AI-Agent Case Study

AI Powered Conversational and Creative Agent Studio

Client

Indigenius

Industry

AI

Year

2021 - 2025

Role

UX & Implementation

The Brand

Indigenius is a multi-channel AI agent Voice platform that enables businesses to automate workflows across email, chat, SMS, and voice channels. As the lead product designer, I was responsible for creating an intuitive interface that made complex AI automation accessible to non-technical users.

The Product

An advancing African Language Models Designed to support the development of African-focused voice AI, making digital services more accessible through natural conversations across multiple regional dialects.

The challenge

This case study focuses on digitalizing local languages for Africans businesses drowning in repetitive customer communication tasks across multiple channels, but existing automation tools were either too complex for non-technical users or too limited in functionality.

As the sole Product Designer on the Indigenius team, I worked closely with a product manager, business analyst, frontend and backend engineers, and QA. I was responsible for the full design process from discovery to production, while also participating in cross-team design reviews and contributing to shared component work across Indigenius design system.

Problem

AI tools were not designed for Africans, languages, or contextual realities. Existing agents struggled with linguistic diversity, cultural nuance, and low-quality localized training data. Creating inaccurate responses or limiting adoption across key African markets.

Approach

My work focused on translating research insights into interaction models for agent creation, onboarding, task execution and embedding feedback loops that continuously refined model interactions based on user behaviour and support signals.

Impact

Measurable improvement in user satisfaction, with platform ratings increasing to 4.0/5. Simplified onboarding and improved task architecture reduced user confusion and led to a noticeable drop in support requests within six months.

Key Insights

Issues that made the adoption process slow and inconvenient.

First, 78% of users found existing automation platforms too technical and difficult to set up.

Second, Teams managed 4-7 different tools for email, chat, SMS, and voice support.

Third, Support teams spent 22.5 hours/weekly on repetitive, automatable tasks.

Fourth, 64% of customer conversations lost context when switching channels.

Conversation Agent

The Conversational Agent system is designed as a modular intelligence layer that enables users to create and deploy agents tailored to specific goals. Each agent can be configured with contextual awareness, allowing it to understand ongoing tasks, maintain state, and adapt responses based on user intent and previous interactions.

Beyond conversation, agents are extensible through integrations with both widgets and workflows. This enables them to trigger structured actions such as data retrieval, content generation, or multi-step processes, all from within a single conversational interface. By linking agents to these execution layers, the system moves beyond chat into task orchestration, supporting multiple actions while preserving coherence, continuity, and task relevance across interactions.

Creative Agent

The Creative Agent section is where the system shifts from static tools into something closer to a production studio. Users can generate and adapt content across voice, language, and media formats without switching tools or workflows.

At its core, it supports voice selection across a range of available characters and tones, including African and global languages to ensure cultural and linguistic relevance. Users can generate audio through text-to-speech, convert spoken input using speech recognition, and apply dubbing to localize or reframe existing Video/audio content. Each voice option is designed to maintain natural cadence while adapting to different emotional and contextual needs.

Beyond audio, the agent extends into multimodal creation. Users can generate video content directly from prompts, combining narration, visuals, and timing into a single output flow. This makes it possible to move from concept to finished media in one continuous system rather than fragmented tools.

The result is a unified creative layer where voice, language, and video are treated as interchangeable building blocks rather than separate processes.

Key Learnings & Reflections

Conversational AI focused on improving natural dialogue, intent handling, and reducing friction in task completion through structured onboarding and feedback loops. Creative AI explored generative expression, evaluation quality, and user control in ideation workflows. Together, both highlighted the importance of grounding intelligence in user context, iteration, and measurable product impact.