2025 Final Take: Impressions of AI's Impact on Consumer Experiences

We’ve talked a lot about AI, and it hasn’t changed how we live our lives day-to-day. Headlines speak of AI bubbles and layoffs. Partly out of FOMO and partly out of reality, executives insist their employees use it. The reality is a mix of unbelievable business value and “the emperor has no clothes.” 

Here are my big takeaways (or themes) from 2025: 

Technology offerings exceed consumer interest - and more importantly needs. Today’s AI consumer experiences fall into three categories for me: 1) use of the generic or utility tools (e.g., Anthropic’s Claude, Google’s Gemini, OpenAI’s ChatGPT) 2) specialized apps (e.g., Canva, Character.ai, Gamma, Suno, or any retailer’s chatbot), and 3) devices (e.g., Friend pendant or the next Jony Ives’ device). Consumers are having fun creating videos or getting help editing, building fitness plans, proposing to a significant other, writing letters, and more. Influencers are creating content and extending their reach by using tools to alter the language, persona, or format for various social media platforms. 

My take: The uptake of the utility tools is impressive. Beyond that, “we’re experimenting, building services, and hoping consumers show up.” Quote from Albertson’s, “... the Albertson’s AI shopping assistant moves beyond search to execute complex end-to-end grocery tasks ….” My question, “why are grocery tasks complex?” They are promoting reducing shopping time from 46 minutes to 4 minutes with agentic commerce. Meanwhile, Sam Altman and his team have basically been quoted saying that the smartphone isn’t the right device for AI. They could be right IF there is a major shift in consumer behavior away from media creation and consumption (e.g., playing video games, reading the news or books, and scrolling through social media). Not everyone wants a Remy on their shoulder telling them what to do all day. I even find real-time spelling and grammar correction annoying. 

Enterprise CEOs got off easy by blaming layoffs on AI - and markets rewarded them. Companies such as Amazon, Intel, Microsoft, Accenture, Salesforce, Oracle, IBM, Cisco, etc. collectively announced tens of thousands of employee layoffs with references to increase AI capital spend (e.g., data centers), to work with fewer employees aided by AI, and more generally to reimagine businesses, products, and work with AI. 

My take: Needing capital to invest in AI is real, especially if you are one of the big hyperscalers or still mostly privately funded LLM-creators (e.g., Anthropic, Mistral, OpenAI, Perplexity.) There is proof that AI is creating efficiencies by automating workflows - especially code creation, using conversational bots (voice or text) to help customers with simpler queries, analyzing vast amounts of data in research scenarios, and creating content. Simplifying processes and saving 10 minutes (or hours) here and there doesn’t tend to add up to eliminating whole FTEs. And we don’t seem to be anywhere near the one-person, one-billion-dollar company Sam Altman thinks is possible. Anecdotally, scaling AI seems to be creating new roles for humans, so far. Moreover, not everyone knows what to do with the tech, and vibe coding isn’t just something anyone can do - really. 

Gloves are off - OpenAI needs to make money just like everyone else. OpenAI has released a number of products for enterprises and consumers. Beyond model updates, OpenAI has released a browser, shopping research, and the ability for third parties to integrate with ChatGPT (e.g., Figma, Instacart). On December 11, 2025, Disney announced a $1B investment in OpenAI, in which OpenAI is licensed to use more than 200 characters and elements in AI-generated images and videos. From the headline, one might think OpenAI would be paying Disney for those rights. No. Finally, in mid-October, Sam Altman announced that it would allow adult users of its platform to generate erotic content - a market that the Economist said would reach $2.5B this year and grow at 27% a year until 2028.

My take: While I hope I am wrong, we are circling back into traditional patterns. Offer consumers a great tool (for free). Build an audience. Collect consumer data. Monetize the consumers by using their data to improve your product (okay) or selling their data indirectly (e.g., advertising) or directly. Once you have a large enough audience and data, you can start to disintermediate other brands. You then offer partnerships to third parties to reunite them with their customers. Brands will borrow moments from these utility LLM apps if they truly drive revenue. In the meantime, they will fight to provide phenomenal experiences to their customers directly based on the data they already have. 

AI may be a bubble - that doesn’t mean it isn’t a transformational technology. With the circular flow of money, investments from private markets, government involvement, creative funding instruments, and relatively little revenue, many speculate that there is an “AI Bubble.” I am not a financial analyst. Oracle’s stock is down 20% in the past month. CoreWeave is down 20%+ in the past week. OpenAI last raised money with a valuation of $500B but is reportedly in negotiations to raise more money at a valuation of $750B to $830B. Expected 2025 revenue is in the range of $10B. 

My take: don’t confuse the idea of a bubble with the impact of this technology. I spoke to a well-educated friend over the weekend who believes there is an AI bubble because the power or impact of the technology is overstated. This technology will change the way we work, manage our health and households, conduct research, educate children, create and consume entertainment, and more. There are legitimately open questions on when and how much change we’ll see how fast. We’ll see change more quickly in low-risk use cases (e.g., creating fun videos for IG) and slower change in healthcare and other regulated industries, robotics, or experiences that impact customers directly. 

AI can be dangerous; we are moving too fast for society to truly comprehend its impact. 

Threats range from nebulous concepts such as “bias in the training data” and hallucinations to arguments around blame and accountability to open threats to society if AI really eliminates jobs or drives up the price of water and electricity by elevating demand beyond supply. In early December, Stein-Erik Soelberg killed his mother and then himself. Her estate is now suing OpenAI for wrongful death. (source: WSJ) Character.ai is turning off access to teenagers. A November 26th WSJ article, whose authors interviewed teens, reported that some were spending hours each day speaking with virtual companions. There is a long tail here of obvious and less obvious sources of harm, including deepfakes, control, manipulation, spread of misinformation, etc. States like California want to regulate AI. The federal government is fighting to prevent that, as are those who stand to gain the most. 

My take: There are too many ironies to unpack here. I’ll choose one: these applications can be like an echo chamber on steroids. GenAI offers consumers many gifts: superpowers, personalization, convenience, access, comfort, and adaptive content (e.g., education and entertainment). Virtual companions fit within comfort. Imagine living in a world where everything - our friends, co-workers, teachers, coaches, significant others - adapts to us and tells us what is easy or what we want to hear? Forget the decline in critical thinking that worried us a year ago. Now we’ll have adults who cannot adapt and resolve conflict, in addition to being unable to discern reality from fake information (or images or videos). Keep in mind that algorithms can be tweaked once product adoption is high. Take Waymo. I’ve watched them evolve from student drivers in 2000/2001 into careful, competent drivers in 2025. I often boasted to friends that I thought they were far safer than human drivers. They may still be, but they are also now rude drivers.