Consumer tech and the build dilemma

Racing to the personal AI assistant

“You’ve got to start with the customer experience and work backwards to the technology. You can’t start with the technology and try to figure out where you’re going to sell it.”

— Steve Jobs

Recently two AI consumer products Humane AI pin and Rabbit R1 have been reviewed by highly influential tech review YouTube Marques Brownlee (aka MKBHD, ~19M YouTube subscribers).

The two videos have provided some very interesting insights. Disregarding the eye-catching purpose of these titles, they still said a lot about the products, “The Worst Product I've Ever Reviewed... For Now” and “Rabbit R1: Barely Reviewable”.

The conclusion of the reviews is while the visions of these AI assistant products are very compelling, the products are half-baked and do not give customers strong reasons to buy (for now).

I always hold high respect for founders who made 1 from 0. But MKBHD had a point: “Customers should make the buying decision based on what the products are today, not what the product could be in several years”.

The majority of customers are not VCs. They pay money in exchange for the current product value, not to fund the startups’ future roadmap (although some tech enthusiasts might do so).

The two startups in question:

• Humane Inc., which made the Humane AI pin (700 USD starting point + 20 USD subscription), has raised 230 million USD.

• Rabbit Inc., which made Rabbit r1 (200 USD), has raised 30M USD.

The concept of AI personal assistant is very compelling with many value propositions, such as:

• Jobs-to-be-done: taking notes, answering messages, searching for information, planning vacations, booking tickets, ordering food, etc.

• Pain reliever: reducing the efforts of doing the tasks above

• Gain creator: saving time

One could hire a human personal assistant and many do so. It is a privilege for a few entrepreneurs, executives, celebrities, etc.; it costs a lot of money; it is not 24/7. So it is not hard to imagine the value and market size if we can create a tech-enabled product providing true personal assistant experience.

Many big techs have already tried to build personal assistants through tech products: Apple’s Siri, Amazon’s Alexa, Google Assistant, etc. None of them has provided a true personal assistant-like experience and none has become a phenomenal success.

Then came the Generative AI / large language model, popularized by OpenAI’s ChatGPT. The world was amazed by what it could do. It truly offered a new way of interacting with machines. No wonder Bill Gates said he was blown away only by two technologies: graphical user interface (GUI) and ChatGPT. GUI enabled clicking on buttons rather than writing MS-DOS command lines and ChatGPT/LLM enabled our command by our natural languages. Andrej Karpathy referred to LLM as a kernel, that could be built into software in a way that couldn’t be done before.

We understood the value proposition of personal assistant and we got the new technology. Naturally, a wave of startups came out to build it.

Most startups are lean startups that build first and then learn from users’ feedback. That has worked very well for the past decade, especially for SaaS startups, creating many unicorns.

In the lean startup concept, the minimum viable product (MVP) is the key. It is the first version product with the bare minimum required features that are presented to the customers/users. It is the founders’ judgment call on what level the MVP should be. As one founder said wisely “You don’t know what is minimum and you don’t know what is viable” and “sometimes you don’t even know what is the product”, Eric Ries, cofounder of the lean startup movement.

During the process of building something out of nothing, there isn’t much data. It is mainly based on the founders’ conviction and judgment, plus slices of early market feedback. What founders are doing is juggling between the cash in the bank, speed-to-market, and the specs. Seeking for perfection, founders risk running out of money and somebody else might launch before them. Launch too soon with a product too half-baked, founders risk bad feedback from the market. That is the build dilemma.

This method is less risky for BtoB SaaS companies as founders could get early feedback from a selected few customers and build together. In addition, when the product is on cloud, it makes it easier to update the product. Some product-led companies often launch multiple changes per day — continuous learning and continuous improvement.

That became way harder when it came to consumer tech, particularly hard for consumer hardware tech products. In consumer products, the brand is a non-negligible factor to consider. The product materializes the brand. Bad products create bad brands, and bad brands remain in customers’ minds, preventing them from buying future products from the same brand — vicious loup.

In addition, social media and YouTube make it too easy for people to know about bad products. For example, MKBHD’s review on the Humane AI pin had 7 million views during only the first 2 weeks of the video since its upload.

Many founders see GenAI as an opportunity of a decade to change the consumer hardware category and many of them see something else powered by GenAI as a replacement for the smartphone, including Sam Altman, the CEO of OpenAI, who is partnering with Jony Ive, former Chief of Design, to invent a new native AI product.

It is very tempting to take a share of Apple’s 200 billion USD iPhone revenue. It takes great value propositions to change consumers’ behavior - either carrying an additional product aside from their smartphone or replacing the smartphone entirely. “You have to have a 10x better product to beat (established) competition. “, Ben Horowitz, General Partner of A16Z.

Humane AI pin and Rabbit r1 are cool and their future roadmaps are certainly compelling. However, for a consumer, the question is what do I do with them? It is easy to watch some tech reviews on YouTube. It is another decision-making process to actually spend money in exchange for the value delivered by the products. As MKBHD wisely pointed out, the majority of consumers buy based on the current product and their current problems/wants/needs.

So I think founders have to go back to the basics and first principles - aiming to deliver undisputable value propositions and extraordinary customer experience and work backward to the technology. That means to deal with the builder’s dilemma. No matter how cool a technology is, it remains fluff as long as it doesn't improve the customer experience.

My advice to founders who are tempted to build in the category:

• Leverage heavily on prototypes and customer feedback, experiment, and iterate ferociously.

• First-mover advantage is tempting but eventually the best product wins.

• Price is 50% of the value exchange equation. You cannot expect customers to pay full price for half-baked prototypes. Or you get roasted by the tech reviews.

• It might be easier to start with software. E.g. build a personal assistant app first (build new on old) and maybe once it is there, build hardware for it to achieve superior integration and user experience.