Case 02 · Marketing · CRM · Products
A long form loses people. A well-built conversation guides them.
The financing chat replaces a credit form with a guided conversation: it only asks what applies to the profile, estimates an approximate payment, suggests cars that fit, books the visit and hands the full case to the salesperson so nobody repeats questions.
- Project
- Financing chatform
- Context
- Car dealership in Panama · now a ZoDesk module
- By Antonio Duque
- Reviewed against the code and documentation on
01
Context and need
People asking about financing want to know quickly whether they have an option and roughly what they would pay. A long form loses them, and when a lead does arrive the salesperson gets loose data without knowing which car they were looking at or which credit route makes sense.
The business also needed to know which question makes people leave, so it could be simplified.
02
My responsibility
- I designed the conversational journey, the product rules and the experience; I directed the build and its move into ZoDesk as a reusable module.
- The version that serves real customers lives on the dealership's website, maintained with its web team: that is where the final verdict is issued and conversions are recorded.
03
Strategy and decisions
- One question at a time, with buttons, so answering is easier than leaving.
- Conditional questions: each profile (salaried, professional, self-employed, taxi, school transport or ride-hailing) only sees what applies.
- The result is guidance, never approval: the chat never says “approved” or shows rates, and it states that the bank or lender decides.
- Measure the step without storing what the person types, and never double-count between the website and ZoDesk.
- One source of truth for the rules: the website server re-evaluates every case.
04
What I built
- A welcome with shortcuts and variants: resume where you left off (answers kept in the browser for seven days, without contact details), arrive from a specific car's ad, or from a type, brand or budget filter.
- Vehicle selection by use and type, or a no-car profile that answers “how much can I borrow?” with up to three cars that fit.
- A profile covering income, stability, documents, credit history and down payment, with different minimums for banks and lenders.
- An approximate biweekly payment based on the car's age and the credit route, always labeled as approximate.
- Visit booking with free days and times, a calendar file and a WhatsApp hand-off with a reference that contains no personal data.
- Lead and appointment submission to the website with an idempotency key and spam protection, and a log of every step for the funnel.
Real screenshots
Screens from the real system. ZoDesk ones come from its demo mode, with an invented dealership; those marked “Live” are public pages captured on November 14, 2026.
05
How it works
- Customer
- System
- Customer
Entry
From the website, an ad with UTMs, a specific car's link or a direct link.
- Customer
Intent
Shortcuts: finance a car, see the most viewed, “how much can I borrow?”, budget or sell your car.
- Customer
Vehicle
Pick a car, describe it, or let the chat suggest by use and budget.
- Customer
Profile
Conditional questions on occupation, income, stability, documents, history and down payment.
- System
Guidance
A possible bank or lender route, or a kind exit if they don't qualify today, with an approximate payment.
- Customer
Visit or WhatsApp
Book the visit or continue on WhatsApp without repeating anything.
- System
Hand-off
The website re-evaluates, saves the lead and notifies sales; ZoDesk receives it for assignment.
Demo
Try it: the journey with fictitious data
Financing advisor
Demo · fictitious cars, rates and answers
Good afternoon! What would you like to do today?
What the salesperson receives
- Source
- utm_source=demo · utm_campaign=portafolio
- Car
- —
- Answered steps
- 0
- Route
- In progress
- Internal notes
- None yet
Internal notes are never shown to the customer. The real system re-evaluates the case on the server before saving it.
Funnel events
- asesor_inicio
Each question is logged when asked, so the last one recorded shows where someone left. What people type is never stored. Try “Start over” halfway through.
06
Tools and what they do
- HTML and JavaScript
- Lightweight conversational interface with optional voice and local state to resume.
- FastAPI (ZoDesk)
- Public chat routes, funnel event log and attribution.
- Laravel (dealership website)
- Case re-evaluation, lead storage, sales notification and conversions.
- Turnstile and idempotency
- Spam protection and submissions that never duplicate on retry.
- Web Push
- Alerts to the salesperson when nobody attends the lead in time.
Available evidence
- Chat and route code, with automated tests for the website connection (24), WhatsApp pre-qualification (17) and WhatsApp forms (16).
- A decision document dated November 11, 2026 that turns the chat into a ZoDesk module.
- A per-question funnel panel built the same day (see the dashboards case).
07
Current status
- In production
Chat on the dealership website
Serves real customers; the verdict is issued by the website server.
- In production
Chat copy served by ZoDesk
Live since November 11, 2026 for demos and sales to other dealerships only.
- Built and tested
Lead and appointment submission
With idempotency, spam protection and automated tests.
- Connected
Per-question funnel and drop-off
Logs the step and field, never the typed text. Data since November 11.
- Built and tested
WhatsApp pre-qualification
Built and tested; the assistant is switched off in production.
- Pending
Own rules engine and per-client settings
Phase 1 of the product plan, not yet built.
“In production” is only used when the project’s documentation confirms it. “Built and tested” means automated tests exist, not that it is in use. “Proposal” is a documented idea, not something built.
Next steps
- Move the rules into an engine configurable per dealership.
- Offer the chat as an embeddable script and unify the inbox.
- Use drop-off data to simplify the question that loses the most people.
08
Lessons learned
A conversation works when each answer opens the right next question, not when it imitates a person.
Saying “this is not a credit approval” does not reduce trust; it increases it.
Measuring where people leave without storing what they type is possible, and it is the right way.
Related services
Sales pipeline
Lead capture, management and follow-up
Centralize leads, statuses, owners and next steps so sales works from a list instead of a memory.
Automotive
Systems for Car Dealerships and Workshops
ZoDesk brings together acquisition, follow-up, documents, collections and operations for dealerships, with a workshop variant.
CRM and sales follow-up
CRM Consulting and Implementation
I design the sales process before touching the tool: stages, assignment rules, alerts and loss reasons, connected to the website and channels.
Connected cases
Monitoring and action dashboards
Dashboards that don't stop at the number: every out-of-range indicator suggests a concrete action.
ZoDesk, a full-funnel ecosystem
From first contact to contract, collections and operations: the system that connects the pieces of the other cases.
Inventory-driven dynamic marketing
Inventory as the single source for catalogs, content, conversion signals and ad decisions.