All posts

What you can automate (and what you shouldn't) in design and marketing

Automation isn't about doing everything with a tool. It's about getting repetitive work off your plate so you can give your time and focus to what needs judgment. A post with examples, limits and a real case.

A hand holding a phone showing an AI announcement.

There’s a lot of FOMO around automation, and a lot of fantasy too: you press a button, the machine does everything and you sit back while the work solves itself. The reality is duller and more interesting at the same time. Automation doesn’t replace judgment: it puts judgment where it belongs.

In design and marketing, automation is good for one very specific thing: taking repetitive, predictable, low-value tasks off the table so you can free up time for the work that actually needs a person. If you’re not clear on what that is, automating can be a very efficient way to make more noise.

First organize, then automate

The rule I’ve been using is simple: I don’t automate a process I don’t understand. Automating a mess only produces a faster mess. Before thinking about tools, you have to map the process, remove steps, standardize what repeats and only then look for where the machine fits.

One example: monthly reports. For years, the end of the month at Cross MKT was a string of manual tasks: log into each platform, export data, paste it into a spreadsheet, build charts, write comments, turn it all into a PDF, send it by email. The repetitive part was collecting and formatting. The valuable part was interpretation and decision-making. Yet both took the same amount of time — if I was lucky.

Once I separated the two, automation showed up. Not to replace the analysis, but to give it more time and better quality.

What you can automate

  • Collecting and consolidating data from different sources.
  • Recurring reports with the same structure.
  • Scheduled emails, reminders and alerts.
  • Exporting assets in different sizes and formats.
  • Publishing and scheduling social content.
  • Backups, updates and maintenance tasks.
  • Invoicing, quotes and payment tracking.
  • Campaign tagging, UTMs and link tracking.
  • Repeatable quality checks: broken links, missing images, metrics out of range.
  • Client onboarding with forms and documents.

They all have something in common: they’re frequent, rule-based tasks with an easy-to-check result. If the process changes every month, automating it is a waste of time. If the result can’t be checked, don’t bother.

What you shouldn’t automate

  • Strategy. No template knows what to do with a client who has changed markets, audiences or priorities.
  • Creative direction. AI can generate options, but choosing an identity, a tone or an idea takes judgment and context.
  • Client relationships. An automated email doesn’t replace a difficult conversation.
  • Final judgment. Quality review, reading a weird number, deciding to pause a campaign.
  • Crisis. When something breaks, automation doesn’t think for you.
  • Copy with a voice. You can assist it, but you can’t delegate it. A brand’s voice is an editorial decision.
  • Pricing and prioritization. They’re business decisions, not production tasks.

Automating these things isn’t just risky: it’s a way to hide a lack of judgment behind a tool.

The hidden cost

Every automation has a maintenance cost. Integrations break, APIs change, data gets dirty, credentials expire. If you automate something critical, it needs an owner, documentation, alerts and a manual plan B. An automation without maintenance is a silent time bomb.

There’s also a dependency cost. If your whole process lives inside one tool, migrating becomes expensive. Before you automate, ask yourself: can I export the data? Can I do it by hand if the tool disappears? How long would it take to rebuild?

How I decide

I use three questions:

  1. How often does it repeat? If it’s once a year, it’s not worth it.
  2. Does it follow clear rules? If it needs judgment every time, it’s not a candidate.
  3. Is the error detectable and reversible? If a silent failure can be expensive, it needs human oversight.

If all three answers point the right way, I automate. If not, I document the process and leave it manual. Not everything has to be a machine.

The Clarika case

Clarika was born out of this problem. At Cross MKT, monthly reports were eating up days. Not because the analysis was impossible, but because collecting and formatting were manual. I built a tool that consolidates each client’s data and generates clear, branded reports in minutes. What used to be a week of work is now a review.

Clarika doesn’t automate strategy. It automates the part that shouldn’t be eating up time. That distinction is what makes the tool useful: the time it frees up goes back into analysis, content and clients.

The right question

The question isn’t what I can automate, but what a person should still be doing. Let the machine do the repeatable. Let someone with time to think do the work that requires judgment. That’s the only automation worth doing.

Next post

Side projects without neglecting the main job