Helping low-income entrepreneurs manage AI agents from their phone

Executive Summary
Unesco designed a mobile platform to help low-income entrepreneurs manage recurring business tasks, such as customer service, with AI agents.
As the first product designer, I was responsible for the design system and initial MVP.
- Rapid Iterative Testing and Evaluation (RITE) rounds
- +12
- Entrepreneurs Impacted
- 10,230
- Retention Rate (90-day)
- +47%

Problem
Low-income entrepreneurs struggled to delegate recurring business tasks to their AI agents. Further issues included:
Formal Education
42% did not receive formal education, and 7% were functionally illiterate
Informal Workflows
74% managed their business primarily through their smartphone
AI Adoption
Only 6% ever used an AI agent to complete a business task




Research
I conducted 50 user interview sessions with the target audience.
My objective was to understand what made AI agents difficult to manage during their daily work:
86%
Struggled to track their agents’ progress and know when they needed to steer AI models
84%
Wanted to delegate recurring business tasks without dealing with multiple chats at once
46%
Wished their AI agents required less prompting
Customer: Good morning. I’m ready whenever you are.
Moderator: Good morning, Rafael. Imagine you want help answering customer messages. Where would you start?
Customer: I’d open this agent. It says customer service, so I think this is the right one.
Moderator: Go ahead. Talk me through what you’re looking at.
Customer: There’s a space to describe my business. Can I explain it by speaking instead?
Moderator: What would you expect to happen when you press that button?
Customer: I’d tell it what customers usually ask. Then I’d want to check the answers before anything gets sent.
Moderator: And where would you look to see how the agent is doing?
Customer: Here, where it shows the tasks. I’d look for anything waiting for me to review.
Moderator: Thank you. Let’s pause here and look at that together.
Design Process
Talking to users helped me define the product direction and specifications before using AI for rapid prototyping.
I designed a mid-fidelity interface for tracking AI agents’ progress from a mobile device. I also conducted moderated usability testing to improve existing features.
Round 0
Monitoring workflows: I designed a home screen so entrepreneurs could check their agents’ progress without opening each conversation.
Round 1
Voice interactions: Users with low literacy rates struggled to describe business problems through written prompts. I tested a voice call format so they could explain tasks aloud.
Retention rates increased overall by 21%.
Round 4
Reusable agents: I created templates for common tasks such as customer service, so users could run AI workflows without prompting from scratch. We also added a community page for sharing templates.




Results
The final prototype combined a centralized AI agent interface with voice input, reusable workflows, and a dashboard for tracking progress.
- Net Promoter Score
- +24%
- Session Retention
- +47%


Reflection
For low-income entrepreneurs, making AI accessible also meant reducing the effort of working with it. That became the focus of features like voice prompting, agentic monitoring, and reusable-workflow templates.
That's why we started the design process with discovery interviews, which helped better align the product direction. For instance, it was only after directly talking to users that we understood the value of workflow templates, since many low-income entrepreneurs did not know how to prompt AI.

Credits
- Creative Direction
- Julio Caggiano
- Executive Leadership
- Natalia Montibeller, Alex Lima, Brandon Tavares




