Most agency conversations about profitability start in the same place.
And now, inevitably, AI.
All of these are valid. They sit comfortably inside the familiar language of business management because they are measurable, trackable, and easy to surface on dashboards. They feel like control.
Yet many agencies doing exactly what they have been told is “the right thing” still feel permanently stretched, fragile under pressure, and quietly dissatisfied with the returns they generate.
That’s because margin is rarely the real issue.
Margin is an outcome, not a lever
Margins don’t disappear on their own, and they don’t improve simply because targets are raised or people work harder. They grow as judgment spreads throughout an organisation, and they erode as a consequence of accumulated design decisions that go unexamined.
You see this when senior people spend time where their judgment adds little value.
When teams are optimised for busyness rather than problem-solving. When pricing models remain anchored to hours in a world where time-based value is collapsing and when founders are still making decisions long after the organisation should be able to carry them.
Pulling harder on operational levers in these conditions rarely restores margin. More often, it increases frustration and exhaustion.
What is actually happening is that the organisation’s capacity to think, decide, and absorb complexity is being exceeded.
Capacity is not utilisation
Utilisation tells you how busy people are. It does not tell you whether the organisation is coping.
High utilisation can look healthy while the quality of judgment degrades unnoticed. Decisions are deferred or centralised. Trade-offs are avoided rather than made. Short-term fixes crowd out long-term learning, and the same problems reappear in different forms.
An agency can be efficient and still be overwhelmed.
Capacity is not about hours available. It is about how much decision load the system can carry without breaking trust, quality, or meaning. When organisations are designed primarily around predictability and throughput, humans are reduced to units of execution and judgement is slowly designed out.
AI hasn’t solved this - it is exposing it further
On the surface, AI should help. It accelerates delivery, reduces execution time, and lowers the cost of producing outputs.
What it does not do is remove the need for judgment.
In fact, it increases it.
Clients know AI exists. They know it compresses time. They can see that what once took eight hours now takes four, and they expect pricing to follow. If you don’t respond, a competitor will. This quickly becomes a race to the bottom rather than a rethink of value.
When time is no longer a reliable proxy for value, pricing becomes harder, not easier. Clients want transparency rather than mystery. Teams need clearer boundaries around where human judgement genuinely adds something different, and decisions about what really matters become more frequent and more visible.
Agencies that were already stretched feel this first, not because AI is the problem, but because it removes the buffer that time once provided.
Where margin really leaks
Margin erosion usually appears where the organisation’s capacity to apply judgment is weakest.
Senior thinking is consumed by pitching that teaches the organisation nothing.
Scope conversations are avoided because people lack the confidence to step outside the process and apply judgment. Founders slip from stewardship into micro-management under the banner of quality control, and teams are rewarded for activity rather than outcomes, learning, or ownership.
None of this is solved by better tracking. It is solved by designing how judgment is distributed, supported, and developed.
The uncomfortable shift agencies have to make
The agencies most likely to adapt are changing emphasis in subtle but important ways.
Counter-intuitively, this often requires slowing down rather than speeding up.
Slowing down long enough to understand where automation genuinely adds value, and where it erodes it. Where standardisation creates clarity, and where it destroys differentiation. Where value sits in the price a client is willing to pay, not the hours recorded, and where saying no protects future capacity rather than limiting growth.
This is not about reducing ambition. It is about protecting the very capacity that sustains ambition.
What this means in practice
Agencies do not need another framework. They need to ask different questions.
Margins improve when these questions are answered honestly. Not because costs are squeezed harder, but because value becomes clearer and easier to stand behind.
Most agencies do not fail because they are inefficient.
They struggle because they outpace their capacity to think.
Margin is the symptom.
Capacity - human judgment, deliberately designed and shared - is the work.
And that work cannot be automated.