For two decades, every pound a marketing team spent on software supported six pounds spent on the people running it. That ratio is breaking, and the commercial consequences are going to land first inside the marketing budget.
For every $1 a business spends on software, it spends roughly $6 on the people who deliver the service that software supports according to a recent article by Sequoia titled Services: The New Software.
That certainly applies to marketing. The CDP, the ESP, the marketing automation licences are the cheap part. The team running it, the agency operating it, the analyst building the reports, the contractor cleaning the data: that's where the budget is really spend. Software captured the pound, the operators captured the six.
For 20 years, the ratio has held. Agencies, consultancies and data teams sat comfortably on the £6 side. AI agents are now starting to absorb that £6 share and the implications for the buyer side are large enough to redraw how marketing budgets get built.
The shift the loudest voices in San Francisco are calling Service as a Software is simple in its proposition. The model sells the outcome, fully delivered, in place of the tools that help a human team deliver it.
The same agent that helped an analyst write SQL last year now writes the query, runs it, builds the segment, pushes it into Braze and ships the post-campaign read. The same model that helped a strategist sketch a customer journey now deploys the journey, plugged into the CRM, instrumented end to end. The output approaches what a competent team would have shipped. The price approaches the cost of software.
Where the model works, the price of the outcome compresses from services pricing toward software pricing. The question for any marketing leader reading this is which parts of their marketing operation sit on the wrong side of that shift.
Every marketing function has three layers of work. Each is being absorbed at a different speed.
Layer 1: Production. Hours-for-money execution. Campaign builds. Audience exports. Tag QA. Performance pulls. Creative variations. Reporting. List hygiene. The work most agencies bill by the hour and most in-house teams complain about. An agent performs this work directly, end to end. This layer is being absorbed first, fastest, and most visibly. A £15,000-a-month retainer covering this scope of work is facing a price ceiling that is collapsing in real time.
Layer 2: Pattern application. Translating a known marketing problem into a working answer using a playbook the discipline has refined for two decades. RFM segmentation. Lifecycle modelling. Channel mix shifts inside an MMM. Win-back logic. Replenishment triggers. The work is genuine. The shape of the answer is also known, which means an agent that has processed ten thousand of these can produce the eleven-thousandth in minutes. The labour cost of this layer is moving an order of magnitude.
Layer 3: Strategic direction. Picking which customer is worth fighting for. Naming what the brand actually stands for. Deciding when to stop optimising and rebuild. Reading a market and rejecting the wrong work. This layer is becoming more valuable as the cost of execution collapses because the cost of picking the wrong thing to execute scales in proportion.
A function paying for Layer 1 and 2 by the hour is now buying capability that agents deliver at a fraction of the cost. The functions gaining advantage are those that buy Layer 3 expertise alongside packaged 1+2 outcomes.
The shift is already commercially visible in our own work. We built a Single Customer View for a client in under 48 hours, with growth segments identified inside 72 hours. Traditional analytics priced the same scope as a three-month engagement. The £3.9m of revenue opportunity it surfaced was not new data. It was the same data, read at a speed that was not commercially possible eighteen months ago.
Three things, in order of how quickly they will hit the P&L.
First, the labour cost of marketing execution is repricing. Agencies and partners delivering Layers 1 and 2 by the hour will either match the new price and absorb the margin loss, or hold the price and lose the work. Some will pivot, most won't.
Second, the value of the people who direct the work goes up. Senior strategists, technical leads and marketing operators with a point of view all become more valuable. The job is shifting from doing the work to making sure the right work is being done, at the right depth, by an agent that has been correctly briefed.
Third, the procurement question changes. The right question now is: what outcome are we buying, on what timeline, for what fixed cost and who owns the model when the engagement ends. Day rates and open-ended retainers were built for a world in which the £6 share of labour was structurally unavoidable. That world is closing. Procurement teams are already writing outcome-based briefs in place of resource-based ones.
Marketing isn't the same category as tax returns or insurance compliance filings. Brand and customer relationships are trust-dependent in ways that a regulated services workflow isn't. The bull-case version of Service as a Software, in which an agent ships the campaign and a team is dismissed, overclaims for our category and will keep over-claiming for at least another two years.
McKinsey's State of AI 2025 makes the point: only 39% of organisations report any measurable EBIT impact from generative AI and for most of those the impact sits under 5%. Adoption is wide but scaled value is rare. The companies converting AI use into real numbers are the ones redesigning the work, not the ones layering AI on top of unchanged workflows.
Strategic direction, brand judgement and the messy human work of deciding what is worth doing are not getting absorbed by agents. Anyone claiming otherwise is selling something they can't deliver.
What does change, and fast, is everything underneath that. The production and pattern layers. The work that has historically consumed the £6 share of the marketing budget.
That's the layer Data Agents builds for.
We build the AI agents that own production and pattern work. We sit alongside marketing leaders to direct what those agents do, what data they read and what outcomes they ship. We charge fixed price, fixed time, fixed output, because the model lets us and because the day-rate alternative is failing the buyer.
The marketers winning ground over the next two years are those already buying marketing capability by the outcome. They compound the saving every quarter the agent layer expands.
Want to see how Data Agents can help you do more with less? Book a quick strategy call →