DESIGN INTELLIGENCEJuly 24, 2026·Mary · DEPIX Design Intelligence

Execution Went In-House. Authorship Didn't.

The design world loves a turf war, and the current one has three sides: the in-house team, the outside agency, and the AI. The framing is always the same - who will do the work? It is the wrong question, and it hides the only change that actually matters.

Start with the facts, because they are not subtle. Over 64% of brands now run some form of in-house creative capability. What began as a small social team has, at many companies, grown into a full studio producing films, campaigns and always-on content at a speed external shops struggle to match. The obvious read is that in-house is winning and the agency is dying. The truer read is quieter: in-house teams have captured the execution layer, and pushed everyone else to justify their existence somewhere higher up.

Now add AI, and the same thing happens one level down. The cheap, fast, high-volume production of assets - the exact work in-house teams just absorbed - is the work a model does best. AI systems are already being sold as an extension of the in-house team rather than an outside vendor, capturing a brand's rules and spinning out on-brand assets on demand. So execution is being commoditised twice over: pulled in-house to go faster, then handed to a machine to go faster still. If your value was doing the work, that value is heading toward zero.

Here is the part the turf war misses. Every one of these shifts moves execution around - and leaves authorship exactly where it was. The market is not splitting into in-house versus agency. It is splitting into generic versus authored: generic work trending to free, human-led and taste-driven work going premium. The agencies that survive are the ones that stopped selling production and started selling the thinking layer - the positioning, the brand platform, the one authored idea an internal team or a model can then execute against. Ownership of the vision, not ownership of the pixels.

And this is not a soft distinction any more; the law is starting to draw the same line. Work generated by AI with no meaningful human authorship is not protected the way human-made work is. Courts increasingly look for significant human intervention - direction, editing, recomposing, the visible fingerprints of a decision-maker. Which produces a strange, clarifying result: the authored concept is now the part of the work you can actually own, and the executed output, if nobody authored it, is the part you cannot. Authorship stopped being a matter of pride and became a matter of title.

For anyone deciding how to build a design function, this reframes the whole question. Do not ask whether to staff up in-house, retain an agency, or buy an AI tool - you will end up doing all three, because they are three ways of buying execution, and execution is now cheap. Ask instead: where does the authored vision live, and who is accountable for it? That role - the person or small group holding the single point of view that everything else executes against - is the only one that does not commoditise, because it is the only one doing the concept-phase work: deciding what the thing is, what it must feel like, and what it refuses to be.

The in-house studio, the agency and the AI are, in the end, all the same kind of thing: extremely capable hands. Hands have never been the scarce part. A company can now assemble more execution capacity than it has ideas worth executing, which is precisely the failure state to avoid - a hundred perfectly-produced assets and no conviction underneath any of them. The teams that win the next decade will not be the ones with the most hands. They will be the ones who kept the authorship in the room, gave it a name, and refused to let it drift out to whoever happened to be doing the work that week.

Execution went in-house. Then it went to the machine. Authorship is still sitting where it always was - in the concept phase, waiting for someone to own it on purpose.

Sources: