MY STORY
THIAGO BRASIL
Product Designer with 7+ years of experience specializing in voice-first and conversational experiences for enterprise contact centers (IVR, chatbots, WhatsApp). Trusted by brands like TIM, Vivo, and Raízen to define flows, intents, prompts, and escalation paths that balance user experience with measurable business impact.
Proven track record of optimizing complex journeys, achieving a 71% CES score for Raízen’s HR platform and slashing financial onboarding times from 454s to 9s while boosting completion from 30% to 80%. Expert in data-driven instrumentation, GenAI/LLM prototyping, and seamless automated-to-human handoffs.
Drives end-to-end alignment alongside product, engineering, and business teams to transform regulated, high-stakes journeys into effortless user interactions.
Career
Professional Journey
Nearly a decade designing digital products from e-commerce and marketing roots to AI-driven UX across iGaming and conversational platforms. Click the arrow on any role for the full story.
Nearly a decade designing digital products, from e-commerce and marketing roots to AI-driven UX across iGaming and conversational platforms. Click the arrow on any role for the full story.
Chatbot
2022 — 2025
Take Blip — AI Product Designer
Designed conversational flows for enterprise chatbots at TIM, Vivo, Raízen, and Superbid, achieving a 71% Customer Effort Success (CES) score for Raízen’s HR conversational services (benefits, payroll) by redesigning complex user journeys, drastically improving resolution efficiency while accelerating new-flow delivery by 35% via a shared design system.
Designed conversational flows for enterprise chatbots at TIM, Vivo, Raízen and Superbid, retaining 72% of Raízen’s HR calls without human intervention and accelerating new-flow delivery by ~35% via a shared design system.
NLP
Design System
Machine Learning
↗
iGaming
2025 — PRESENT
Regulated B2C SaaS Platform — Product Designer
Reduced onboarding and KYC abandonment by 28% and cut 30-day churn by 15% through AI-powered personalization and continuous behavioral analysis across Casino and Sportsbook journeys.
Casino
Benchmark
AI
↗
Construction
2018 — 2022
Construliga — UX/UI Designer
Designed, prototyped and coded LIGAJOBS, an engineering jobs platform that scaled to 30,000 users through research-led personas, acquisition campaigns and analytics.
Personas
Acquisition
Qualitative Research
↗
Education
Always Learning
From a business degree in Brazil to a postgraduate in Australia, a UX bootcamp, and now back in the classroom for a diploma in experience design.
In progress
2026
Diploma in Experience Design Technology
Cruzeiro do Sul — Brazil
Postgraduate
2009 — 2010
Postgraduate Diploma in Marketing & Business
AIPE — Australian Institute of Professional Education, Australia
Bachelor’s
2003 — 2005
Bachelor’s Degree in Web Design Technology
CEINTER — Brazil
Certifications
UX Bootcamp — AELA, Brazil (2021–2022)
Artificial Intelligence Foundations: Machine Learning
Introduction to NLP and LLMs: Principles and Practical Applications
English / IELTS Prep — Greenwich College, Australia (2009)
How I work
Process that drives real outcomes
I don't follow a template. Each engagement starts by understanding what actually matters, then I work backward from outcomes, not forward from deliverables.
01.
Discovery
I start by naming the problem space, not the feature request. That means mapping which channels the conversation actually lives on, agreeing the KPIs we will be judged on, and writing down the guardrails and compliance limits before anyone designs a single turn.
Channel mapping
KPI alignment
Guardrails
Compliance scope
02.
Research with AI
Real transcripts, support tickets, and call logs go through an LLM to surface recurring intents, the queries that actually dominate volume, and how people really phrase them. I use it to compress domain knowledge into a first knowledge base, and to draft user and bot personas I can test tone against.
Transcript mining
Intent clustering
Knowledge base
Bot persona
03.
Flow and Logic
This is the architecture: happy paths, edge cases, and fallback strategy mapped in Voiceflow or FigJam. I define the intent-entity model or the RAG routing, and pin down every integration point, from API calls and CRM lookups to the exact conditions that hand a conversation to a human.
Happy paths
Fallback strategy
Intent-entity model
RAG routing
Agent handoff
04.
Draft with AI
With the logic fixed, AI does the volume work. It generates sample utterances and phrasing variants per intent, drafts candidate dialogues, system prompts, and few-shot examples against my flow, and produces synthetic user responses so I can walk complex branches in minutes instead of days.
Utterance generation
Phrase variants
System prompts
Few-shot examples
Synthetic dialogues
05.
Human Refinement
This is where the craft goes in. I rewrite the generated copy for brevity, tone of voice, and turn-taking, then assemble a working prototype to test whether context actually survives across turns. Error recovery gets the most attention: a fallback should sound like a person helping, not a system failing.
Tone of voice
Turn-taking
Context retention
Error recovery
Interactive prototype
06.
Validation and Testing
Nothing reaches production untested. I run the prototype with real users to check comprehension and task completion, red-team the system prompts for jailbreaks, off-scope questions, and hallucinations, then benchmark accuracy across edge cases with evals.
Usability testing
Red teaming
Hallucination checks
Evals
Edge case benchmarks
07.
Handoff
Engineering gets assets, not a slide deck: tested system prompts, few-shot examples, and guardrail instructions. Alongside them go flow diagrams covering logic, state changes, and API payload specs, plus documented fallback triggers, analytics events, and intent mapping.
System prompts
Flow diagrams
API payloads
Analytics events
Intent mapping
some My favorite Tech Stacks Tools
I build with intention. Modern workflow that bridges the gap between traditional UX design and high-velocity AI development.
Notion
My scratchpad before anything earns a spec: research notes, call transcripts, and the messy first draft of a dialogue. It doubles as my personal task list and lighter project wikis.
Confluence
My formal spec repository: flows, state IDs, and the approved message copy engineering builds from. It keeps dialogue decisions visible to product and engineering across sprint cycles, including why a fallback returns to the last step instead of ending the conversation.
Voiceflow
Where the conversation itself gets designed. I map the happy path, intents, fallbacks, and hand-offs, then walk the dialogue end to end before any bot logic is built.
Figma Tools
My primary environment for everything around the dialogue: message templates, component libraries, and the flow diagrams I use to review a whole journey with stakeholders.
Google Stich
When timelines are aggressive or budget rules out a full build, Stich turns Figma components into a working, interactive environment fast, so I can put a real screen in front of someone instead of a static mock.
Framer
For prototypes that behave like real code. I switch here when a flow depends on timing, transitions, or micro-interactions that a clickable mock cannot honestly test with real users.
Claude
My technical partner for the work around the dialogue: pressure-testing flow logic against edge cases, troubleshooting code-based implementations, and automating repetitive spec and documentation tasks.
Jira
Keeps conversation design inside the sprint rather than beside it. Every state, template, and exception becomes a tracked ticket, so dialogue work ships against the same deadlines and technical constraints as the rest of the build.
MY STORY
THIAGO BRASIL
Product Designer with 7+ years of experience specializing in voice-first and conversational experiences for enterprise contact centers (IVR, chatbots, WhatsApp). Trusted by brands like TIM, Vivo, and Raízen to define flows, intents, prompts, and escalation paths that balance user experience with measurable business impact.
Proven track record of optimizing complex journeys, achieving a 71% CES score for Raízen’s HR platform and slashing financial onboarding times from 454s to 9s while boosting completion from 30% to 80%. Expert in data-driven instrumentation, GenAI/LLM prototyping, and seamless automated-to-human handoffs.
Drives end-to-end alignment alongside product, engineering, and business teams to transform regulated, high-stakes journeys into effortless user interactions.
Career
Professional Journey
Nearly a decade designing digital products from e-commerce and marketing roots to AI-driven UX across iGaming and conversational platforms. Click the arrow on any role for the full story.
Nearly a decade designing digital products, from e-commerce and marketing roots to AI-driven UX across iGaming and conversational platforms. Click the arrow on any role for the full story.
Chatbot
2022 — 2025
Take Blip — AI Product Designer
Designed conversational flows for enterprise chatbots at TIM, Vivo, Raízen, and Superbid, achieving a 71% Customer Effort Success (CES) score for Raízen’s HR conversational services (benefits, payroll) by redesigning complex user journeys, drastically improving resolution efficiency while accelerating new-flow delivery by 35% via a shared design system.
Designed conversational flows for enterprise chatbots at TIM, Vivo, Raízen and Superbid, retaining 72% of Raízen’s HR calls without human intervention and accelerating new-flow delivery by ~35% via a shared design system.
NLP
Design System
Machine Learning
↗
iGaming
2025 — PRESENT
Regulated B2C SaaS Platform — Product Designer
Reduced onboarding and KYC abandonment by 28% and cut 30-day churn by 15% through AI-powered personalization and continuous behavioral analysis across Casino and Sportsbook journeys.
Casino
Benchmark
AI
↗
Construction
2018 — 2022
Construliga — UX/UI Designer
Designed, prototyped and coded LIGAJOBS, an engineering jobs platform that scaled to 30,000 users through research-led personas, acquisition campaigns and analytics.
Personas
Acquisition
Qualitative Research
↗
Education
Always Learning
From a business degree in Brazil to a postgraduate in Australia, a UX bootcamp, and now back in the classroom for a diploma in experience design.
In progress
2026
Diploma in Experience Design Technology
Cruzeiro do Sul — Brazil
Postgraduate
2009 — 2010
Postgraduate Diploma in Marketing & Business
AIPE — Australian Institute of Professional Education, Australia
Bachelor’s
2003 — 2005
Bachelor’s Degree in Web Design Technology
CEINTER — Brazil
Certifications
UX Bootcamp — AELA, Brazil (2021–2022)
Artificial Intelligence Foundations: Machine Learning
Introduction to NLP and LLMs: Principles and Practical Applications
English / IELTS Prep — Greenwich College, Australia (2009)
How I work
Process that drives real outcomes
I don't follow a template. Each engagement starts by understanding what actually matters, then I work backward from outcomes, not forward from deliverables.
01.
Discovery
I start by naming the problem space, not the feature request. That means mapping which channels the conversation actually lives on, agreeing the KPIs we will be judged on, and writing down the guardrails and compliance limits before anyone designs a single turn.
Channel mapping
KPI alignment
Guardrails
Compliance scope
02.
Research with AI
Real transcripts, support tickets, and call logs go through an LLM to surface recurring intents, the queries that actually dominate volume, and how people really phrase them. I use it to compress domain knowledge into a first knowledge base, and to draft user and bot personas I can test tone against.
Transcript mining
Intent clustering
Knowledge base
Bot persona
03.
Flow and Logic
This is the architecture: happy paths, edge cases, and fallback strategy mapped in Voiceflow or FigJam. I define the intent-entity model or the RAG routing, and pin down every integration point, from API calls and CRM lookups to the exact conditions that hand a conversation to a human.
Happy paths
Fallback strategy
Intent-entity model
RAG routing
Agent handoff
04.
Draft with AI
With the logic fixed, AI does the volume work. It generates sample utterances and phrasing variants per intent, drafts candidate dialogues, system prompts, and few-shot examples against my flow, and produces synthetic user responses so I can walk complex branches in minutes instead of days.
Utterance generation
Phrase variants
System prompts
Few-shot examples
Synthetic dialogues
05.
Human Refinement
This is where the craft goes in. I rewrite the generated copy for brevity, tone of voice, and turn-taking, then assemble a working prototype to test whether context actually survives across turns. Error recovery gets the most attention: a fallback should sound like a person helping, not a system failing.
Tone of voice
Turn-taking
Context retention
Error recovery
Interactive prototype
06.
Validation and Testing
Nothing reaches production untested. I run the prototype with real users to check comprehension and task completion, red-team the system prompts for jailbreaks, off-scope questions, and hallucinations, then benchmark accuracy across edge cases with evals.
Usability testing
Red teaming
Hallucination checks
Evals
Edge case benchmarks
07.
Handoff
Engineering gets assets, not a slide deck: tested system prompts, few-shot examples, and guardrail instructions. Alongside them go flow diagrams covering logic, state changes, and API payload specs, plus documented fallback triggers, analytics events, and intent mapping.
System prompts
Flow diagrams
API payloads
Analytics events
Intent mapping
some My favorite Tech Stacks Tools
I build with intention. Modern workflow that bridges the gap between traditional UX design and high-velocity AI development.
Notion
My scratchpad before anything earns a spec: research notes, call transcripts, and the messy first draft of a dialogue. It doubles as my personal task list and lighter project wikis.
Confluence
My formal spec repository: flows, state IDs, and the approved message copy engineering builds from. It keeps dialogue decisions visible to product and engineering across sprint cycles, including why a fallback returns to the last step instead of ending the conversation.
Voiceflow
Where the conversation itself gets designed. I map the happy path, intents, fallbacks, and hand-offs, then walk the dialogue end to end before any bot logic is built.
Figma Tools
My primary environment for everything around the dialogue: message templates, component libraries, and the flow diagrams I use to review a whole journey with stakeholders.
Google Stich
When timelines are aggressive or budget rules out a full build, Stich turns Figma components into a working, interactive environment fast, so I can put a real screen in front of someone instead of a static mock.
Framer
For prototypes that behave like real code. I switch here when a flow depends on timing, transitions, or micro-interactions that a clickable mock cannot honestly test with real users.
Claude
My technical partner for the work around the dialogue: pressure-testing flow logic against edge cases, troubleshooting code-based implementations, and automating repetitive spec and documentation tasks.
Jira
Keeps conversation design inside the sprint rather than beside it. Every state, template, and exception becomes a tracked ticket, so dialogue work ships against the same deadlines and technical constraints as the rest of the build.