Fheem : A Specialized AI Assistant for Advocacy Campaigns in Arabic
Client: Internal ADSI Product Pivot
Sector: AI / Civic Technology / Advocacy Tools
Services by ADSI: Product Pivot Strategy, UX Copywriting, Language Localization, AI Architecture, Advocacy Design
1. The Challenge
The original Fheem app was built as a generic AI assistant, designed to serve a wide range of queries with fallback model support and dynamic response routing. While functionally solid, it lacked a clear user niche and narrative. ADSI challenged its own team with a redesign:
“Can we pivot Fheem into a high-impact assistant that supports Arabic-speaking activists, civil society groups, and advocacy campaigns—starting with Sudan?”
The goal was to repurpose the existing AI core into a purpose-built assistant for planning, writing, and strategizing Arabic advocacy materials—something that didn’t yet exist in the civic tech ecosystem.
2. Context
Many civil society actors across Sudan and the region operate with limited resources, inconsistent internet, and no access to Arabic-native tools for:
-
Campaign strategy and messaging
-
Writing compelling Arabic content
-
Translating complex policy goals into clear calls to action
-
Understanding audience sentiment and communication framing
While AI tools were becoming more accessible, few offered localized, culturally-aware, and advocacy-specific capabilities—especially in Arabic.
3. ADSI’s Approach
To reposition Fheem, ADSI applied a rapid design and strategy framework:
Vision Reframing
-
Defined a sharp niche: Arabic-first AI for advocacy design
-
Outlined key use cases: campaign planning, content writing, slogan generation, script drafting, localization advice
Language & Cultural Calibration
-
Set Arabic as the default language for all outputs, with Sudanese dialect used in key messaging where appropriate
-
Removed generic or insensitive terminology like “القضايا العربية” and replaced with nuanced alternatives
Interaction Design
-
Modified prompt-response architecture to reflect advocacy flow: from problem framing → narrative crafting → content execution
-
Prioritized ethical prompt behavior, culturally aware tone, and context-sensitive outputs
Technical Enhancements
-
Retained fallback model chaining for consistency (OpenAI, Groq, etc.)
-
Planned integration with Supabase Edge Functions to protect API keys and enhance serverless performance
-
Maintained support for future multimodal inputs (image, voice)
4. Design Solutions
The redesigned Fheem is:
-
Natively Arabic in tone, logic, and purpose
-
Structured around advocacy campaign flows, not just chat
-
Lightweight and browser-based, with responsive design
-
Contextualized for Sudan but expandable to other Arabic-speaking civic environments
-
Equipped with custom response logic that can output slogans, write-ups, strategic advice, and visuals on demand
5. Outcomes
-
Product Identity Clarity: Fheem now serves a clear audience with a sharp use case
-
Cultural Responsiveness: Key updates to language, tone, and framing resolved major alignment issues
-
Improved UX: The app’s interactions now feel purpose-built rather than generic
-
Deployment-Ready: The system is functioning and only requires minimal updates for full launch under the new brand direction
6. Reflections
Fheem proves that AI tools don’t need to be universal to be powerful. By narrowing the audience, refining the tone, and deeply embedding cultural understanding, the assistant became more impactful.
AI can be strategic. In a region where campaigns are often silenced, Fheem helps them speak louder—with clarity, cultural grounding, and purpose.
– ADSI, Product Strategy Team