Differentiate the user experience to customize it for novice and expert users.

Provide a progressive experience, do not show all AI features at once, first display easy & familiar capabilities; and then, as the user interacts with the app, provide more advanced functionalities.

Maintain an open mind to use Google AI Design Patterns but go beyond, by applying them, experimenting, learning, and identifying new ones applicable to your product, industry, and use case.

Prioritize recommendations for users, not all recommendations are relevant to the context, needs, and moments of shopping. 

Do not over-explain functionality and focus on feedback. Are recommendations meaningful to users? How are they perceiving the application? How is their experience evolving?

Consider relevant recommendations beyond the traditional frequency & shopping patterns. Consider, for example, healthy, bio, and sustainable products, which are based on users' lifestyles, values, and interests.

Take into account physical aspects that complement & enhance the digital experience, e.g., humans (staff) supporting/guiding users in case of questions in the shop, arrangement of products, and the overall ecosystem that supports the UX.

Critical point: users want to feel good about completing the experience. Once they finish shopping, make sure to provide them with meaningful notifications, bonuses, vouchers, or memberships that make them feel good about using your app, buying your product, and visiting your shop.

It is important to think about key criteria/decision-making for ML to gauge users' needs to provide relevant recommendations, e.g., diet, labels, etc., to incorporate into the onboarding.

Humans remain a key alternative to errors in ML, to help customers. How much human touch do users want/expect?

Important: provide alternatives to users to move forward! Alternatives to pop up windows/messages. How should we inform users when products have been added to the cart? Immediate, effortless notifications.

Decision-making tools before identifying errors in AI could be helpful to better determine possible errors, esp. for not especialized people.  Determine what are false positive & negatives in advance for the particular use case.

Always inform users of their actions. How to do it without annying users? Vibration? what level of info/attention we want to take from users. Different severity/level notifications.

Provide users incentives to improve the recommendation engine for themselves (individually) and the whole community/users, e.g., go beyond how it helped accuracy to include more human things, e.g., it helped x amount of people eat healthier/buy more sustainably. Let them know how that helped improve recommendations.

Knowing who are the users liking/disliking, recommending products. For AI we need more accuracy, it is more than like/dislike, need of a new pattern? We need to know a bit more of the users providing recommendations/reviews about a product so that users know if they can trust it. 

Follow up with users to let them know how the organization is planning on improving the experience, once they complete shopping and during beta version.

What is the impact of exposing users to unwanted unethical experiences,e .g., unpaid products because they were not identified. Be proactive, inform users in advance that something may go wrong, its not their fault. 

Integrate AI features in the product marketing strategy. Inform users about enhanced experience. Embed it in the branding, build familiarity.

Set expectations in the onboarding, so that users know how the experience will run/work. Do a test run.

Generate an early & quick prototype of the application, to learn early about users' mental models, possible sources of errors/mismatches, need for explainability, edge cases, limitations, etc. before going through the AI design workshop.

How to create trust if people haven't even tried the product before? Triptech (Google method). How to tell about the shopping experience? Handle mental models.

Data privacy remains a big concern.

Important to have interdisciplinary teams to address different aspects of product design and development, e.g., business development, marketing, design, software engineering, etc.