Adnova

Adnova is an AI-driven marketing dashboard for performance teams. The brief was straightforward to write and hard to design: turn fragmented campaign data into decisions, without burying the user in charts or hiding the data behind a chat box.
Year 2025
Role Product Designer
Focus AI UX, dashboards, design system
Website hamedbahadori.com

The real question

Marketing dashboards usually fail in one of two ways. They either dump every metric on a single screen and force the user to make sense of it, or they wrap everything in an AI chat and hide the data behind a prompt. Neither works for a marketer who needs to ship an answer before the next standup.

How do we put AI inside the decisions without taking the data away from the user?

Five decisions that shaped the dashboard

01 · Hierarchy

Insights above charts

AI alerts and trend summaries land first; raw data sits one click below — visible, not hidden.

02 · Trust

Traceable AI suggestions

Every AI tip cites the campaign, metric, and timeframe it came from. No black-box advice.

03 · Input

Prompt Builder, not a prompt box

Source, metric, and timeframe pickers compose the prompt — no blank-page tax on the user.

04 · Context

Embedded AI Advisor

Chat lives next to the campaign it discusses, not in a separate tab.

05 · System

Variant-driven design system

Atomic components, light + dark themes, AA contrast. Same primitives across every module.

★ Outcome

Decision-led, not chart-led

The dashboard answers "what should I do next?" before it answers "what happened?"

Trade-offs I made on purpose

Tempting Single big chart dashboard Looks comprehensive in a screenshot, makes the user do the synthesis.
Chose Insight-first surface with data one tap below The user gets the answer; the data backs it up when they want to verify.
Tempting Open-ended AI chat as the main UI Demos well, fails when the user doesn't know what to ask.
Chose Structured Prompt Builder + contextual chat Pick source · metric · timeframe → get a real analysis, not a guess.

What the dashboard covers

Page

Dashboard

KPI overview, AI alerts, user behavior graphs.

Page

Campaigns

Budget, retention, ROAS, contextual smart tips and warnings.

Page

User Quality

Segment-level insights and retention performance.

Page

Prompt Builder

Compose GPT analyses from source / metric / timeframe pickers.

Page

AI Advisor

Chat-based strategic suggestions, embedded next to the campaign.

Page

Settings & Profile

Account, integrations, theme, accessibility.

Design system

Inter UI 8px grid Light + Dark WCAG AA Figma variables Variant-driven

Reusable cards, data tables, toggle tabs, alert banners, modals. Light and dark themes are powered by Figma variables, so the team can swap themes without rebuilding components. Microinteractions — hover states, collapsible charts, tooltips, loading feedback — exist where they aid comprehension, not as decoration.

Impact

Workflow

Lower cognitive load

Marketers see the answer first, the data second.

Trust

AI with receipts

Suggestions cite their source — confidence goes up.

System

Reusable SaaS kit

The component set ports cleanly to fintech, analytics, ops dashboards.

Learnings

  • 01Variant-based systems future-proof multi-theme products. Theme is a property, not a redesign.
  • 02AI + UX only works when the model shows its work. Context and traceability earn the trust.
  • 03Figma discipline pays off at handoff. Clear naming and tokenization save the next person's week.