Expertise and efficiency stopped being a trade-off.

Marketing services have always priced expertise and volume against each other. The senior operator who understood your category was expensive and slow, because their time was the product. The agency that could produce 50 assets a month was affordable, because the people making them were junior and the process was industrial.

Most companies picked one and lived with the cost. You either got advice you trusted and not enough output, or output you could schedule and had to keep correcting.

What actually changed

The thing that forced that trade-off was production time. Writing, designing, formatting, and versioning consumed most of the hours in any engagement, which meant senior judgment could only ever be applied to a fraction of the work.

AI collapsed that part. Drafting, restructuring, adapting one argument across six formats, producing the fifth variant of a headline — that is now hours instead of weeks. Which means the constraint moved. It is no longer how fast you can produce. It is whether what you are producing is right.

AI compresses the execution. It does not compress knowing what to execute.

Why speed makes domain knowledge matter more

This is the part that gets backwards in most conversations about AI and marketing. Faster production does not reduce the need for expertise. It raises the cost of not having it, because a wrong strategy now executes at volume before anyone catches it. The visible symptom is content that all sounds the same — competent, publishable, and interchangeable.

In security, risk, and resilience this is not an abstract risk. A studio that does not know the difference between a CISO and a Chief Risk Officer, or why a claim about detection rates invites a procurement question, or what a resilience buyer has to prove to a regulator, will produce plausible material that quietly disqualifies you. It reads fine. It fails in the review.

Where the old studio model breaks

Traditional agencies were built around the production constraint. The pyramid staffing, the account layer, the billable hour, the discovery phase that exists partly to load context the team does not have — all of it made sense when execution was the expensive part. That structure is also why the agency and fractional comparison comes out differently than it used to.

Bolting AI onto that structure produces the same work faster, at the same margin, with the same context problem. The output improves marginally. The thing that determined whether the work was any good — whether the people doing it understood the category — is untouched.

• Discovery is a symptom. A team that needs 8 weeks to understand your market will still be guessing in week 9.
• Layers cost accuracy. Every handoff between the person who understood the conversation and the person producing the asset loses something.
• Throughput without judgment is a liability, not a feature. Volume amplifies whatever direction you set.
• Speed you cannot review is not speed. Work that has to go back 3 times was not faster.

What the model looks like instead

If production is no longer the constraint, the studio should be organized around knowing the problem set before the engagement starts. That means senior operators doing the actual work rather than supervising it, and it means specializing narrowly enough that the context is already loaded.

The practical difference is what happens in week 1. A studio that already knows your category starts with a position and an argument. A studio that does not starts with a questionnaire.

In practice that has looked like naming and launching a category with Fusion Risk Management, building the market story for iluminr microsimulations — whose customer AXA Ireland went on to win Most Original Exercise Programme at the BCI Europe Awards 2026 — and helping resilience teams explain their programs internally.

This is the reason we built Eudai around a narrow set of categories rather than a broad service list. Not because breadth is bad, but because the advantage AI creates only compounds if the judgment is already there. Efficiency applied to the wrong direction just gets you there sooner.

What this does to price

The obvious objection is that senior operators are expensive. Compared with an hourly rate for junior production, they are. That comparison is also the wrong one, because it prices the input rather than the outcome.

What an engagement actually costs is the rate plus the discovery you fund before anyone is useful, plus the correction cycles when the work misreads the buyer, plus the delay while all of that happens. Removing the production constraint does not lower a senior rate. It lowers the number of hours that rate has to be paid for, and a specialist removes the discovery line entirely.

So the honest framing is not cheaper or more expensive. It is a different purchase: fewer hours of a more expensive person, with less rework, instead of many hours of a less expensive team plus the cost of getting it wrong first.

How to evaluate a studio now

Asking whether a studio uses AI is no longer a useful question. Everyone does, and the ones who say they do not are usually not being honest about it. Better questions:

• What do you already know about my buyer that I did not tell you?
• Who is doing the work, and have they operated in this category or only marketed to it?
• Show me something you decided not to do, and why.
• What would you refuse to do on our behalf?

The answers tell you whether you are buying judgment with efficient production attached, or efficient production with judgment quoted separately. The same test applies to execution: what actually works in outbound is mostly knowing what to say, not how fast you can send it.

The trade-off is genuinely gone. But it only disappears for teams that already know the problem, because expertise is the one input that cannot be generated on demand.

Part of our work as a fractional CMO for security & resilience.