Tarrabiz: Designing Guided Conversations That Simplify Finance for Small Businesses

As a Generative AI Conversation Designer, I was responsible for creating a comprehensive, confidence-building guidance flow for Tarrabiz. My focus was on simplifying complex financial and technical concepts while guiding users seamlessly through key tasks.

To achieve this, I leveraged Natural Language Understanding, Large Language Models, and Application Programming Interfaces to design responsive, user-centered interactions. My process involved analyzing conversation transcripts to identify top user needs and call drivers, reviewing existing knowledge base content, and structuring clear, intuitive dialogue flows.

I also applied prompt engineering, intent modeling, and continuous refinement through Quality Assurance and User Acceptance Testing to ensure the experience remained accurate, helpful, and easy to navigate.

Scenario 1: “When do I get paid?” (Payment setup anxiety)

Scenario 2: “This is too much!” (Feature Overwhelm)

Scenario 3: “Am I doing this right?” (Confidence Gap)

User Challenge

A new user wants to set up payments but:

  • Doesn’t understand the required steps.

  • Fears “doing it wrong.”

  • Is intimidated by financial language

My approach:

  • Sets expectation (time = reduces uncertainty)

  • Uses reassurance (“I’ll guide you”)

  • Avoids jargon

Scenario: “When do I get paid?” Contd.

I ensure the Tarra bot boosts the user’s confidence by reinforcing their choice and using a light, positive tone without being unprofessional. It’s also crucial to reduce pressure by giving the user full control of the process.

Scenario: “When do I get paid?” (Setup completion)

The Tarra bot celebrates task completion to reinforce progress while smoothly guiding the user toward the next meaningful action.

User Challenge

User logs in and sees too many features:

  • Invoicing

  • Automation

  • Analytics

  • Customer management

They don’t know where to start.

My approach:

  • The proactive Tarra bot replaces dashboard overwhelm with guided entry points

  • Frames features as user goals, not system capabilities

Scenario: “This is too much!” Contd.

The Tarra bot is designed subtly to educate users while keeping choices limited to reduce cognitive load.

Scenario: “This is too much!” Contd.

The Tarra bot continues to maintain the momentum of setting time expectations while focusing on plain language as opposed to rules and logic.

User Challenge

User completes a task but feels unsure:

  • “Did I set this up correctly?”

  • “Will this work?”

My approach:

  • Validates user first (confidence).

  • Uses checklist format (clarity).

  • Highlights improvement gently.

Scenario: “Am I doing this right?” Contd.

The Tarra bot suggests improvement and includes a clear benefit to the user. It also avoids sounding too corrective or critical.