Are we in favor of Amazon’s design decisions? There’s no doubt that we are all extremely reliant on Amazon as a delivery service. In fact, as I’m writing this, I had sworn to limit my Amazon spending and purchases in 2025… buuuut buying online and getting packages delivered directly home is so convenient that sometimes I break my own rules.
Amazon is also a powerhouse in terms of its streaming services, as well as its drop-shipping services and AWS services for small businesses and enterprises alike. Amazon is a giant, a corporate behemoth, too big for its own good.
And we as consumers walk a fine line between over-reliance on the convenience their services offer, while also being critical of the way that they run things….
How many times have we read negative news headlines about Amazon’s poor treatment of warehouse workers, delivery drivers, even under-cutting small businesses that sell via their e-commerce platform?
But let’s dig further into Amazon’s UX e-commerce practices, some might even be described as dark patterns that they utilize to sell aggressively to consumers:
Their one click buy is probably the most famous. Amazon is the only company to have patented the one-click buy. Apple happens to license this capability for their Apple music store.
Amazon defaults to subscribe when you try to purchase an item from the e-commerce platform, rather than just letting you make a one-time purchase right away.
Amazon routinely asks you to upgrade to Prime for faster deliveries, making it very hard to say no, especially when you want a rush order.
Let’s also get into Amazon’s Alexa, and when it made news headlines because of putting in accidental orders after ‘hearing’ things in passing or from TVs in homes.
This brings us to AI design and designing for uncertainty. Alexa’s AI in this case worked by taking in inputs (i.e., overhearing our voices, noises, commands) and creating an output (i.e., retrieving information, ordering).
Every AI output has some amount of error rate- for example, an AI system can be 99% certain that the output is correct or it can be 40% certain that the output is correct. Depending on how it was designed, it will suggest the output anyway.
Clearly in Alexa’s case, Alexa executed on orders with high uncertainty rates (i.e., overhearing conversations or the TV) which only shows how Amazon leveraged dark patterns to keep aggressively selling via AI and Alexa. No surprises there.
But what was the cost of their dark patterns and high error outputs from Alexa’s AI? It cost them winning the AI race. Everybody stopped trusting Alexa, only using the device for basic tasks like checking the time, calling a friend or family member, etc.
And then ChatGPT came along… The clarity of ChatGPT outputs, giving us options when the AI was not so certain about our requests or inputs, is reassuring. It builds trust. And this is why ChatGPT is where it is today, while Amazon’s Alexa is left collecting dust in people’s homes. Yikes. Comparing ChatGPT to Alexa is like comparing night and day.
This is why ethical AI design is so incredibly important. Amazon leveraged dark patterns and short term sales goals, completely missing out on billions of dollars of AI market share because they favored short term profits over long-term investment into an innovative AI product.
On the other hand, ChatGPT focused on rigorous training of their large language models (LLMs) with thoughtful, nuanced and complex responses (AI outputs) to user requests (inputs). And I’m not saying that ChatGPT is perfect! All products always have room for improvement, which is what makes design so exciting, a constant work in progress.
We created a 0-1 AI Product Innovation course ($1490) that covers building AI products and features, prototyping them with existing LLMs and also considering the user perspective to truly create something worth using.
The course goes above and beyond using basic LLMs to cover:
What is AI? How does AI work? What are some parameters to understand when working with AI systems? How do complex AI functionalities get created by chaining multiple AI systems together?
How AI & People Connect: How do AI inputs and outputs influence the way AI connects with people? To what extent should a user experience be automated? To what extent should a user have control over the outcomes of the experience? Think driverless cars and the spectrum between driving yourself versus AI completely taking over. What happens if there’s an obstacle on the road? Who takes charge? How is control handed off between an AI system and their human user?
AI Design Heuristics: What do we need to communicate to users who are using AI features and products? What do they need to know before using AI, during using AI, and after using AI? To what extent are they comfortable with AI learning about them via their data over time to become more accurate with outputs and personalize/customize outputs for that user?
Prototyping AI Products: How might we leverage LLM’s like ChatGPT and Gemini to understand the scope of what AI is capable of today? How might we leverage Wizard of Oz prototyping to understand how comfortable users feel when interacting with AI capabilities we might be dreaming up? We need to consider both technical prototypes and user-centric prototypes when building AI features and products.
AI Chatbot Design: To what extent should AI chatbots and social AI systems mimic human characteristics? How relatable should an AI system feel to users? How might we balance our AI designs for relatability, but also distinguish to users that they are using AI?
AI Generated Content: How might we distinguish between AI and human creations in a world where AI generated content will increase all over the internet? To what extent can we trust communications we see online?
AI Simulations: How might we use AI simulations of human behavior to create economic, behavioral, and social models that accurately predict human behaviors and interactions and outcomes? How might we better understand the long term effects of AI products we design?
AI Ethics: To what extent will AI displace jobs and how might we correct for the negative effects of AI? What are the unintended consequences of AI products that we design?
Enroll today and schedule 1:1 mentorship calls with me to work through your ideas, as part of the course.
Also! We just updated our VIP remote UX jobs board and entry level UX jobs board (subscribe to paid to get access to passcode) with over 60+ new entry level and remote UX designer, product designer, and UX researcher roles.
As always, rooting for your success!
Samaya, your UX woman.
P.S. Subscribe below to gain access to VIP entry level UX jobs board ($6.67/mo), updated weekly. When you subscribe, you will also gain access our VIP job search resources, including:
Our UX resume template that has landed UX Woman alumni interviews at Nike, HP, Google, Reddit and more.
Our UX portfolio template that has landed UX Woman alumni interviews at Nike, HP, Google, Reddit and more.
Our UX 6-page job search checklist to remain accountable during your UX job search and put your best foot forward when applying to jobs.
Our 5-page LinkedIn job search strategies playbook for the entry level UX job seeker.
Our 3-page UX interview playbook to help you stand out as a candidate during the final rounds of your interview
Become a paid subscriber for access! (see below)
