x
Reimagining Lead
Generation for Fiat
Reimagining Lead
Generation for Fiat
Reimagining Lead
Generation for Fiat
Designing an AI-powered WhatsApp conversation experience to drive higher-quality leads for Fiat's new models.
Designing an AI-powered WhatsApp conversation experience to drive higher-quality leads for Fiat's new models.
Role
Visual Communication
Designer
Timeline
2 weeks
Team
1-2 people
Platform
Whatsapp — AI
Role
Visual Communication Designer
Timeline
2 weeks
Team
1-2 people
Platform
Whatsapp — AI
Fiat Topolino Features in form of Whatsapp AI conversation
The Impact
Shifting from a static, low-activation chatbot to a personalized AI conversation flow — producing measurably more qualified leads and significantly higher engagement than the Topolino baseline test.
Shifting from a static, low-activation chatbot to a personalized AI conversation flow — producing measurably more qualified leads and significantly higher engagement than the Topolino baseline test.
+63%
moved closer to purchase decision
+390%
more leads vs. traditional bots
Problem
Traditional channels were running out of road
Traditional channels were running out of road
Fiat's lead-gen relied on SMS and static web forms — channels with declining returns. An initial WhatsApp campaign showed strong ad engagement but the bot barely activated. The gap between click and conversation revealed a structural problem: the flow wasn't built for real people.
Fiat's lead-gen relied on SMS and static web forms — channels with declining returns. An initial WhatsApp campaign showed strong ad engagement but the bot barely activated. The gap between click and conversation revealed a structural problem: the flow wasn't built for real people.
"271,000 link clicks. Under 3,000 bot activations. The ad wasn't the problem."
01
Diminishing returns on traditional channels
SMS and form-based capture were yielding fewer, lower-quality leads each cycle.
02
Low bot activation despite high ad engagement
The Fiat Topolino test: 271K link clicks — under 3K bot activations. A 99%+ drop-off at step one.
03
Static, one-size-fits-all conversation flows
Menu-driven bots couldn't adapt to different intents — browsing, comparing, or ready to buy.
Fiat Topolino — WhatsApp Test Funnel
Link clicks
271,000
WhatsApp opens
~18,000
Bot activated
2,970
Qualified leads
~356
99%+ of ad clicks never reached the bot — the static flow was the bottleneck.
Fiat Topolino WhatsApp campaign data. Used for diagnostic framing.
Research
Users haven't moved to messaging — they already live there
Users haven't moved to messaging — they already live there
Shifts toward messaging over calls and forms were well-established before this project. Add shrinking attention spans and privacy-driven signal loss weakening traditional ad targeting, and the case for conversational channels becomes structural, not tactical.
Users aren't asking for less information — they want it delivered conversationally and calibrated to where they are in the decision. Browsing intent and ready-to-buy intent require completely different responses.
Shifts toward messaging over calls and forms were well-established before this project. Add shrinking attention spans and privacy-driven signal loss weakening traditional ad targeting, and the case for conversational channels becomes structural, not tactical.
Users aren't asking for less information — they want it delivered conversationally and calibrated to where they are in the decision. Browsing intent and ready-to-buy intent require completely different responses.
85%
of users prefer messaging over calls or email for brand interactions
Meta / McKinsey messaging preference data, 2023.
↓ Signal
Cookie deprecation weakens traditional ad targeting signals
3×
higher open rate — WhatsApp vs. email campaigns
What we learned
Ad-to-chat is the new ad-to-lead-form. Each conversation becomes a first-party data source — richer and more reliable than an anonymous form submission.
Ideation
From scripted menus to a context-aware agent
From scripted menus to a context-aware agent
Early exploration compared two approaches: a rigid menu-driven bot versus a context-aware AI agent capable of real-time Q&A and personalized follow-up. The menu-driven path couldn't close the activation gap — the same failure mode as the Topolino test.
Early exploration compared two approaches: a rigid menu-driven bot versus a context-aware AI agent capable of real-time Q&A and personalized follow-up. The menu-driven path couldn't close the activation gap — the same failure mode as the Topolino test.
Old approach
Static Chatbot
✕
Pre-written menu trees
✕
Same flow for every user
✕
No open-ended Q&A
✕
Drops unexpected inputs
✕
Requires manual updates
Proposed
AI Agent
✓
Context-aware responses
✓
Adapts to user intent
✓
Real-time Q&A on any model
✓
Handles open-ended inputs
✓
Learns from conversation
Testing interaction flows early
The prospect-scoring approach identifies intent through opening questions — budget, timeline, current ownership — and branches accordingly. Low-propensity users get exploration content; high-propensity users move directly toward dealer booking.
The prospect-scoring approach identifies intent through opening questions — budget, timeline, current ownership — and branches accordingly. Low-propensity users get exploration content; high-propensity users move directly toward dealer booking.
Prospect Intent Scoring
Low
Browsing, no timeline
→ Educate & inspire
Medium
Comparing models
→ Compare & clarify
High
Ready to buy
→ Book test drive
Four-stage design process — each phase had a defined output and a stakeholder sign-off gate.
Designs
One thread.
Every step of the journey.
One thread.
Every step of the journey.
The final experience lives in a single persistent WhatsApp conversation — no app downloads, no form redirects. From the first tap on a Meta ad through model discovery, dealer selection, and test-drive booking, everything happens in the thread.
The final experience lives in a single persistent WhatsApp conversation — no app downloads, no form redirects. From the first tap on a Meta ad through model discovery, dealer selection, and test-drive booking, everything happens in the thread.
End-to-end WhatsApp AI flow — ad click through test-drive booking.
The Personalization Engine
Opening signals — timeline, ownership, budget — feed a propensity model that determines which branch the AI takes. This data also populates the CRM, giving the sales team qualified leads with intent context attached.
Opening signals — timeline, ownership, budget — feed a propensity model that determines which branch the AI takes. This data also populates the CRM, giving the sales team qualified leads with intent context attached.
Opening signals — timeline, ownership, budget — feed a propensity model that determines which branch the AI takes. This data also populates the CRM, giving the sales team qualified leads with intent context attached.
💬
Entry Questions
Budget, timeline, current car — surfaces propensity signals
🔍
Low · Score 1–3
Exploration content, inspiration, soft CTA to configurator
⚖️
Medium · Score 4–6
Model comparison, pricing deep-dive, FAQ resolution
🚗
High · Score 7–10
Dealer locator → test-drive booking → calendar confirmation
📊
CRM Handoff
Lead data + intent score passed to sales team with full conversation log
Prospect scoring logic — determines AI branch and CRM lead quality tier.
Lessons
What this project taught me
What this project taught me
Conversational personalization affects lead quality more than raw volume. The Topolino test generated thousands of clicks; the problem was that almost none converted into meaningful intent signals. A smarter first exchange is worth more than optimizing ad spend.
The design challenge wasn't building a chatbot — it was building a conversation that felt natural enough to keep someone in the thread past the first message. At that layer, interaction design is closer to writing than UI.
What I'd validate further: running this flow head-to-head against the original Topolino benchmark, and testing whether WhatsApp-booked test drives show at a higher rate than form-generated leads.
Conversational personalization affects lead quality more than raw volume. The Topolino test generated thousands of clicks; the problem was that almost none converted into meaningful intent signals. A smarter first exchange is worth more than optimizing ad spend.
The design challenge wasn't building a chatbot — it was building a conversation that felt natural enough to keep someone in the thread past the first message. At that layer, interaction design is closer to writing than UI.
What I'd validate further: running this flow head-to-head against the original Topolino benchmark, and testing whether WhatsApp-booked test drives show at a higher rate than form-generated leads.
💡
Quality over volume
A smaller number of high-intent conversations beats a large pool of passive clicks every time.
✍️
Interaction design is writing
The hardest decisions were about tone, pacing, and what the AI says — not layout or components.
📐
The activation gap is a UX problem
271K clicks to 3K activations isn't a media failure — it's a friction design failure at the entry point.
🔬
Numbers need validation
+390% and +63% come from early-stage data and need controlled benchmark validation before interview-ready claims.
Problem
Traditional channels were running
out of road
Fiat's lead-gen relied on SMS and static web forms — channels with declining returns. An initial WhatsApp campaign showed strong ad engagement but the bot barely activated. The gap between click and conversation revealed a structural problem: the flow wasn't built for real people.
"271,000 link clicks. Under 3,000 bot activations. The ad wasn't the problem."
01
Diminishing returns on traditional channels
SMS and form-based capture were yielding fewer, lower-quality leads each cycle.
02
Low bot activation despite high ad engagement
The Fiat Topolino test: 271K link clicks — under 3K bot activations. A 99%+ drop-off at step one.
03
Static, one-size-fits-all conversation flows
Menu-driven bots couldn't adapt to different intents — browsing, comparing, or ready to buy.
Fiat Topolino — WhatsApp Test Funnel
Link clicks
271,000
WhatsApp opens
~18,000
Bot activated
2,970
Qualified leads
~356
99%+ of ad clicks never reached the bot — the static flow was the bottleneck.
Fiat Topolino WhatsApp campaign data. Used for diagnostic framing.
Research
Users haven't moved to messaging — they already live there
Shifts toward messaging over calls and forms were well-established before this project. Add shrinking attention spans and privacy-driven signal loss weakening traditional ad targeting, and the case for conversational channels becomes structural, not tactical.
Users aren't asking for less information — they want it delivered conversationally and calibrated to where they are in the decision. Browsing intent and ready-to-buy intent require completely different responses.
85%
of users prefer messaging over calls or email for brand interactions
Meta / McKinsey messaging preference data, 2023.
↓ Signal
Cookie deprecation weakens traditional ad targeting signals
3×
higher open rate — WhatsApp vs. email campaigns
What we learned
Ad-to-chat is the new ad-to-lead-form. Each conversation becomes a first-party data source — richer and more reliable than an anonymous form submission.
Ideation
From scripted menus to a context-aware agent
Early exploration compared two approaches: a rigid menu-driven bot versus a context-aware AI agent capable of real-time Q&A and personalized follow-up. The menu-driven path couldn't close the activation gap — the same failure mode as the Topolino test.
Old approach
Static Chatbot
✕
Pre-written menu trees
✕
Same flow for every user
✕
No open-ended Q&A
✕
Drops unexpected inputs
✕
Requires manual updates
Proposed
AI Agent
✓
Context-aware responses
✓
Adapts to user intent
✓
Real-time Q&A on any model
✓
Handles open-ended inputs
✓
Learns from conversation
Testing interaction flows early
The prospect-scoring approach identifies intent through opening questions — budget, timeline, current ownership — and branches accordingly. Low-propensity users get exploration content; high-propensity users move directly toward dealer booking.
Prospect Intent Scoring
Low
Browsing, no timeline
→ Educate & inspire
Medium
Comparing models
→ Compare & clarify
High
Ready to buy
→ Book test drive
Four-stage design process — each phase had a defined output and a stakeholder sign-off gate.
Designs
One thread.
Every step of the journey.
The final experience lives in a single persistent WhatsApp conversation — no app downloads, no form redirects. From the first tap on a Meta ad through model discovery, dealer selection, and test-drive booking, everything happens in the thread.
End-to-end WhatsApp AI flow — ad click through test-drive booking.
The Personalization Engine
Opening signals — timeline, ownership, budget — feed a propensity model that determines which branch the AI takes. This data also populates the CRM, giving the sales team qualified leads with intent context attached.
💬
Entry Questions
Budget, timeline, current car — surfaces propensity signals
🔍
Low · Score 1–3
Exploration content, inspiration, soft CTA to configurator
⚖️
Medium · Score 4–6
Model comparison, pricing deep-dive, FAQ resolution
🚗
High · Score 7–10
Dealer locator → test-drive booking → calendar confirmation
📊
CRM Handoff
Lead data + intent score passed to sales team with full conversation log
Prospect scoring logic — determines AI branch and CRM lead quality tier.
Lessons
What this project taught me
Conversational personalization affects lead quality more than raw volume. The Topolino test generated thousands of clicks; the problem was that almost none converted into meaningful intent signals. A smarter first exchange is worth more than optimizing ad spend.
The design challenge wasn't building a chatbot — it was building a conversation that felt natural enough to keep someone in the thread past the first message. At that layer, interaction design is closer to writing than UI.
What I'd validate further: running this flow head-to-head against the original Topolino benchmark, and testing whether WhatsApp-booked test drives show at a higher rate than form-generated leads.
💡
Quality over volume
A smaller number of high-intent conversations beats a large pool of passive clicks every time.
✍️
Interaction design is writing
The hardest decisions were about tone, pacing, and what the AI says — not layout or components.
📐
The activation gap is a UX problem
271K clicks to 3K activations isn't a media failure — it's a friction design failure at the entry point.
🔬
Numbers need validation
+390% and +63% come from early-stage data and need controlled benchmark validation before interview-ready claims.