
