CONVERSATIONAL DESIGN · PUBLIC HEALTH · AI · 2025

Impulse Bot

Argentina's first chatbot fully dedicated to sexual health.

I designed and launched the conversational channel that supported 130+ people during Pride 2025 — from RSVP for the Impulse truck to post-event follow-up.

Undetectable and Untransmittable — Impulse Bot's two characters embracing on an orange background

2,249

Messages exchanged

196

Active conversations

82%

Opt-in for future communications

#1

Sexual health chatbot in Argentina

Role

Product Designer & Project Lead

Client

Impulse — AIDS Healthcare Foundation Argentina

Timeframe

Jun 2025 – Apr 2026 (pilot)

Channel

WhatsApp Business API

Status

Pilot closed — evolving to new technology

The context

WhatsApp was where people already were.

Impulse is the Argentine chapter of AIDS Healthcare Foundation, the world's leading HIV prevention organization. As part of its Pride 2025 presence, Impulse deployed an active truck in the march with testing, information and guest artists. The operation needed a channel that could RSVP and follow up with guests in real time, answer sexual health questions in a warm — non-clinical — tone, and build a contact base that could be reactivated in future campaigns, all with a small team and no call-center budget.

Traditional channels (forms, email, social DMs) broke down on at least one of those points. WhatsApp was where people already were.

Until this pilot, Argentina had no conversational channel built specifically for sexual health: information lived in static websites, long FAQs, or in-person consultations. A dedicated chatbot opened the door to something different — accurate info, in the user's language, in the user's channel, backed by a trusted organization.

How I built it

Five decisions that defined the product

01

Conversational mapping before touching the tool

Before configuring anything, I mapped flows on paper: what are the real questions? What tone do people use when asking about sexual health? What happens when the bot doesn't know? I spent afternoons reading real conversations on other Impulse channels to distill patterns.

02

Tone design

I defined a voice system: close, free of unnecessary jargon, non-judgmental, humorous when appropriate. The hypothesis was that if the user felt judged, they'd leave. The data confirmed it.

03

Intent-based flows

Three main branches: (a) RSVP and truck logistics, (b) sexual health info and testing referrals, (c) post-event follow-up. Each branch with a human fallback when the question stepped outside the script.

04

WhatsApp as the only channel

A counterintuitive call — usually you design multichannel. Here, the bet was to concentrate everything on WhatsApp to avoid fragmenting the experience or the support team.

05

Metrics from day zero

Dashboard for received vs. sent messages, active conversations, new contacts, opt-in. Without this, the pilot couldn't justify its evolution.

The bot's full decision tree: three main branches, global keyword handling, and human handoff at every escape point.

Visual system

Making "Undetectable" and "Untransmittable" human

For a sexual health chatbot to work, you first need to dismantle public health language. People don't enter a conversation with the phrase "undetectable viral load" in their heads — they come with doubts, embarrassment, urgency.

The bot's visual system was built around two characters that embody the global U=U slogan (Undetectable = Untransmittable): if a person living with HIV has an undetectable viral load thanks to treatment, the virus is sexually untransmittable. It's one of modern prevention's most powerful messages, and it's almost always communicated in medical language.

Undetectable (blue) came first. Untransmittable (red) arrived six months later — the actual time treatment needs to become untransmittable. They're twins. They hug. They jump. They feel scared, they hesitate, they celebrate. They show up in WhatsApp and break the idea that talking about HIV has to be solemn.

Undetectable — blue character from Impulse Bot

Undetectable

The blue one. Represents an undetectable viral load — the goal of antiretroviral treatment.

Untransmittable — red character from Impulse Bot

Untransmittable

The red one. Arrives six months later — the actual time treatment takes to achieve untransmittability.

"The twins aren't a sticker pack. They're a product decision: if the brand doesn't scare you, the conversation continues."
Download character manual (PDF)

Results

A small pilot that unlocked investment in a new platform.

Two activation waves — Pride June 2025 and the BA Pride March in October — generated the two spikes of the pilot. Between waves, the bot stayed active and contacts remained in the base, allowing reactivation with no friction.

Messages by period · Jun – Oct 2025

Activation wave Base activity
~900
Jun Pride BA
Jul
Aug
Sep
~730
Oct BA March
Sent vs. received
1,457
792
Opt-in vs. NGO avg (<30%)
82%
30%

196

Active conversations

133

New contacts

2,249

Total messages (792 received · 1,457 sent)

82%

Opt-in for future comms (108 of 131 contacts)

1.84:1

Sent / received message ratio

2

Activation waves: June Pride + October BA March

The 1.84:1 ratio is design, not noise

The bot speaks more than it listens because a useful sexual-health conversation includes context, options and referrals — not just binary answers.

82% opt-in is exceptionally high for an NGO

Market averages for NGO comms subscriptions usually sit below 30%. It signals the channel built real trust.

The two spikes aren't noise — they're the story

The bot acted as an amplifier in the two key moments of the Argentine LGBTIQ+ calendar. Zero churn between waves: contacts stayed in the base ready to reactivate.

What I learned

Four principles the pilot left behind.

Channel beats sophistication

A rule-based WhatsApp bot beat any "AI-first" solution we could've shipped on another channel. People don't open new apps to talk about sexual health.

Tone is product

Not a cosmetic layer. Every copy decision moved the conversation-continuation metric.

Designing for the human fallback is designing the whole system

Chatbots fail; what matters is how they fail. The human handoff was designed with the same seriousness as the main flow.

A well-measured pilot is the best case for evolution budget

Without the metrics, this case wouldn't be migrating to a more robust platform today.

What's next

The pilot closes and the project evolves into a new platform with generative-AI capabilities. The 131 opt-in contacts are the starting point; the tone and flow learnings are the conversational brand playbook.

Project presentation at Latin American Summit