Nothing illustrates how software has created a paradigm shift from "money as granting ownership" to "money as granting access" quite like BMW's heated seats.

In 2020, BMW introduced a new version of its operating system that allowed the company to introduce subscriptions and micro-transactions. Per an article in The Verge:

A monthly subscription to heat your BMW’s front seats costs roughly $18, with options to subscribe for a year ($180), three years ($300), or pay for “unlimited” access for $415… Other features that BMW is locking behind subscriptions (as per the company’s digital UK store) include heated steering wheels, from $12 a month; the option to record footage from your car’s cameras, priced at $235 for “unlimited” use; and the “IconicSounds Sport package,” which lets you play engine sounds in your car for a one-time fee of $117."

What was new about this situation wasn’t the idea that a carmaker might charge extra for high-end features, but that the software, not the hardware, would be the limiting factor. “In the case of heated seats, for example, BMW owners already have all the necessary components, but BMW has simply placed a software block on their functionality that buyers then have to pay to remove.” 

This is in line with the broader shift indicated by a term that hit exit velocity in 2023: "the subscription economy." It was supposed to pinpoint the new paradigm of how consumer-facing companies could make money on the internet. Although none of this was especially new to the year 2023 (subscription giants like Netflix, Spotify, Hulu, Amazon, Chewy, had all been around since at least the mid-2000s), the term was supposed to pinpoint that same financial shift I opened with – we used to pay for ownership, now we pay for access. 

From a business standpoint, the advantages were always clear. Upfront payments, recurring revenue, building loyalty, mining data from users. The idea is that these efficiencies would be passed along to the consumer in the form of reduced prices, more stuff, and a smoother transaction process.

However, that means that, now, the purchase is being constantly re-evaluated. The implications of this ongoing purchase process are far-reaching. Instead of a first "flash and splash" marketing moment to draw someone in and make the sale, the subscription model means that money becomes something new: a function of the continuous attention you can draw from your consumers. 

This is at the heart of one major tactic to draw attention that has defined the VC-powered internet startup age: you release an early version of a product/service that people like by enticing them with hot offers and low subscription fees, you operate at a loss but you have enough funding to continue to acquire users because your product is good and people like it. But then the funding starts to run out and you need to make a profit. What do you do? Increase subscription prices, introduce ads, create a million tiered options, increase the cost of shipping -- in short, make the product worse. People leave and companies run out of room for growth.

This happened with the flood of DTC startups who spent tons of money on targeted social ads only for Apple to allow users to opt out of being tracked for advertising purposes: suddenly they had to answer tough questions about "unit economics" instead of just showing consistent growth, albeit growth that wasn't turning a profit. It's also happening with streaming companies who are upping prices and introducing advertisements as more and more people cancel their subscriptions.

If you're going to sell access rather than ownership, you need a way to make that access feel worth it all the time. And if the “offer great product that you make worse when it’s time to make money” model is dying, I think it’s worth looking ahead to try and understand what other tactics companies might use to try and permanently hold the attention of their users, and why they would or wouldn’t work. 

Intimacy and the Parasocial

Attention is currency in a digital economy. But what makes something worthy of our attention? 

Amidst the hand-wringing over the future of subscriptions, there’s been a mini-rise in what I might otherwise call the “patronage” economy, best represented by platforms like Substack and Patreon and GoFundMe, to a lesser extent. What these platforms do is give you the chance to pay a particular person or group of people directly. Instead of reading The Verge, you're paying Casey Newton. 

The difference between paying a person versus paying a company is fundamentally an emotional one. They’re representative of what is popularly called “parasocial relationships” – ones in which one person expends emotional energy and the other person is completely unaware of the other person's existence. Our dollars are a sign of an investment in a person's success; we feel connected to our favorite podcasters. We might even feel that, given the right circumstances, they could be our friend. Their voices might be comforting in a way our friends' voices are comforting.

In these situations, it's not just our attention that we're giving, it's intimacy, which ultimately glues us to our favorite "content creators" much more effectively than paying for ad-free versions of a service. 

It's not as if large subscription services have completely given up on the quest to build intimacy with their subscribers. Personalized recommendations are a real step in the direction towards a two-way relationship: "we know what it is that you want" is not such a far step from assuming that "we know you." However, this is largely maintained by the shady black box that is "the algorithm." It shows you things, sometimes you know why, sometimes you feel like you need to train it to know you better. Can we ever feel the same attachment to our algorithm as we do to the people to whom we send our money? Whose voices might help us fall asleep or keep us company while we do mindless work tasks?

If we can, what would that look like?

“Like AI from the movies”

AI is becoming more emotional. That’s at least the message that the launch of “GPT-4o” conveyed. Making reference to the movie Her in which a lonely Joaquin Phoenix falls in love with his phone’s operating system, appearing to rip-off Scarlett Johannson’s voice from the same movie (lawsuit incoming), and repeatedly mentioning that this was now like “AI from the movies,” the presentation’s emphasis was about how the conversational interface of GPT-4o can now mimic and respond to human emotions in a more convincing manner.

Now, there’s nothing core to a chatbot’s functionality that suggests it needs to be able to respond with emotional nuance. You are asking and receiving questions from a computer. It does not need to use sarcasm or speak in the first-person to do this. But, seen through a different lens, this is an entirely appropriate response to the needs of users who want their digital experiences to make them feel something. Of course, the movie they chose to illustrate GPT-4o’s new functions is just as telling.  

Autonomous computers have long been a central motif to the development of the sci-fi genre. Some of them are campy; some of them are evil; some of them are nice; some of them are nice and then evil. Usually they have names like HAL, or K, or TARS, or Gigolo Joe, or Ava, or "The Terminator." Perhaps the starkest difference between Her and the rest of these movies is that the AI bot, Samantha, is mostly just there to talk to Joaquin Phoenix. In fact, they talk so much that Joaquin Phoenix falls in love with it. Whereas past iterations of AI bots have served as Frankenstein-style warnings about our ability to create things we can't control, this is a more minor key example of a bot filling an emotional void for a sensitive and slightly depressed Joaquin Phoenix. 

The relationship might not be real in the way it is between two people, but it might feel real in the same way we feel like we are friends with our favorite podcasters. It mimics the intimacy of a digital service that has become, well, irreplaceable by way of tugging at our heartstrings. One would assume that Joaquin Phoenix was more than happy to pay his monthly subscription fee. 

It’s easy, of course, to dismiss this as a speculative example from a movie. But that’s missing something crucial about our propensity to become intensely attached to all sorts of stories, objects, voices, and more. There are tales of medieval knights who fall madly in love upon hearing only of tales of a courtly lady's "chastity and virtue"; a major theme of Shakespearean comedies is that people will fall in love with anything (people in disguise, people with the head of a donkey, etc.). More recently, people have married video game characters, holograms, pillows, the Eiffel Tower, a duvet, the Berlin Wall. I could keep going.

From “the algorithm” to LLMs to digital assistants, we’ve become increasingly accustomed to the presence of abstract digital entities in our lives and it wouldn’t be so outlandish to say we form some type of relationship with them. GPT-4o then appears to follow a long lineage of non-human objects designed (explicitly or implicitly) to draw us toward them by tugging on our heartstrings. But whereas these other examples don’t really seem to benefit any large corporations or moneyed interests – the emotionally-nuanced chatbot seems to have a real economic purpose. The digital economy is reliant on continuous and renewed user attention; attention is stickiest when it’s emotional. The timing would be apt. Just as it seems that the “enshittification” of the internet by way of the declining quality of online products and experiences is complete, here comes a tool that promises a more intimate internet.  

Zooming Out, Economically 

My simplified schema of the brief history of the consumer-facing digital economy is this:

1. Subscription as value-add: companies whose services basically make sense as digitized platforms you pay to access (e-commerce and entertainment are the obvious examples)

2. Personalization without emotion via algorithmic recommendations. The era of digital “over-abundance” and “digital exhaustion” and “enshittification.”

3. Personalization via emotion. Technology that replicates the feeling of a two-way relationship as a way to build deeper bonds between what’s ultimately a company and its consumer.

In this sense, the AI Chatbot is the logical progression towards the solution of a problem that was endemic to the digital economy as such: it lacked heart. Paying to unlock the heated seats that we’re already part of the car is one version of “software eating the world”; the next version might be paying for a car that’ll convincingly replicate talking to a friend on a road trip. Which of these seems the likeliest to lead towards customer retention? 

Curated tech-driven emotional experiences are like the final boss of personalization. However, the danger of personalization is that we end up in our own echo chambers, being spoon-fed things we're already assured to enjoy. The idea that we don't yet know what we want, or that what we want does not equate to who we are, simply doesn't exist. 

There is no notion in which we might be transformed by the people, things, and experiences we are brought into contact with, that part of the joy of human relationships, parasocial and otherwise, is that we find something new out about ourselves in the process of being in contact with another person. It's possible that a bot could facilitate that kind of experience. But in order for that to happen, it wouldn't need to find ways to awe and amaze us like a magic trick from “the movies,” it would need to find new ways for us to become surprised by ourselves.