Marketing automation or headcount? A decision grid for scale-ups, the lessons from Klarna and IBM, and the order of operations that protects your payroll.
Three months after the funding round, the marketing hiring plan lands on the table. A growth manager, two junior acquisition profiles, a data analyst, a CRM manager. Five positions, easily 350,000 € in fully loaded payroll per year. My question at that point is always the same: how many of these job descriptions describe work that will still exist in eighteen months?
This is not a provocation. It is the new core trade-off of every scale-up: every euro spent on an execution role is a euro that funds neither creative, nor media, nor a senior profile capable of making calls. And for two years now, a growing share of marketing execution has become software.
The hiring reflex comes from a world that no longer exists
For fifteen years, the rule was simple: a new problem, a new hire. More leads to process? An SDR. More reporting? A junior analyst. More creative variations? Another designer. It was rational, because execution required hands.
The math has changed. A junior hire means three to four months of ramp-up, management time taken from your senior people, and a fixed cost that stays in the P&L when growth slows down. On the other side, a well-built workflow runs at night, never resigns, and costs a fraction of a first month's salary. Asking "hire or automate" before opening any position is no longer a tech founder's quirk. It is budget hygiene.
Marketing automation in a scale-up: what actually gets replaced
Let's be precise, because this topic attracts dream sellers. Here is what gets automated today in production, not in demos:
- Reporting. Multi-platform aggregation, CPA drift alerts, a weekly summary delivered to Slack. Nobody should pay a human to copy numbers into a spreadsheet.
- Creative variations. Formats, languages, hook variants from a validated master. In our studio, AI has divided production costs by five on some video formats.
- Lead processing. Enrichment, scoring, routing to the right sales rep in seconds instead of two days.
- Competitive monitoring. Competitor prices, promotions and messaging, collected and summarized every morning.
The underlying shift is massive. Gartner estimates that 33% of enterprise software will embed agentic AI by 2028, up from less than 1% in 2024, and that 15% of day-to-day operational decisions will by then be made autonomously. The right unit of measurement to steer all this is the full-time equivalent freed up: a workflow that saves 0.5 FTE compares very simply against a hire. We detailed the method in our article on the ROI of automations.
Klarna, IBM: the lessons of those who moved too fast
Klarna first. February 2024: the fintech announces that its AI assistant absorbed 2.3 million conversations in one month, the equivalent, by its own account, of the work of 700 full-time agents. Fifteen months later, a change of tune. In May 2025, Sebastian Siemiatkowski acknowledged in a Bloomberg interview that cost had weighed too heavily in the trade-off, that quality had suffered, and that Klarna was hiring humans back into customer service.
IBM next, the same month, a different trajectory. Arvind Krishna explained to the Wall Street Journal that AI agents had replaced about 200 HR roles. But the group's total headcount went up: the freed-up budget went to developers and salespeople, jobs built on creation and relationships.
Both stories say the same thing. You automate tasks, not jobs. Klarna tried to replace an entire job and paid for it in perceived quality. IBM replaced a block of tasks and reinvested the difference where human judgment pays the most. The second approach wins every time.
The "workflow first" rule
Before opening a position, write down the process the person would execute, step by step. If you reach the end of the description, it is a workflow: automate it. If you cannot, because it requires forming hypotheses, negotiating or reading a market, it is a job: hire. And hire senior.
The decision grid: tasks versus judgment
Three scenarios, and they cover most situations.
- Automate what is repetitive, high-volume, governed by explicit rules, and where a mistake is recoverable. A botched Monday-morning report gets fixed by Monday noon.
- Hire for what requires forming hypotheses, building relationships or arbitrating budgets. There, mistakes are expensive and only become visible months later.
- In between, keep a human in the loop. An agent prepares, a human validates. It is the rule we apply to all our critical workflows, detailed in our article on human-in-the-loop.
One example to make it concrete. At Essity, the campaign that generated 14 million impressions at around 6 € per lead owed nothing to extra execution capacity: it was a creative hypothesis, a gamified social mechanic that gave people a reason to participate rather than yet another reason to buy. No AI agent will ever propose that bet. A small, senior team will. Automation only has value if it frees up time for that kind of call.
What a scale-up marketing team looks like in 2026
A short core: a head of growth who makes the calls, a strong creative profile, a data profile. Around them, workflows for repetitive execution and partners for specialized expertise. The org chart I defend fits in one line: few people, very senior, very well-equipped.
One guardrail before you dive in, though. Gartner predicts that over 40% of agentic AI projects will be scrapped by the end of 2027, due to escalating costs, unclear business value and poorly controlled risks. The same firm counts only about 130 genuinely agentic vendors among the thousands claiming the label; the rest is "agent washing", repainted chatbots. In other words, this deserves the same seriousness as a data program: a process audit, clear priorities, progressive rollout. That is exactly what our AI & Automation practice does.
My order of operations, to close: automate first, hire second, for what remains. In that order, every hire becomes a deliberate choice to put human judgment where it pays. In the other order, you lock in fixed costs on tasks that will be software within a year. If you are hesitating on a position right now, bring the job description to our free audit: we will run it through the workflow-first filter together.
Read also
The Jour de Chance Team
Digital acquisition and media strategy experts.