Almost everyone starts from the tool. They open an automation platform's website, look at the plans, pick one, and then wonder what to put inside it.
That is the wrong order, and it always shows the same way: six months later the tool is paid for and it contains two welcome emails.
How do you know a process is ready to automate?
If you cannot explain it to a new hire in five minutes, it is not ready to be automated.
That is not a joke. An automation is the transcription of a procedure: if the procedure has holes, the automation runs those holes at machine speed, every day, without anyone noticing.
Before automating anything, three conditions have to hold:
- NOIIt is already written down — a document exists, not just somebody who "knows how it goes"
- NOIIt is repetitive — it happens at least a few times a week, otherwise the setup time never pays back
- AIIt is measurable — you can say whether it worked without opening five screens
If one of the three is missing, the work is not automating. The work is writing the procedure.
What should you automate first?
| Order | What | Why first |
|---|---|---|
| 1 | Transactional emails | fixed text, immediate effect, minimal risk |
| 2 | Lead follow-up | where most money is lost to pure slowness |
| 3 | Cart recovery | measurable return in days, not months |
| 4 | Recurring reports | frees up people's hours every week |
| 5 | Segmentation | needs clean data, so it comes later |
| 6 | Content | the most visible and the riskiest: last |
The order is not arbitrary. Those four share one property: if they break, you notice immediately. An email that does not go out is visible; a report that does not arrive is visible. Bad segmentation, by contrast, works quietly for months before anyone notices people are getting the wrong campaign.
Cart recovery ranks this high because the pool is huge: the Baymard Institute puts the average documented cart abandonment rate at 70.22%, across 50 studies.
Lead follow-up, which is nearly always the point
If I had to name one, it is this.
A lead who fills in a form and hears back the next day is worth far less than the same lead contacted straight away. That is not about politeness: it is that in the meantime they have written to three others, and whoever answers first takes the conversation. That most companies answer online leads too slowly is not our hunch: it is the finding of Oldroyd, McElheran and Elkington's research in Harvard Business Review back in 2011.
The useful minimum is three pieces: an immediate confirmation saying when the real answer will arrive, an internal notification to whoever has to send it, and an automatic reminder if nobody has after 24 hours. Nothing else is needed, and this gets built in an afternoon.
What should never be automated
Three things, and they are always the same three.
The first reply to a complaint. Automating it is the fastest way to turn an unhappy customer into a lost one. You automate the notification to whoever has to answer, not the answer.
Pricing decisions. Automatic discounts, cart-abandonment reductions, dynamic pricing: they look clever and they teach customers to wait for the discount.
The judgement about what to publish. Generating drafts is fine, and we do it. Deciding a draft is publishable is an editorial call, and delegating it to the machine shows. It is exactly how AI slop gets made. And the risk is not theoretical: per Eurostat, 34.70% of EU enterprises using AI in 2025 used it for marketing or sales.
What you can genuinely hand an agent is in the AI agents guide; the tools we use and what they cost are in the stack.
How long does automation take to pay back?
On the top two of the list, days. An automatic follow-up gets built in an afternoon and the difference shows on next week's leads.
On the rest, weeks. And there is one thing worth budgeting for up front: an automation is not finished when it goes live, it is finished when somebody checks on it now and then. The ones that break quietly cost more than they ever saved, and they always break at the point nobody is looking at.
Frequently asked questions
Where do you start with marketing automation?
With transactional emails and lead follow-up: fixed text, immediate effect, and if they break you notice at once. You do not start from the tool, you start from the process. Picking the platform before knowing what goes in it is why six months later it contains two welcome emails.
When is a process ready to be automated?
When it is already written down, repetitive and measurable. The quick check is this: if you could not explain it to a new hire in five minutes, it is not ready. Automating a procedure with holes means running those holes at machine speed, every day.
What is not worth automating?
The first reply to a complaint, pricing decisions, and the judgement about what to publish. The first two damage the customer relationship; the third is exactly how AI slop gets made: generating drafts is fine, deciding a draft is publishable is an editorial call.
Do I need a paid tool to start?
Not at the start. Transactional emails and follow-up run on what almost everyone already has: the email system and an internal notification. The tool becomes necessary when volume grows, or when there are enough automations that you need to see them in one place.
How long before there is a return?
On lead follow-up, days: it gets built in an afternoon and the difference shows on next week's leads. On the rest, weeks. Budget for maintenance too: an automation is finished when somebody checks on it now and then, not when it goes live.
Does automation replace marketing people?
It moves their time from repetitive execution to strategy and control. The work that disappears is the work nobody wanted to do: copying data, sending the same three messages, filling in reports. What remains, and grows, is deciding what is worth doing and verifying the automations still do what they were built for.
Want to know what to automate first?
Tell us how you work now and we will say where we would start. Seven questions, founders answer within 48 hours.
ApplyKeep reading
Founder of As Growth, an AI-first growth partner. He runs growth and strategy across partner businesses: OroEtic, Giojello, Equity Plus and others. Apply if you want to work with us.