2026 · Client project with NCAD Innovation & Engagement
Deece
Making an AI briefing tool clearer, more consistent and easier to trust.
Overview
Deece is an Enterprise Ireland-backed AI tool that helps marketing teams write clearer briefs and stay aligned with their agencies. The challenge was presenting complex AI outputs, scores, risk flags and multi-step workflows without overwhelming busy, non-technical marketers. Through NCAD’s Innovation & Engagement department, I was one of four designers working as Deece’s design research partner, from a full UX audit to Figma prototypes of two new features.
Client
Deece
Role
UX Design Researcher
Team
4
Timeline
4 Months
UX audit
Finding where people lost confidence.
Two people, three journeys
We started with interviews and task-based usability testing, then built two personas: John, a marketing director juggling too many tools, and Katie, a junior marketer who leans on Deece to write briefs she can stand behind. We mapped their journeys through onboarding, creating a brief and sending it to an agency.
Personas for John and Katie
What we found
Brand details entered during onboarding didn’t save, so people had to upload everything again. Logos stretched on upload. The guided brief flow gave no sense of progress, key actions like Evaluate were icon-only, and AI feedback sat in a different panel from the text it referred to. Every export also needed around ten minutes of manual reformatting.
Onboarding journey map, with pain points and opportunities. Click to enlarge.
Heuristic evaluation
Scoring the platform against usability heuristics gave us five priorities: show what the AI is doing, give people clear exits and undo, use real marketing language, favour recognition over recall, and prevent errors before they happen.
Competitors
Learning from tools people already trust.
Grammarly, Shopify and Tracksuit
Grammarly ties every AI suggestion to the exact text it’s about, with clear Accept and Dismiss actions. Shopify guides new users towards the next step. Tracksuit keeps dense dashboards calm and consistent. From these we drew six principles for Deece, including linking AI output to its source, labelling important actions and giving every colour a job.
Comparing Deece with Grammarly
Design system
Clarity, consistency, confidence, connection.
Design System v2.0
We rebuilt Deece’s visual foundations so colour signals something specific instead of decorating. The brand gradient is now saved for key moments, chips read clearly as either labels or controls, and every button and field has default, hover, pressed, disabled and validation states. Status messages always pair what happened with a clear next step.
Button states from Design System 2.0. Click to enlarge.
Alerts and status messages, each with a clear next step.
Accessibility by default
We reviewed the platform against WCAG 2.1 AA and the EU Accessibility Act, then built the fixes into the component library: labels every screen reader can announce, visible focus states for keyboard users, a proper heading structure, and contrast that holds up in every state. New features inherit all of it from the start.
Campaign Navigator
Scoring agency work against the brief.
The problem
When an agency presents back, marketers have to judge the work against the original brief from memory, round after round. Campaign Navigator lets them upload each agency presentation, link it to the brief, and get a scored review across brief alignment, legal and risk, with every flag ranked by severity.
Campaign Navigator user flow. Click to enlarge.
Designing for trust
Taking cues from Adobe GenStudio and Hawky, every score shows its reasoning: what the brief asked for, what the agency proposed, and where the proof is. Marketers see the headline score and critical flags first, open a flag to investigate the evidence, then dismiss it or flag it to the agency. Each new round carries forward what’s been resolved and what’s still outstanding.
A round under review, with flags by severity
Investigating a flag
Agencies in the loop
Agencies only see the briefs they’re invited into. They can comment and add context, and any change they suggest waits for the marketer’s approval. We deliberately kept Campaign Navigator from judging creative work or ranking people, so the marketer always stays in charge of the feedback.
The agency’s side of a round
Effectiveness Analyser
Closing the loop after launch.
From brief to results
Once a campaign is finished, it moves into Effectiveness Analyser. It compares results line by line against the brief’s success criteria, flags KPIs the report never addressed, and turns findings into learnings saved for the next brief. A chat drawer lets marketers ask questions about any campaign.
Effectiveness Analyser wireframes. Click to enlarge.
Results
A full UX audit, journey maps, competitor analysis, accessibility review, Design System v2.0, and Figma prototypes of Campaign Navigator and Effectiveness Analyser, handed over to Deece’s developers.
Retrospective
What I took away.
AI has to show its working.
Almost every fix came back to trust. People won’t act on an AI’s output until they can see what it did and why.
Consistency is a feature.
Small gaps, like a button with no pressed state, added up to a product people second-guessed. Fixing the system fixed a lot at once.
Design for both sides.
Treating agencies as real users, not just recipients of a PDF, changed how we designed sharing, comments and approval.
Next project