Redacted

2025

Driving an 86% Jump in Dealer Satisfaction

TL;DR - After I designed a complete 0 → 1 dashboard, dealers had data but no fast way to turn it into decisions. I led the design of a context-aware AI that lets users speak or type a prompt and instantly get analysis, charts, and PDF reports.

Feature design
AI Integration
SaaS
Scalable design system
Automotive

The Context

My Role

Senior Product Designer

Team

Founder, CTO, 3rd party dev team, QA

Space

Redacted, SaaS, Automotive, AI

What I walked into

Dealers couldn’t pull historical data fast

One-off feature requests were eating up time and attention

The DMS was central to daily work, but individual insights where manual tasks

Space on screen was already tight, new features had to fit without being in the way

The Problem

Dealers needed a way to create reliable insights and reports in seconds, without manual searching, analysts or extra tools.

Aging stock ties up capital; late insights mean missed pricing moves; manual reporting steals hours from sales. If we can compress “question → actionable answer” to seconds, we improve margin, turnover, and confidence in our users.

The Discovery

My starting point

Interviewed 6 dealerships (owners, sales managers, finance admins, technicians)

Mapped current reporting flows and time-to-report

Ran CSAT surveys to gauge where we currently sit with reporting

Collected screenshots of workarounds

Tested comfort levels with AI (most already used ChatGPT casually)

What I heard

"I want to see what’s sitting too long, not mess with filters."

“Every time the owner asks for a report, it’s a half-day in Excel.”

“How can I trust AI if I can’t see where the numbers came from.”

“How can I trust AI if I can’t see where the numbers came from.”

“We make decisions on instinct because it’s faster than pulling the data.”

“We make decisions on instinct because it’s faster than pulling the data.”

“I want to walk into a meeting with a chart ready, not raw data.”

“I don’t care about another dashboard unless it actually saves me time."

What does all this mean?

Dealers don’t want “search,” they want answers

Every output needs to link back to the source data

The key value isn’t new dashboards, it’s speed-to-report

Adoption depends on making it feel like part of their workflow, not “new tech”

My guardrails

Every output links back to its table/view

Entry point must be intuitive, simple and not frustrating

AI auto-detects context (Inventory, Finance, etc.)

Design exploration

Modal vs. Drawer

Modal won: lets dealers see data while prompting and navigate freely. Drawer took too much space and broke navigation.

Clipping issue

Original search design inputs clipped long prompts. Replaced with expandable stacked input for clarity, it also used layer blur, but in practice caused all the text to distort so switched to a box shadow.

Sidebar placement

Broke mental model, sider bar = navigation not action.

Top Bar placement

Stole space from core CTAs, repeats across every screen.

Floating input

Always there, minimal friction, fits “ask anywhere” idea, but can cause issue as it blocks lower screen data.

Final Design

Floating Input Button

  • Bottom-centre button styled like a text field

  • Glass background so it doesn’t block dashboards

  • Expands into full overlay on click

Text + Voice

  • Dealers type or record voice

  • Voice is transcribed to text before sending

  • Preview, confirm, or discard

Contextual

  • Links to the current dashboard/dataset automatically

  • Context stays active even when switching screens

  • Dealers can switch context to different datasets

Conversation

  • Full chat history

  • Start new or pick up recent conversations

  • Minimise panel to keep working while AI processes

  • Little space as dealers required quick, short answers in text form, so graphs, charts and deep insights are sent as a downloadable link. (also saves on generation tokens)

More Menu

  • Mark chats as private or team-shared

  • Access complete history

  • Switch between models (own API key or managed plan)

  • Settings is accessed in account area

The Impact

Dealer satisfaction nearly doubled, from 38% to 71% (86% increase CSAT Scores)

Reporting became part of the meeting, not prep work before it

Teams started basing decisions on data instead of instinct

Dealers told us the DMS “felt complete” with insights built in

Learnings

Dealers didn’t care about “AI features.” They cared about speed, clarity, and trust. By placing this inside their daily flow and keeping outputs transparent, we built something dealers used every day, and something they could actually rely on.