David Sugg has spent 26 years building and operating media supply chains at major studios. He's led large technology programs, managed teams of hundreds, and learned most of what matters by being in the room when things went wrong. He now consults on technology implementations, supply chain strategy, and operations, with a particular focus on how technology actually performs in production environments.
Every large media company I have any visibility into has more supply chains than it wants, and most are trying to consolidate. The instinct is sound and the execution usually stalls, because consolidation gets treated as the goal instead of as a way to get control. Fixing the middle is very hard, not because building one pipeline is hard but because consolidating the deliverables those pipelines produce takes a very long time. You are better off normalizing the two ends, how content comes in and how it gets ordered. Those are not the front and back of a pipeline. They are the two doors through which supply and demand enter, and the people-heavy work, translating demand into orders, chasing materials, handling exceptions, happens there. My Human Layer post from May argued that the industry has used people as the integration layer between its systems for decades. The two doors are where that integration lives. Normalize them and the control and the savings follow, and the savings buy the time to fix the middle as you can.
Andy’s recent Deep Cut, “Everything Is Replaceable,” made the architecture version of this argument: a system can be modular and still hard to change, because the switching cost sits in everything attached to the components, so that “every piece was somebody’s, and the sum was nobody’s.” What follows is the media supply chain version, from the side I know, building the tools operations runs on.
Consolidation is not the goal. Own intake and ordering, and let the middle change in the time that buys you.
What a studio supply chain looks like
At major studios the media supply chain is rarely one supply chain. It is several, sometimes a dozen or more, each grown out of an acquisition or a business line, each full stack from order management through asset management to distribution. Broadcast and non-broadcast is the simplest case. Mergers merge some of the org chart and rarely the supply chains, so the count only goes up.
Those supply chains overlap in one important place, the recipients, and the overlap is in name only. Amazon gives the Southeast Asia supply chain one delivery spec, the Central European supply chain another, and the North American one a third. Streaming and broadcast arms of the same company are on different specs, rightly so. A few thousand recipient companies become tens of thousands of recipient profiles, each a valid delivery to the same company from a different supply chain and a different starting point.
Consolidating a supply chain’s middle means replacing every processing component, and every one of those recipient profiles is wired to one of them. Each took longer to onboard than almost anything else the supply chain does. Replace one component and you pay for everything wired to it, and replace the middle of a dozen supply chains and you pay for tens of thousands of profiles, each owned by somebody and the total owned by nobody. The recipient profiles are the wiring.
The obvious version of the consolidation argument is “I deliver to NBC, Amazon and Netflix, therefore I should make one file per recipient and be done.” It treats consolidation as the end in itself, and it fails because the merge is larger than the thing it was supposed to enable, and nobody sees a benefit until it finishes.
Normalize intake and ordering first
If the middle cannot be consolidated first, what can? Every one of those supply chains runs the same process: accept demand, understand it, turn it into orders and tasks, receive the materials, store them, transform them, deliver them. A dozen supply chains is a dozen copies of that process. The approach I drove, across over a dozen of these supply chains, was to normalize the two ends, intake and ordering, and leave the middles alone. It went in phases. Intake first, because one front door is the easier case to make. Then ordering, which is the less obvious side, because the instinct when consolidating distribution is to put everything through one transcoding platform, and the thing to make single is the way content is asked for. The project hadn’t finished when I moved on, and it was never meant to finish in one pass. Doing it in passes is what buys the time.
Content planning and intake become one front door. Materials come into the company one way, to one spec, into one central library, whichever supply chain will process them later. Most of the reasons content used to come in differently went away years ago. The ones that remain, rights, territory, security, timing, are exceptions to handle at the one door, not reasons for a second. One door also fixes a long-standing failure: content that enters a regional supply chain through its own door and never gets back to the center. From there the regional supply chains pull what the library announces, get pushed what they will need, or ask for something and the library supplies it or raises the order to acquire it. That last case is what lets them stop acquiring anything on their own.
Ordering becomes one place to ask. Every request for content, from every source, comes through it, including the messy ones. Everyone knows content goes to Netflix, Amazon, the licensees and the broadcast partner. What most supply chains forget is the editor who needs the master to cut a clip for social. Those ad hoc requests are many, and a supply chain that says “if you want content you order it” and then ignores them has built its own shadow paths.
The central order system knows the Southeast Asia supply chain owns the Southeast Asia Amazon recipient, so an order for that recipient is placed centrally and routed to the regional supply chain that already has it onboarded. The dozen supply chains keep their recipients and their processing and stop doing their own intake and ordering. Over time you take the one-off paths out of each supply chain, one use case at a time, which is the pattern software people call a strangler fig: the new path grows around the old one until the old one can be retired.
Normalize how content comes in and how it is ordered. Route everything else to the supply chain that already does it.
The middle becomes processing
Say you get three quarters of the way through that. Intake is one system. Ordering is one system, with everything that hangs behind it. And there are a dozen processing systems whose job is to turn content into specialized delivery profiles. Who cares?
The work people touch, demand, orders, metadata, content management, task management, now sits on the normalized ends. The work people do is simpler, and the savings management is asking for are there. The middle is transformation, and once the people-heavy work is out of it, keeping a dozen transformation systems alive costs far less than forcing them together, licensing and scarce expertise included. Software people would call this consolidating the control plane without consolidating the execution plane. On-premises or cloud, build or buy, becomes a calm conversation, because none of the people-heavy work depends on how the middle is built. The tens of thousands of recipient profiles stay until Amazon or Netflix normalize their own specs, on their own schedule, and carrying them until then costs little.
You did not consolidate the middle. You made it so the middle could stay as it is without holding anything else back.
The standard belongs at the edges
Someone will ask whether an industry standard lets you skip all this. It does not. No standard or SDK will rebuild a middle that already exists, several times over. What a standard can do is define the language everything outside uses to talk to the supply chain: how a title is requested, how materials are requested, how materials and the data about them are submitted. Intake and ordering are the two edges every supply chain has, which is why this piece is about them, and the same rule holds wherever something outside connects. That argument is the next piece.
A standard will not rebuild the middle, but it can define the language everything uses to reach it.
What you control
With intake and ordering normalized and everything reaching them in one language, you know what came in, what was ordered, where it went and what it took, and that is the operating record. It is the one part of the supply chain you never have to reconstruct, because it is written as the work happens. The last two pieces argued that the controls on a model have to sit on your side, because the vendor will not sign for them. They get written down at intake and ordering, and once they are there the middle can change, or not, on its own schedule.
AI agents multiply the importance of that operating record. The Human Layer post said what agents need before they can do useful work: demand that is legible, task state that lives outside someone’s head, orchestration that runs without the person who understands it in the room. Normalized intake and ordering are that foundation. Point an agent at the middle before they exist and it speeds up the duplicate work, a dozen supply chains still taking their own orders, faster. The number of decisions made per day is about to go up by an order of magnitude, and the record those decisions run against had better be yours.
Own the ends
Which brings the series back to where it started. Much of what is sold today as an AI platform or an orchestration platform does one of two things. It consolidates the middle, a common transformation layer, which is the wrong place to start and the part that least needs it. Or it sells you the whole thing, built on the vendor’s reference model of how a media company works, with your recipients and your demand fitted in afterward. Either way, the first written version of how your supply chain works ends up inside a vendor’s product. That operating model has lived in people’s heads and runbooks, and the platform is where someone else writes it down for the first time. Most companies do not notice it was theirs until they try to leave.
Let the middle be whatever the economics say, built or bought. You can buy intake software and you can buy an order system. What you cannot buy is what they mean in your company, the demand, the policies and the record, and that is the part nobody can sell you.
A foundation model is a supplier. The model your vendor runs on your content is a supplier inside your supplier. The layer that decides what both of them are asked to do is the one thing in the stack that has to be yours.
Sources and further reading:
Andy Beach, Everything Is Replaceable (Sept 20, 2026); Four Ways to Reach the Model (Aug 21, 2026).
David Sugg, The Model Is a Supplier (Sept 13, 2026); The Supplier Inside Your Supplier (Sept 18, 2026); When the Human Layer Is the Control Layer (May 23, 2026).
AWS, How 5 migrated to AWS Elemental MediaPackage v2 (June 29, 2026), the Channel 5 case Andy discusses. On the switching cost framing Andy takes from AWS: Mark Schwartz, Switching Costs and Lock-In (Dec 29, 2018), and the AWS whitepaper Unpicking Vendor Lock-in.
On standards at the edges. The nearest existing specs, none of which covers the whole of what the piece describes: MovieLabs, Asset Ordering, Delivery and Tracking (TR-META-AOD), asset orders, availability and delivery status between studios, platforms and service providers; MovieLabs, Ontology for Media Creation, a shared vocabulary that describes itself as “not a schema”; MovieLabs, Interoperability in Media Creation (2023), which notes that “in many cases, however, no standards exist for such APIs”; SMPTE, ST 2021 Broadcast Exchange Format (BXF), request and reply messages for content transfer, dub and purge between broadcast traffic, automation and asset management systems; EBU and AMWA, FIMS, service interfaces for capture, transfer, transform and repository jobs, last released 2017; and, from the music industry, DDEX, the closest adjacent example of a supply chain message standard.
