The “Agentic Web” and the future business of Music
It’s been a while. Some musing that I wanted to write down somewhere as they have been buzzing around in my head.
Prelude
In 1999, Nick Szabo (yes that Nick Szabo) wrote a paper titled Micropayments and Mental Transaction Costs, which goes a long way to describe how business will work on the internet. It is twenty-five years old, and yet still just as insightful as the day it was written.
As a fairly short piece, it’s definitely worth reading, but the key takeaway is that the limiting factor on business models will be psychological more so than technological, “We have seen how customer mental transaction costs can derive from at least three sources: uncertain cashflows, incomplete and costly observation of product attributes, and incomplete and costly decision making. These costs will increasingly dominate the technological costs of payment systems, setting a limit on the granularity of bundling and pricing. Prices don’t come for free.”
Over two decades of Moore's Law later, services and content on the internet are paid for by advertising, subscriptions, or a combination of the two. Business models that abstract the per-action cost away from the user, exactly as Szabo predicted.
Ben Thompson summarised the current state of things well in his piece a couple of years ago, The Unified Content Business Model. In it, he argues that the logical position that every content platform will adopt will be one of a mix of advertising and subscription, citing the New York Times, YouTube, and Netflix as examples. Music isn’t explicitly mentioned, however, Daniel Ek spoke to a podcast a few months later about the piece, saying, “Yeah, exactly. It is one of the areas that I'm the most intrigued about. I think Ben Thompson had this piece very recently. I think he called it the unified content business model piece. I don't necessarily agree with everything he said, but I think his main takeaway is obviously that all media models ought to move to freemium. He's someone who's been saying that for 15 years.”
Arguably, Ben Thompson’s most impactful observation on how business is done on the internet is what he calls Aggregation Theory. “Zero distribution costs. Zero marginal costs. Zero transactions. This is what the Internet enables, and it is completely transforming not just technology companies but companies in every single industry. Old moats are gone — and new ones can be built — and Aggregation Theory helps you identify both”. Through Aggregation Theory, Thompson explains how Aggregators (Google, Facebook, Amazon, etc.) have utilised asymmetric leverage against supply chains to balloon in size and power, creating the internet that we have all become accustomed to.
In its simplest form, disruption initially occurs when a new technology enables a sought-after equivalent service or product to be provided for cheaper than its current market offer from incumbents. Disruption sticks when the new cheaper proposition empowers a user to do something new that they couldn’t previously do, and this new thing becomes a key user requirement.
The discussion on AI x Music to date has been frustratingly shallow on this point. Focus has centred on genAI music as an increasingly credible, cheaper alternative to music underpinned by copyright. While that is the thing in front of the industry’s nose right now, lets game out what this could actually mean with regards to future business models for content.
Back to Ben Thompson, last year he out out a piece titled Meta’s AI Abundance in which he laid out the case for why Meta is best placed to succeed in the incoming AI era, at least on the first clear battleground, AI x Advertising. Meta emerged from App Tracking Transparency (ATT) regulatory changes with a reinforced position as a global leader in digital advertising (a classic case of regulatory action raising the capex threshold for competition), and owns the world’s content feed, which it is increasingly populating with genAI content:
“Meta is the best positioned to do that in the short-term, thanks to the obvious benefit of applying generative AI to advertising. Meta is already highly reliant on machine learning for its ads product: right now an advertiser can buy ads based on desired outcomes, whether that be an app install or a purchase, and leave everything else up to Meta; Meta will work across their vast troves of data in a way that is only possible using machine learning-derived algorithms to find the right targets for an ad and deliver exactly the business goals requested….at some point ads will be indistinguishable from content. You can already see the outlines of that given I’ve discussed both generative ads and generative content; they’re the same thing! That image that is personalized to you just might happen to include a sweater or a belt that Meta knows you probably want; simply click-to-buy. It’s not just generative content, though: AI can figure out what is in other content, including authentic photos and videos. Suddenly every item in that influencer photo can be labeled and linked — provided the supplier bought into the black box, of course — making not just every piece of generative AI a potential ad, but every piece of content period.”
More recently, in an interview with Stratechery this month, Mark Zuckerberg was very clear in that he sees the future of advertising as being a turnkey solution provided to brands and consumers: ”Yeah, or we just make it for them. I mean, obviously, it’ll always be the case that they can come with a suggestion or here’s the creative that they want, especially if they really want to dial it in. But in general, we’re going to get to a point where you’re a business, you come to us, you tell us what your objective is, you connect to your bank account, you don’t need any creative, you don’t need any targeting demographic, you don’t need any measurement, except to be able to read the results that we spit out. I think that’s going to be huge, I think it is a redefinition of the category of advertising.”
Advertising on the internet is set to become even more ubiquitous, even more targeted, and much cheaper. The supply chain underpinning serving ads, from creative through to SEO and profiling, will be compressed into Zero distribution costs. Zero marginal costs. Zero transactions., the next run of Aggregation Theory.
The final bit of set up that I want to reference to create the background for this piece, and I guess the one that prompted me to write at all after a bit of a break, is The Agentic Web and Original Sin, again, from Ben Thompson. Citing recent announcements from Microsoft concerning accessibility and standardisation of the “Open Agentic Web”, he argues that the future of an internet filled with AI agents, which many are now building toward, may well open up the possibility for micro-transactions to become the dominant business model. Unburdened by the psychological friction of pricing that we humans have imposed onto the web, agents instead will algorithmically find the path of least cost, even if that requires orders of magnitude more decisions to get there. What I take from this is that the biggest trap that we can fall in to when thinking about future content business models is adopting a skeuomorphic mindset, taking what works now, advertising + subscriptions, and simply adding bots to the equation. The rules that built the current, human-centred internet will not be the same in the future, human + agent-centred web.
Music
OK fine, but the business of music on the internet has evolved in a fundamentally different way from other content. For one simple reason - ads don’t work very well with audio in the premium experience of a continuous streaming environment. Adding ads into the flow creates a disjointed user experience, and is largely the reason why Spotify can’t follow the Netflix / Cable TV model, injecting ads into the subscription tier to increase profitability. Instead, in music, ads have evolved to be more in line with the “rip the roof off third class” rather than use them to ”pay for champagne in first class” business model; intentionally making the cheaper/free tier a worse experience so that users are encouraged to upgrade their ticket.
And so, it follows that the current status quo of music streaming could well be relatively sheltered from the incoming disruption to the advertising industry highlighted above, when compared to other content platforms that are more reliant on ad$s, such as Twitch or online news sites, and at least in the short term.
However, when looking through a slightly longer lens, it is important to call out something else that Ben Thompson highlights in his Agentic Web and Original Sin essay, that there will inevitably be a content drought on which the algorithms of tomorrow can train. Thompson sees this as an opportunity on which a new marketplace can be developed whereby the AI itself is the primary customer for new content, aggregating it in to make adverts, affiliate marketing content, etc.. He cites this as a “possible - not probable”, but should the market move the way that evidence is currently pointing, then I feel it is should be categorised as economically inevitable.
Furthermore - and before making this point I want to state that the conversation of “is there even copyright involve in training data” is boring and is being parroted by boring people, because, of course there is, and of course regimes will settle that way eventually, and inevitably there will be a big pay-out for all the world’s content created to date…and then everyone will start looking forwards rather than backwards - this takes the “how should future content be licensed to AI” debate to a different place. Buy-outs.
Because, with the introduction of a new class of primary customers (AI), between the creator and distribution platform, and then the consumer, we are likely to see a reset in the power balance. Currently, music streaming platforms require all music to be competitive, and as a result, there is collective strength in the negotiating position of the supply chain. An AI model will want as much music as it can get, but importantly, it doesn’t need it. As the primary customer and aggregator, and unburdened by any kind of transaction costs, it can act as the largest aggregators of the past 20 years have done so, and force fragmented creators to its own terms. Which will absolutely require full copyright control in the form of a buy-out.
But why, as a creator in the future, engage in this system? The incentives would have to be right, and there would have to be no credible alternative.
I said earlier that the music streaming subscription business is probably relatively sheltered from the earliest disruptive forces of genAI, which seem set to impact content businesses overly reliant on ad$s. However, the economics of music streaming, already very head & tail-y, is fundamentally schisming into a two-tier ecosystem as the payout models evolve in response to being DDOS’d by noise. Arguably, there are already very few ways to meaningfully monetise music recordings online for all but the very select few. The proposition of a marketplace in which you can sell recordings to an AI model is likely quite compelling, if only as the next generation of work for hire. Work for hire, the dependable revenue stream that has supported artists for as long as there have been artists.
Competing with free / will free remain free / the new free
The Open Web is funded through ad$s, if you don’t pay for it, then you are the product etc, etc. With the digital advertising industry facing imminent disruption, it follows that the Open Web may well not remain as open and free as it currently is.
Ever since the average user could sail the high seas, the music industry has been in competition with free. Twenty years of litigation and picking winners has brought us to the place we are in now, a forced worse free to promote the value of premium. So it’s kind of ironic that it may actually be the rise of AI that does more to shut down free as an option than anything else beforehand, by significantly diminishing the revenue stream that a free site can generate.
But in place of the current free comes the new free, the genAI content feed; Meta, X, Google, OpenAI. How well does premium subscription stand up as a proposition then?