Cracking Consumer AI
The Search for Magic Apps
Stepping off the management consulting diving board and plunging into the world of startups, I was introduced to a new definition of a familiar word: “magic.”
When building a consumer application, magic is the gold standard for success – the sign of product market fit.
How do you know when you’ve created a magic app?
Oh, you’ll just know.
Symptoms of app magic include creating a product so beautiful, fun, and easy that consumers can’t help but tell all their friends about it. Under the hood, magic looks like steep, hockey stick-like curves of user counts with high, active user and retention numbers.
The diagnosis is easy, but getting there is incredibly hard.
Consumer AI applications have been, up until this point, an underwhelmed market. The biggest wins in AI have gone to the LLM developers themselves (OpenAI is now valued at $300B), and then, more recently, to B2B applications like HippocraticAI, Harvey, and Hebbia, which are proving real utility in streamlining workflows and seeing around corners of employee responsibilities. Building for Consumer AI is the hardest challenge of all. Supplemental to business basics like “building something people want”, to build a strong consumer AI agent you need three core things that are all individually incredibly challenging to achieve.
I. Data
People often talk about “trust” and “personalization” when evaluating AI applications. To me, those things go hand in hand and are both unlocked with good data. Like real relationships, trust between a consumer and an AI agent is built over time by repeatedly meeting or surpassing expectations – by being truly helpful. The way to be truly helpful is to truly understand the person you are trying to support. This is only possible with good data on both 1) who the user is and then 2) how to generate the most helpful output.
The challenge with Consumer AI is that data on consumers is not as readily available as it is with businesses. Companies have been storing their data for decades, providing a deep data lake for AI applications to connect to. With consumers, information is spread all over businesses and the internet, much of which sits in “walled gardens” where data is privately owned. Compared to B2B tools, consumers have even less patience to try and help an AI agent build data and understanding.
Social media platforms, arguably the best publicly accessible storage units for consumer data, dish information out to apps through APIs. These APIs restrict the types of accessible data, never providing the full picture. Incumbents like Meta and Google who already hold and control so much consumer data are, in my opinion, at a clear advantage when building for Consumer AI, unless you really knock the second pillar out of the park.
II. Design
Once a consumer AI agent understands you, it must tell you so in a way that’s clear and fun – like the best motivational speakers do. Good design is a way to differentiate in the crowded consumer ecosystem and turn an application into something that’s fun to use. Design is always important to apps, but even more so in the age of AI agents. Agents provide information. The more fun and less burdensome getting to the information feels, the more likely the consumer is to pick the app versus its competitors.
III. Making the impossible possible.
With killer data and design, a Consumer AI agent is probably most of the way to “magic”. Now, all it needs is one more sprinkle of something special: disbelief. Magic is realized in witnessing something that you never thought would be possible. This shock and surprise typically translate into an urgent need to tell all your friends, driving rapid, organic growth of the product. The go-to-market dream!
Where is the magic happening now?
Generative AI has been a known, popular technology for two years now, and I’ve tested countless consumer AI apps. But I never felt magic with anything, until recently when I first used Lovable. Lovable takes consumers from an “idea to an app in seconds.” All you do is type in what you want it to build, and it simply does.
As someone who has never been able to code but always wished to build products that require it, I was able to design and build frontends of websites, mobile, and desktop apps on Lovable just by prompting it through good old, regular words.
Lovable made me feel more capable. It allowed me to turn ideas into reality. It made me feel like I had magic powers. And so of course, I had to tell my friends.
*This is not a Lovable ad at all, even though it sort of reads like one.
Sounds very empowering and magical, indeed! 😊