Disintermediation Does Not Remove the Work

I teach a case in an Executive MBA program that asks the classic disintermediation question about the music business: which players disappeared, and who took their place. Every year the groups come back with a clean list of casualties. And every year the most interesting correction comes from checking whether those players actually vanished, because most of them did something else. They lost their volume, their income moved to a different payer, and the work they used to do turned up somewhere else in the system.

What disintermediation actually moves

When an industry gets rewritten, the first question in the room is which players are about to disappear. I have sat in that conversation in retail, in insurance, in sports and in medical devices, and it is the wrong place to start, because it confuses the company with the function.

A function is something the system needs doing. Somebody has to put the product in front of a buyer. Somebody has to finance its production, decide what gets promoted, carry the inventory risk, hold the relationship with the customer. A company is a temporary container for a bundle of those functions, held together by whatever made that bundle efficient at the time. When the conditions change, the container comes apart, and every function it was holding has to be picked up by somebody.

So the question worth asking is where each function went, and it has three parts: who performs it now, at what scale, and who pays for it. The third part is the one almost nobody checks, and it is usually where the new business model sits. A function can survive intact while the client who pays for it changes sides.

How the container came apart

The collapse of recorded music usually gets told as a morality tale about piracy. As a mechanism it is far more useful, because each step removed one specific piece of friction, and once enough of it was gone the old bundle stopped paying for itself.

Compression made the file small enough to move. An album that lived on a disc became a handful of files that fit through a dial-up connection, and from that moment the physical object was one way of owning music rather than the only one.

Napster added the directory. A single index telling you who had what, which happens to be the function a wholesaler performs for a record shop. The industry sued, and the service went dark in 2001.

Then came the hydra. Kazaa, eDonkey, BitTorrent, each generation pushing the index further away from any single server, until shutting down the center stopped shutting down the network. This is the part that tends to get skipped in the retelling: the lawsuit succeeded and the problem grew, because by then the function had been copied into software that no company owned.

In Spain the physical channel took a parallel hit. Street vendors sold burned discs for a couple of euros on the same shopping streets as the record shops, and home copying did the rest. For a while the shop was competing against its own product at a tenth of the price.

The industry’s response pushed in several directions at once, and most of them moved the customer further away: lawsuits against individual users in the United States between 2003 and 2008, copy protection that stopped legitimate buyers from playing what they had paid for, and the Sony BMG rootkit in 2005, which installed hidden software on the buyer’s own computer. I spent six years inside this business, building the B2B and B2C e-commerce operation of the largest music wholesaler in Spain, and what I remember is not denial. Everybody could see the ground moving. Nobody had a model ready to replace the one that was going.

The replacement came from outside the industry. Apple opened the iTunes Store in 2003, a hardware company solving a distribution problem the labels had not managed to solve, and it worked because it matched a decision the listener had already made. What people wanted was access to the music. Owning a copy had been the means, and the means had stopped mattering.

Spain’s recorded music market shows the size of the fall. It peaked at 626 million euros in 2001, and by 2014 it had lost 76% of that, according to Promusicae’s own market report. Three quarters of an industry gone in thirteen years.

Meanwhile, recording got cheap

While the distribution side was coming apart, a second shift was running in parallel on the production side, and it gets far less attention. Alesis started shipping the ADAT in February 1992, eight tracks of digital recording onto cheap S-VHS tape for around $3,995, which put professional quality inside rooms that could never have afforded a studio. In 1999 Digidesign released Pro Tools LE with the Digi 001 interface, running on the computer’s own processor instead of dedicated cards, and soon after the Mbox did the same job over USB for around $495. The recording studio, historically owned or booked by the labels and the hardest link in the chain to go around, became optional in roughly a decade.

That movement is worth separating from the others. Nobody replaced the studio. The capability itself moved, from a company that rented it by the hour to the artist who now owns it. The caveat matters as much as the fact, because capability and leverage are different things. An artist today can record, release and sell directly, and still needs the investment, the radio and playlist weight and the international machinery that a major supplies, which is why most careers that reach real scale still go through one. What moved to the artist is the capability, and with it the cost and the risk. The bargaining power went to whoever controls access to the audience, which today is the platform.

A map of who took over what

Once you stop counting companies and start counting functions, the picture gets specific. The table below maps the music business by function instead of by player, and it asks four questions of each one: what the function is, who owned it in the 1990s, who owns it today, and who pays for it now. I call it The Function Reassignment Map©, and the fourth column is the one that changes decisions, because a function that survives with a different payer is a different business even when the work looks identical.

Read down the third column and the pattern is hard to miss. One actor absorbed, in whole or in part, four functions that used to belong to four different companies: aggregating catalog, forecasting demand, selling to the fan and controlling access to the fan. That concentration is the actual event of the last twenty years in this industry, and it is a better description of what happened than any count of who went out of business.

Read down the fourth column and something else shows up. Several functions survived with the bill moved to a different party. Financing a recording used to be the label’s risk and now falls on the artist for most of what gets released, though the labels still fund the recordings they sign. Promotion used to be paid for in media budgets and is now paid for in content and time. None of that changed the work itself. What changed is who receives the invoice for it.

The function that shrank instead of dying

The wholesaler is the clearest case, and the one I know from the inside. Before the collapse it did three things for a record shop. It was the single supplier of assortment, because a shop could not buy from thirty labels one by one, and most labels would not have taken the order anyway. It knew the media plan for every release, so it could estimate demand and its peaks before the record came out. And its sales force told each shop what to order and in what quantity, which is a forecasting service delivered by a human being with a route and a catalog.

When the market fell to a quarter of its size, that structure stopped being payable. A warehouse, a fleet, a sales force, credit lines and a returns policy are fixed costs sized for hundreds of shops, and the hypermarkets that had become the volume channel started buying straight from the labels. The large wholesaler did not lose an argument about its usefulness. It lost the volume that paid for its overhead.

The function itself stayed in business at a smaller size. The shops that are still open in Spain buy from specialized distributors, Popstock and Everlasting among them, working with catalogs like XL, 4AD, Matador, Rough Trade and Sub Pop, from Discmedi and Blanco y Negro, which also places product in El Corte Inglés, Fnac and MediaMarkt, from small regional wholesalers, and from European one-stops like Bertus for imports, with around 400,000 references on a B2B platform. A good part of the shelf never passes through a distributor at all, because it is second hand bought from the public, with Discogs acting as the global catalog that makes pricing possible.

The scale is worth stating plainly. The whole Spanish physical market was 41.7 million euros in 2025, around 12% of recorded music revenue, against 302 million in digital, with vinyl at 2.18 million units, according to Promusicae. That is a real business with real logistics, and it is a fraction of what it was. Where the market is big enough, the same function still operates at industrial scale: Alliance Entertainment, the US one-stop, reported $1.149 billion in revenue in its 2026 fiscal year, with $383 million in vinyl and $156 million in CD.

The part that took me longest to see is who pays. The classic wholesaler bought and resold: its customer was the shop and its income was a margin. Today’s distributor charges the label for placing its product in shops and on platforms, and sells unbundled services on top. La Cúpula Music, a Spanish distributor, publishes its rates: registration, ISRC and barcode cost nothing, delivering a single track to every platform costs 2.99 euros, and the business sits in the catalog of services around that, from manufacturing and mastering to artwork and promotion campaigns. None of these companies lives on aggregating assortment alone.

So the honest version of the famous disintermediation story is narrower than it sounds. One company size became unviable, one customer was replaced by another, and the work got split between a platform that absorbed it for free as a side effect of its own operation and a set of small specialists who now invoice the label instead of the shop. When an intermediary loses its volume, the ones that survive tend to survive by changing clients.

The functions nobody picked up

Not every function found a new owner in the same shape. The clearest case is forecasting for the physical shop. The wholesaler delivered it as part of the service, reading the media plan and advising on quantities, and no single actor does that integrated job today. Distributors and sales reps still advise, pre-order and sales data exist, and the shop assembles the pieces itself. The platform, meanwhile, generates better demand data than anyone ever had and keeps it, because it is a byproduct of its own operation rather than a product it sells. And streaming data does not translate cleanly into physical sales: a track with millions of plays can sell very few records.

Two more functions ended up only half claimed. Long-term artist development, the willingness to fund three records before one works, used to sit with the label as a portfolio bet; it survives in a few places and has largely moved onto the artist, who rarely has the balance sheet for it. And editorial filtering with something at stake: a radio programmer or a music journalist put their own reputation behind a recommendation, which is a different mechanism from a system optimizing for retention. Playlist editors do some of this. Creators do some of it. No actor owns it the way the gatekeepers did.

Meanwhile, a short list of things did not move at all. Somebody still has to write a song worth hearing. Somebody still has to play it in a room. Live performance went from being the promotional expense that supported record sales to being the main income for most working artists, which is an inversion of the whole model and the reason the economics of a tour deserve their own analysis.

That list of unclaimed functions is the most practical output of the whole exercise. A function that the system needs, that nobody currently owns, and that somebody is implicitly paying for in inefficiency is the shape of an opening. It is where the next intermediary comes from, and it is usually visible years before anyone builds it.

The same pattern, now with AI

Every wave so far hit the functions around the music: how it was stored, how it moved, how it was discovered, how it was paid for. Generative models reach a function that had stayed untouched, which is making the thing in the first place. That is a difference in kind, and it is why the industry’s reaction in 2025 and 2026 looks so much like 2001 at first glance and so different underneath.

The observable signals are worth separating from the speculation. Major labels sued the main generative music services and then moved toward licensing agreements with them, which is the same sue-then-license sequence that ended the Napster era.

Deezer, the one platform that publishes the figure, reported in July 2026 that fully AI-generated tracks had passed half of its daily uploads, around 90,000 a day, up from roughly 10,000 in January 2025. That music accounts for between 1% and 3% of actual streams, and up to 85% of those streams were identified as fraudulent in 2025. Spotify does not disclose an equivalent number, so one platform is not the market, but the gap between how much gets produced and how little gets heard tells you the bottleneck moved from production to attention. And a claim worth labeling as a claim: the accusation that a platform fills playlists with cheaper commissioned material has been published and denied, so it belongs in the analysis as a contested point rather than a fact.

The structural question the map raises is more interesting than any of those headlines. An actor that already owns discovery, the relationship with the fan and the payment rail has very little distance left to travel before it can own the catalog too. Tencent Music has been doing exactly that for two years. It generates songs with AI inside its own apps, has credited its own content for part of its gross margin improvement, and its CEO told investors in August 2026 that the effort is producing very good commercial returns. If that happens the digital chain goes from five actors to two, and the function that gets reassigned is the one everybody assumed was safe.

I am not going to tell you where that lands, because the useful part is the method rather than the forecast. Take the functions in your own operation, ask which of them a model can now perform at acceptable quality, then ask the two questions that follow: at what scale, and who ends up paying. The answers tend to look less like elimination and more like reassignment, with the bill moving to whoever has the least negotiating power.

Four questions for your own sector

1. List the functions, not the companies. What has to happen, step by step, for a customer in your industry to end up holding the product?
2. For each function, who performs it today, and at what scale compared to ten years ago?
3. Who pays for each one, and has the payer changed? A function with a new payer is a new business.
4. Which functions currently have no clear owner? That is your shortlist, and the system is already paying for their absence somewhere.

The fourth question is the one that produces decisions. The first three describe a market, and most teams can answer them in an afternoon once they stop arguing about which brands will survive. Disruption rarely deletes work. It moves it, concentrates it in whoever controls access to the customer or can perform it at a scale nobody else matches, and leaves a few pieces on the floor for somebody to pick up.

I keep teaching this case because the correction is always the same. The groups hand in a list of companies that died, and the conversation only gets useful when we rebuild it as a list of functions and ask where each one went. The wholesaler I worked for is gone. Almost everything it used to do is still being done by somebody, somewhere, paid for by someone else.

---

© 2026 Oriol Guitart. The Function Reassignment Map© has been developed by its author, Oriol Guitart. All rights reserved, for the duration and extent established by the applicable intellectual property law. Any reproduction, distribution, public communication and/or transformation, in whole or in part, is strictly prohibited without the author’s express written authorization, and the author must in any case be credited as such in any subsequent use.