Wired claimed that Apple’s WWDC just proved that AI* is a feature, not a product. Then Marques Brownlee did a video on this, which is quite insightful.
*Implicitly, people nowadays refer to AI as Generative AI, including large language models to answer questions and diffusion models to generate photos and videos.
I have always appreciated Marques Brownlee’s exceptional consumer product sense. He had some points in his video that more people will use AI stuff in the long run as features more than a standalone thing.
He gave some examples on both sides of the argument:
Clubhouse is a great feature and a fading product as the main platforms (e.g. Twitter/X) have built the feature in their products.
TikTok has popularized the carousel short-form videos. Even though Meta and YouTube have adopted it in their products, TikTok is still the most popular version of it.
He argued that the feature has to reach the level of Snapchat or TikTok in order to succeed as a standalone product to defeat that it could be just a feature somewhere. And Humane and Rabbit didn’t reach that level and are failing as standalone AI products.
In general, I agree with Marques Brownlee’s points - on the consumer tech end, more people will use AI as a feature than as a product.
Incumbents have added AI as features and pushed through their distribution to reach a vast amount of users. Microsoft integrated OpenAI’s ChatGPT in Bing as soon as it hit off and had high expectations to use this feature to grab more market share in Search (the most profitable segment of software). It has been also integrated into Office365 as Copilot. Instantly, millions of Microsoft’s products’ users could access to these features.
The classic innovators’ dilemma didn’t hold back the incumbents this time. Credit to the CEO of Satya Nadella, unlike his predecessor Steve Ballmer who has missed Mobile computing, has been one step ahead in innovation (invested in OpenAI since 2019) and deployed the technology soon after the public launch of ChatGPT. Google under financial market pressure, played catch-up relatively fast, even though it is unclear how Generative AI will disrupt its business model, particularly on Search (Google’s main cash cow).
So far only a few true outliers are succeeding standalone AI products. OpenAI had a tremendous first-mover advantage. ChatGPT was the fastest adopted technology in history and has become a household brand. ChatGPT as a brand has also become one of OpenAI’s biggest assets (now directly under the domain chatgpt.com). The new domain had 2.5 billion visits in May according to SimilarWeb. Perperlexity AI was another very appreciated outlier by users, especially by knowledge workers (e.g. product managers), which had 85 million visits in May. But we haven’t had any other global phenomena. Both two highly mediatized AI startups Humane AI and Rabbit have been in trouble since their latest product launch and bad reviews.
Consumer tech is notoriously difficult to break into. It has a tremendously high bar for the product and user experience. Historically, it is a space where winners take all and it takes an astronomical amount of funding to get there. The last winner in the mobile computing era Uber has raised 22 billion USD to get where they are today. In comparison, OpenAI has raised 11 billion USD so far.
In summary, incumbents have the cash, resources, and strategic awareness to push the technology as features through their existing products and distribution channels. Very few AI startups have truly made it so far in the consumer tech space. It is not counter-intuitive that most people will have access to the technology through incumbents’ product features.
Let’s cross-check with the micro-economic theory of demand.
Demand is a function of (price, income, substitute products and complement product, taste and preference).
Price: the majority of the consumer AI products/features are free, freemium or included in a hardware package. ChatGPT 4o model is free or Plus version is 20 USD per month with more features. Many Android phones already have some Generative AI features like translation, text/image generation or modification.
Substitute products: as long as ChatGPT, Perperlexity, Google Gemini have free versions, by default, consumers have free substitute products. And the most important substitute products are the status quo products that people are using, doing things the same way as before.
Complement products: hardwares (phones and computers) keep getting better to support AI usage.
Taste and preference: Generative AI is a general purpose technology. People are naturally finding their use cases with it. More than a quarter of the population is reported to use Generative AI in their private lives in the six selected countries (USA, Denmark, France, UK, Argentina, and Japan) according to a survey conducted by Reuters Institute. The adoption is fast given we are only 20 months after ChatGPT’s launch. It is showing the change of preference of using Generative AI in populations, moving from early adopters to mass.
The demand is there and keep growing for free AI usage. Thus, it is extra-challenging for startups to come up with standalone consumer AI products as long as there are free versions of ChatGPT, Gemini, Perperxity which are doing an incredible good jobs as general purpose product and more AI features are coming on IOS and Android that are also free to use. On top of that, it comes with the challenge to reach to billions of people, which requires billions if not more funding.
What about unbundling some AI features into standalone products?
“only two ways to make money in business: one is to bundle; the other is unbundle.”
Jim Barksdale, a former CEO of Netscape
I think unbundling is in general working better for BtoB startups. For BtoC consumer tech, we observe more often the reverse, bundling - incombents integrate a cool product as a feature, thinking of Story, short-form videos, and now Generative AI. This is not saying that we won’t see any outlier in the consumer tech space. If some startups come up with >10x better products, they just might win over a segment of the consumers.
However, for the mass, we will access the Generative AI as features, provided by incombents big techs, with exception of ChatGPT by OpenAI, which we could reasonably argue is becoming a big tech.
AI the Product vs AI the Feature