This morning I received an email from Google outlining what they describe as the three key agency plays for the AI era.
The message was clear and familiar: agencies need to move faster, embed AI more deeply, shift towards value-based pricing, and prove their worth through stronger measurement and efficiency.
None of that is controversial. In fact, most agencies we speak to are already having these conversations. It’s a neat summary of where the industry is heading - and, unsurprisingly, it places Google’s tools right at the centre of that future.
To be fair, Google dominates this space. Most of us don’t really have the option of opting out, and these tools are powerful. But as I read the email, one question kept coming back to me:
What happens when the measurement layer we’re being asked to build our future on is no longer complete?
The assumption beneath the AI narrative
A lot of the AI conversation in agency land rests on a quiet assumption:
That we can accurately measure behaviour, confidently prove ROI, and therefore move cleanly from time-based pricing to value-based pricing.
In theory, that makes perfect sense.
In practice - particularly in the UK and EU - it’s far more complicated. Since the introduction of GDPR and UK privacy law, a significant amount of what we used to “just know” about users has disappeared. Analytics platforms like GA4 are now fundamentally shaped by consent. A meaningful proportion of users never opt into marketing cookies. Others never interact with a banner at all.
As a result, large parts of real user behaviour are no longer directly observed.
This isn’t a tracking setup issue.
It isn’t an agency capability issue.
It’s the reality of operating on a more privacy-conscious web.
The data we lost - and how it’s being replaced
Before GDPR, we had a much clearer view of:
- where people came from
- how they moved through a site
- whether they’d visited before
- what influenced their decision to act
Today, much of that insight depends on two things:
- People’s goodwill in opting into marketing cookies
- Platform-led modelling designed to fill in the gaps when they don’t
Tools like GA4, Consent Mode, and enhanced conversions don’t recreate lost data. They estimate it. And those estimates can be genuinely useful - particularly for trend analysis, optimisation, and directional decision-making. But they are not the same as observed truth.
Why this matters more in an AI-led world
AI systems optimise based on the data they can see.
When that data is partial, modelled, and legally incomplete, we need to be clear about what we’re proving versus what we’re inferring.
Google’s guidance quite rightly focuses on efficiency, automation, and scale. But what’s often left unsaid is that as we push harder towards value-based pricing, agencies are also quietly absorbing more uncertainty in the measurement layer underneath.
The risk here isn’t technical.
It’s relational.
If agencies oversell certainty - confident dashboards, definitive percentages, clean ROI claims - we create problems further down the line when reported performance doesn’t align neatly with commercial reality.
Trust is much harder to rebuild than it is to protect.
GA4 isn’t useless - but it isn’t the system of record either
This isn’t an argument against Google’s tools. We use GA4 every day. In many cases, it’s unavoidable, and it’s often the best option available.
But in GDPR markets, it needs to be treated honestly. GA4 is:
- Directional, not definitive
- One signal among many, not the single source of truth
- Strong for trends and optimisation, weaker for absolute claims of value
Pretending otherwise doesn’t serve clients, and it doesn’t ultimately serve agencies either.
What this means for clients - and why we’re open about it
We’re not raising this to lower the bar or reduce accountability. Quite the opposite.
Being a true partner today means being clear about the constraints we’re all operating under. It means explaining where data is solid, where it’s modelled, and where professional judgement still matters.
Marketing hasn’t become less effective - but measurement has become less complete.
Our role as an agency is no longer just to report numbers. It’s to interpret them responsibly, combine them with commercial context, and help clients make good decisions without hiding behind false precision.
That’s the trade-off of a privacy-first web. And navigating that trade-off honestly is now part of the job.
A more mature definition of “value”
If the AI era is genuinely about maturity - not just speed - then value can’t live solely inside platform metrics.
It has to be anchored in a broader picture:
- lead quality, not just volume
- pipeline health, not just conversions
- operational efficiency, not just ROAS
- commercial outcomes, not just dashboards
It also means designing pricing and engagement models that reflect confidence levels, not just aspiration.
This isn’t about being pessimistic. It’s about being accurate.
Maybe I’ve got this wrong
It’s possible the right answer is simply to lean in harder. To trust the models, adopt faster, and let AI smooth out the rough edges over time.
Agencies - including ours - can always improve how we use first-party data, educate clients, and connect marketing activity to commercial results.
But even then, the underlying reality remains.
AI doesn’t remove uncertainty from measurement.
It abstracts it.
The real advantage going forward
AI will reshape agency models. Tools will improve. Automation will accelerate. That’s inevitable.
But I’m increasingly convinced the agencies that thrive won’t be the ones who adopt AI the fastest. They’ll be the ones who are most honest about its limits - especially when it comes to data and measurement.
In a market full of confident dashboards and definitive percentages, measured honesty might turn out to be the most valuable differentiator of all.
If you’re looking for an agency that won’t oversell certainty - and will help you make good decisions even when the data is imperfect - that’s exactly how we work. Let's have a chat.
