Before implementing AI to solve a need or a pain point, it is critically important to assess whether AI truly improves the user experience or introduces complexity.

Examples where integrating AI is better

1. Providing personalised recommendations to individual users based on their past viewing habits and preferences.

2. Detecting fraud in online payments or account hacking

3. Image or voice detection in a robotic system

4. Predicting future events like weather forecast or stock market
   

Often, Hueristic-Based (IF/Else) solutions are simpler, easier to build, maintain and debug.

Examples where AI is not better:

1. Static Interfaces like the "Home" button, that stays in a place where you can always find it.

2. Open-source software where users need clear and understandable code for implementation.

3. A standard Signup form or payment gateway interface

Lack of user data to surface personalised recommendations

How to use this pattern

1. Determine the friction points in the customer journey

2. Assess AI feasibility: Determine if AI can address the challenge effectively. Evaluate costs, dataset availability and ROI.

3. Validate Mental Models

   - Determine if AI solution aligns with existing workflows? 

   - Determine if AI solution erode user expectation? (Automation vs Augmentation vs Manual control)

   - How are probabilistic errors handled?