Automation has a reputation problem in marketing, and most of it is earned. Plenty of “AI-powered” agencies mean an AI wrote it and nobody read it before it went out. That is not automation working well, that is a person’s judgement being removed from the one place it mattered most.
The useful version of automation does something different. It takes the repetitive, rule-based, low-stakes work off a person’s plate so their time goes toward the decisions that actually need a person.
The test that decides the line
There is a simple test for whether a task belongs to automation or a person: what happens if this goes slightly wrong, and who notices first.
If a reminder email goes out a day late, or a report generates with last month’s figures still in a header, the cost is minor and easily corrected. That work can run on rules, checked periodically rather than checked every time.
If a client reads a message that gets their situation wrong, or a prospect receives an email that misreads the context of a recent conversation, the cost is reputational and it lands immediately. That work needs a person’s eyes before it leaves the building, every time, no exceptions.
What genuinely belongs to automation
Data movement and syncing. Keeping a CRM, a spreadsheet and an email platform in agreement with each other is exactly the kind of task a machine does more reliably than a person doing it manually at 6pm on a Friday.
Sequencing and timing. Deciding that a follow-up goes out three days after no reply, or that a lead who fills in a form gets a specific next step, is a rule. Rules are what automation is good at.
Drafting the first pass. Automation producing a first draft of a report, a summary, or a routine update is a genuine time saving, provided the word “draft” is not quietly dropped before a person has actually looked at it.
Monitoring and flagging. Automation watching for a payment that has not landed, a workflow that has failed, or a metric that has moved outside a normal range, and telling a person about it, is exactly the right use of a machine’s patience.
What should never run without a person
Anything a client will actually read and remember. An email, a proposal, a piece of content with the client’s name on it. This is the one line we hold without exception, because it is the moment where a machine’s confident wrong guess costs the most.
Anything involving judgement about a specific relationship. Whether now is the right moment to raise a difficult point with a client, or how to phrase feedback that could land badly, requires knowing the person, not just the rule.
Anything where being wrong is expensive to undo. A published article with a factual error, an email sent to the wrong segment, a proposal with the wrong figures. Automation can prepare all of these. It should not be the last check on any of them.
Running it this way ourselves
We run more than forty automations across our own business, from lead follow-up to reporting to internal tracking. None of them send a word to a client without a person reading it first. That is not a caveat we add reluctantly, it is the actual design.
The honest version of “AI-powered marketing” is a business that says plainly what runs on rules and what a person checks, and holds that line even when it would be faster not to. The dishonest version says “AI-powered” and hopes nobody asks the follow-up question.
If you are weighing up how much to automate in your own marketing, or wondering whether a supplier’s automation still has a person behind it, that second question is worth asking directly. If the answer is vague, that is usually the answer.
