The Challenge
OCP’s sales process is long and complex: supplier quotes, logistics, delivery tracking, and negotiations with clients. Recently, the company began introducing AI agents to automate parts of this workflow.
But a clear challenge emerged:
Sales reps didn’t fully trust the AI.
They lacked visibility into what the AI was doing.
They feared losing control of client relationships.
Design challenge (from brief):
Enable reps to collaborate with their AI agent.
Show what the AI is doing at all times.
Allow humans to take over or guide workflows when needed.
Research & Insights
I created an empathy map and persona (Karim) based on the brief and assumptions about OCP’s sales context. Despite limited access to users, the exercise highlighted recurring themes:
Visibility builds trust → Reps need a clear log of AI actions.
Human-in-the-loop control → They must be able to override or guide workflows anytime.
Simplicity is key → Workflows should live in one central place.
Book a call and lets go through the empathy map and research together :)
Design Principles
From the research, I defined three guiding principles:
Transparency → AI actions must be logged in a visible, accessible way.
Control → Users can approve, edit, or dismiss any AI action.
Collaboration → Interaction with the AI should feel like working with a teammate.
Design Exploration
As the sole designer, I owned the full process from research to prototypes. I began with low-fidelity wireframes to validate flows before moving into high-fidelity design.
The hardest design problem was balancing transparency without overwhelming users.
Too much detail would distract sales reps.
Too little, and they would lose trust.
My solution: an activity feed with layered detail — clear at a glance, but expandable for depth.
Final Solution
1. Deals Overview
A high-level dashboard showing all deals in progress, with AI status surfaced clearly. Reps can instantly prioritize deals needing attention.
Image: High-fidelity Deals Overview screen.
2. AI Panel
A chat-like interface for natural interaction with the AI. Reps can query performance, request summaries, or guide AI actions as if speaking to a colleague.
Image: High-fidelity AI Panel screen.
3. Deal Details
A timeline that shows every AI action — emails sent, quotes generated, follow-ups paused — all with timestamps. Reps can approve, edit, or dismiss actions anytime.
Impact
This was a conceptual exploration, shared with the OCP team in Casablanca (April 2025). While not implemented, the design demonstrates how human-AI collaboration could improve enterprise sales:
Efficiency: Unified tools reduce context switching and manual updates.
Trust: Transparent logs let reps confidently delegate routine tasks.
Scalability: Framework for human-in-the-loop AI that can extend across enterprise workflows.
Reflection
What worked: Designing for transparency and control made the AI feel more like a partner than a black box.
What I learned: Trust in AI is not about intelligence — it’s about visibility, accountability, and giving humans the final say.
Next steps: If this project continued, I’d validate with real sales reps, measure adoption, and refine how much detail the AI feed should show.
