Human Factors in AI: Blending Design Thinking with CRISP-DM in Retail ~via John Andrews
It’s been too long since I shared one of John’s insightful posts. Here he demonstrates something he and I have been saying for a while now… The best use of AI isn’t to replace people, it’s to empower them. /Ted
Artificial Intelligence is often discussed in terms of models, algorithms, and data pipelines — but the human side of AI is just as important. My recent project for the Duke University / Coursera “Human Factors in AI” course explored how we can merge Design Thinking and CRISP-DM to design AI systems that not only work technically, but also deliver meaningful human value.
The Challenge: Smarter Signage in Retail
Retailers rely heavily on signage — both print and digital — to guide shoppers and promote products. Yet they face two persistent challenges:
Waste and inefficiency in printed signage (too much produced, often unused).
Low engagement with digital signage (poor placement, irrelevant content).
This is a human problem as much as a technical one: shoppers are overwhelmed, store associates are stretched thin, and messaging often misses its moment.
Applying Design Thinking: Human-Centered Discovery
The Stanford Design Thinking process helped me start with people, not data:
Empathize: Shoppers described signage as cluttered; associates said answering “where do I find this?” takes up too much time.
Define: Shoppers need timely, relevant signage because current layouts and static content create confusion and frustration.
Ideate: Potential solutions included AR wayfinding apps, AI-powered kiosks, dynamic digital displays, and more intelligent print allocation.
Prototype: Low-fidelity store maps, kiosk mockups, and signage sketches made ideas tangible.
Test: Quick trials showed strong interest in AR navigation and kiosks, especially among time-pressed shoppers.
Applying CRISP-DM: Data-Driven Execution
Once the human need was clear, CRISP-DM framed how data and AI could deliver solutions:
Business Understanding: Reduce signage waste and improve customer engagement.
Data Understanding: Use sales logs, planograms, shopper traffic data, campaign history, weather, and events.
Data Preparation: Map product SKUs to store coordinates; clean traffic patterns.
Modeling:
Model A (Batch): Forecast demand for printed signage per store/campaign.
Model B (Near Real-Time): Optimize digital signage placement/content based on traffic flows and engagement data.
Evaluation: Pilot in test stores; A/B test optimized vs. current strategy.
Deployment: Roll out to stores, monitor for drift, refine continuously.
Why This Matters: Human Factors in AI
This project underscored that AI must serve people, not the other way around. By combining Design Thinking (empathy-driven problem discovery) with CRISP-DM (rigorous ML execution), we can design AI systems that:
Save costs (25–30% less print waste).
Boost engagement (more relevant digital messages).
Improve experiences (shoppers find products faster, associates spend more time serving).
Key Takeaway
AI is not a magic wand; it’s a toolkit. When paired with human-centered design, it becomes a way to create systems that are efficient, ethical, and empathetic.
Design Thinking helps us find the “what” and “why.” CRISP-DM delivers the “how.” Together, they ensure AI works for people, not just around them.
✨ Thanks for reading! This project was part of the Duke University / Coursera “Human Factors in AI” course. I’d love to hear how you’re applying human-centered approaches to AI in your field.
#HumanFactors #AI #DesignThinking #CRISPDM #RetailInnovation #Coursera #DukeUniversity
