Something has changed this year, and it’s worth naming directly.
For most of our industry’s history, the scarce thing was skill. Good design, good development, and good content took time to learn and were genuinely hard to find. That scarcity is disappearing fast. AI can generate a website plan, a content structure, a brief, and a working prototype in minutes. The barrier to building something has nearly vanished.
We’d suggest that doesn’t make agencies less necessary. It makes us more necessary, just for a different reason than before.
The risk moved; it didn’t disappear
When it was hard to produce anything, the work was the production. Now that production costs less and is fast, the real risk shifts to a different point entirely: organisations committing to plans, structures and decisions that look complete but haven’t been tested against how they actually operate.
A brief can read as thorough and still be wrong. A prototype can look finished and still be built on assumptions that fall apart the moment real content, real audiences or real edge cases meet it. We’ve seen sitemaps that looked complete on the page but hadn’t accounted for an archive of hundreds of past productions, or ticketing logic that worked perfectly in the demo and broke the first time a real season went on sale.
The faster these things get produced, the easier it is to commit to the wrong one with total confidence.
The consequences of getting that wrong haven’t become cheaper just because the plan was generated faster. A website can be rebuilt. A content model can be reworked. A ticketing journey can be redesigned. But each correction costs time, money, momentum, and often confidence.
The irony is that as production gets easier, the cost of committing to the wrong direction often goes up, not down. More options can create the feeling of rigour without the substance of it. More speed can mean less scrutiny. That’s why we think the most valuable conversations increasingly happen before a project begins, while assumptions can still be challenged and decisions can still be changed.
What we think matters now
That’s the work we think matters most right now: not generating the plan, but knowing whether it’s the right one. Asking what a brief assumes, and whether those assumptions hold. Pressure-testing a structure against the complexity of a real programme before it’s built, not after. Taking responsibility for a recommendation. That includes the times when it means saying this isn’t the right approach, even when a faster yes would be easier to give.
None of that is anti-AI. We use these tools ourselves, and they’re genuinely useful for exploring options quickly. But generating an option and knowing it’s the right one are different skills. Only one of them carries any risk if you get it wrong.
The advantage is shifting, not disappearing
There’s a more hopeful way to read this shift, too.
If production is no longer the expensive part, the financial barrier that used to separate organisations with a budget from organisations with vision starts to thin out. A smaller venue or a leaner team can now explore, prototype, and test ideas that would once have required a far larger spend just to see whether they were worth pursuing.
That doesn’t mean budget stops mattering. It still buys time, depth and capacity. But it does mean the gap between well-resourced organisations and ambitious ones is no longer purely a question of who can afford to build. Increasingly, it’s a question of who can tell a good idea from a fragile one before committing to it.
That’s a meaningful shift. The organisations that benefit most won’t necessarily be the ones with the deepest pockets. They’ll be the ones who pair whatever resource they have with genuine judgement about what’s worth pursuing, and the discipline to test it properly before they commit.
Where that leaves agencies
We don’t think the agencies that matter in the next few years will be the ones that can produce the most, the fastest. We think they’ll be the ones who can look at something that was produced quickly, by a person, a tool, or both, and tell you honestly whether it will hold up.
That’s a different kind of value than the industry has historically sold. It’s also, we’d argue, a more durable one.