Product Designer • Credit Platform • Agribusiness
Reinventing Rural Credit
Less spreadsheet dependency, better decisions: end-to-end design of a credit intelligence platform for Itaú BBA.
Scroll to explore
Overview
Itaú BBA's rural credit operation ran through a critical, complex journey with high financial impact per proposal. The process depended on multiple tools, parallel spreadsheets, legacy systems, emails, manual reports, and individual knowledge to consolidate data before decision-making.
I worked as the Product Designer on the end-to-end redesign of the experience, connecting discovery, information architecture, prototyping, validation, documentation, and handoff. The product was structured into 13 sequential stages, with 321 screens, block-level validations, standardized components, and an interaction logic close to the sense of control users previously sought in Excel.
The result was the transformation of a fragmented operation into a single, traceable platform designed for faster, more controlled, and more confident credit decisions, without losing the analytical depth required by a highly critical financial product.
Business Context
Agribusiness credit approval requires a different lens from traditional corporate credit because it combines future production, operational risk, assets, succession, reputation, repayment capacity, and commercial dynamics. The project also took place under tight deadlines, different maturity levels across teams, parallel initiatives, and strategic decisions with real business impact.
- Journey involving client, officer, analyst, system, manager, approval, and bank
- Crop cycles, crops, production projections, pricing, costs, and inventory
- Economic groups formed by multiple individual and company IDs, farms, and subgroups
- Land, assets, liabilities, banks, investments, and suppliers
- Low tolerance for operational error and need for end-to-end traceability
The Problem
The official journey did not provide the sense of control the team needed to operate agribusiness credit. As a result, decision-making happened across the legacy system, Excel, email, manual reports, meetings, and parallel documents.
Complex Journey
Many stages, dependencies, business rules, multiple user profiles, and no single end-to-end workflow.
Spreadsheets as a Parallel System
Users maintained their own controls because the official system did not centralize data, calculations, and validations with enough trust.
Cognitive Overload
Analysts and officers spent energy gathering information before they could actually evaluate the proposal.
Risk and Rework
Data outside the system, lack of traceability, and errors discovered too late increased operational risk.
The challenge was not just redesigning screens. It was redesigning how agribusiness credit information should be structured, validated, and turned into decisions.
Product Goals
Structure Complexity
Turn an extensive journey into 13 sequential stages with tabs, progressive logic, and block-level validation.
Reduce Operational Risk
Centralize data, reduce spreadsheet dependency, automate calculations, and surface inconsistencies earlier.
Scale With AI and Evidence
Use AI to accelerate research synthesis, insight organization, hypothesis generation, test analysis, and deliverable standardization without replacing design judgment.
Discovery and Research
Discovery combined desk research, a CSD matrix, As Is journey mapping, in-depth interviews, and alignment with involved areas. The goal was to understand the real journey, not only the documented flow, identifying where the official experience failed and why users created parallel controls.
I used AI as support during discovery to accelerate the organization of qualitative inputs: clustering recurring pain points, comparing user profiles, turning interviews into actionable themes, reviewing hypotheses, and structuring questions for new research rounds. AI helped speed up synthesis, while final decisions remained grounded in evidence, business context, and validation with real users.
Key Insights
- Excel was used because it gave users a sense of control, not because it was preferred.
- The SLA could reach 30 days even when the proposal moved forward without major issues.
- Officers, analysts, and managers had different needs within the same journey.
- The journey needed to combine guided structure with autonomy for complex data analysis.
- The core problem was not too much data, but a lack of clear progression, traceability, and trust in the official system.
MVP Definition
After To Be mapping and workshops with users and stakeholders, the MVP was organized into 13 sequential stages. Each stage was designed to reduce ambiguity, enable progressive validation, and bring the experience closer to the mental model users had built in spreadsheets, but with governance, traceability, and product consistency.
Commercial Structure
Home, proposal list, groups, subgroups, briefing, and group history.
Financial Structure
Assets, liabilities, banks, suppliers, and investments.
Agribusiness Logic
Land, crops, projections, proposal, result, and approval.
Each stage was treated as a validatable module, with a minimum SUM score of 80% before being considered an approved experience. This approach reduced delivery risk, made squad alignment easier, and allowed the product to evolve through independent blocks.
Design Process
-
Discovery and Alignment
I mapped the As Is journey, pain points, tools, involved areas, and business constraints with commercial, credit, management, and risk teams. I used AI to support interview synthesis, organize pain patterns, and turn scattered findings into prioritizable hypotheses.
-
To Be and MVP
I facilitated the design of the future journey, defined business rules, and turned the flow into 13 sequential stages for the MVP, balancing technical feasibility, operational risk, business value, and user clarity.
-
Wireframes
I created low-fidelity flows to validate structure, progression, data hierarchy, and rules before investing in final UI. At this stage, AI supported microcopy review, exception scenarios, and consistency across stages.
-
High-Fidelity Prototype
I designed final screens with a design system, tables, tabs, states, feedback, microinteractions, and composition focused on cognitive safety. I also organized reusable patterns to accelerate delivery across squads.
-
SUM Testing and AI
I validated stages in Maze, observed sessions, tracked doubts and errors, and created an AI agent to standardize and accelerate SUM calculation, reducing operational analysis effort and increasing comparability across tests.
Solution
The solution integrated the control layer that existed in spreadsheets directly into the official journey: centralized data, progressive validations, automatic calculations, tabs, tables, and stage-completion feedback. The proposal was not simply to digitize an existing flow, but to redesign decision-making to reduce effort, increase trust, and make the process auditable.
Control Without Leaving the Journey
The experience incorporated tabs, tables, and familiar Excel-like patterns, but with governance, traceability, centralized data, and less dependency on individual controls.
Progressive Validation
Each block received completion and progression criteria, reducing late discovery of errors during analysis and improving the quality of information sent to credit teams.
Cognitive Safety
Typographic hierarchy, states, microinteractions, and feedback helped users understand progress, pending items, and next steps.
Standardization Across Squads
The journey was componentized to avoid divergence between versions, accelerate iterations, document decisions, and maintain MVP consistency.
Validation and Iterations
In a journey with critical decisions and high financial risk, qualitative feedback alone would not be enough. The team needed to compare versions, prioritize adjustments, and make evidence-based decisions. For this reason, SUM was adopted as the official usability metric, combining success rate, execution time, satisfaction, and perceived error.
During delivery, AI was used to accelerate test analysis, standardize result interpretation, support decision documentation, review naming, consolidate learnings by stage, and turn session observations into actionable adjustments for the team. This increased iteration speed without giving up human curation and user validation.
What Worked
- Stage-based navigable prototypes made objective testing easier.
- SUM enabled clear comparison between versions and journeys.
- The guided structure was perceived by users as reassuring.
- The AI agent reduced calculation effort and helped standardize usability analysis.
What Was Adjusted
- Visual hierarchy of tables, fields, and completion blocks.
- Completion, stage progression, and pending-item feedback.
- Component standardization to reduce variation across iterations.
- Microcopy for complex fields to reduce misinterpretation of financial and agribusiness data.
Results and Impact
- Full delivery of 13 stages and 321 screens for the rural credit journey.
- More than 50 tests and interviews throughout discovery, validation, and pilot.
- Peak involvement of 9 squads, requiring consistency, documentation, and componentization.
- Consolidation of multiple tools into a single working experience.
- Use of AI as an accelerator for research, test analysis, documentation, and deliverable standardization.
SLA Reduction
The cycle went from approximately 1 month to 2 weeks, according to results presented during the pilot phase.
Consolidated Tools
The journey reduced dependency on spreadsheets, emails, legacy systems, and parallel controls by bringing work into a single application.
Tests and Interviews
Real users participated in interviews, stage-based tests, observed sessions, and pilot-phase feedback.
Average SUM
The solution achieved an average SUM approval score above the minimum defined for stage progression.
Pilot feedback highlighted less dependency on spreadsheets, automatic calculations, clearer result visualization, and faster proposal completion. More than a visual improvement, the project repositioned the experience as a tool for work, control, and decision-making.
Learnings
Co-creation Reduces Cost
The 3-day workshop with real users required hours of alignment, but avoided weeks of scope changes after launch and helped turn implicit rules into product decisions.
Structure Is Also UX
The step-by-step guided journey seemed restrictive on paper, but in testing it was perceived as a layer of reassurance and clarity, especially in a flow with high operational risk.
Resilience Sustains Quality
In a critical product, defending method, evidence, and consistency was essential to maintain delivery quality, align squads, and protect the experience under tight deadlines.
What This Case Demonstrates
- Ability to work on complex, regulated, data-heavy products.
- End-to-end vision, connecting discovery, strategy, UI, validation, metrics, and handoff.
- Pragmatic use of AI to gain speed without outsourcing critical thinking.
- Seniority to turn ambiguity into a clear, validatable, and scalable architecture.
Next Steps
- Implement FullStory to monitor live data from the new journey.
- Monitor real behavior to identify continuous improvement opportunities.
- Add perception capture within the journey itself, using Likert logic and Itaú NPS at the end of stages.
- Create monitoring dashboards to connect real usage, friction points, stage performance, and the improvement backlog.