Which Agency Workflow Should You Automate First?

Someone on your team checked campaign pacing this morning. Someone else pulled numbers for a client update or built ad groups from an approved plan. If you’re deciding what to automate first, start with that list.

A demo of a media plan built in seconds or a report that writes itself is more exciting. But it leaves out the part you’ll have to figure out: whose judgment goes into the work, how often the process changes, and who catches a mistake before it reaches a client.

At KORTX, automated monitors track a range of campaign variables and flag issues that need attention. When a monitor catches an issue, it opens a ticket with an owner and a deadline. A person investigates and decides what to change.

That’s a useful first project: a defined task, a result someone can check, and a person still making the spending decision.

Start with last week’s work

Ask your account and ad ops teams to list the tasks they repeated last week. Be specific enough that someone else could recognize the work: checking campaign pacing, building ad groups from an approved plan, pulling numbers for Monday’s client update, or renaming creative files.

These tasks are easy to overlook because they’re familiar. Add up the time across the team, though, and you’ll have a useful place to start.

The scorecard

Score each task from 1 to 3 on five questions. Higher is a better first candidate.

I’d start by looking at tasks that score 12 to 15. Anything under 10 probably needs more work before it’s a good first project, however much time it takes.

Before choosing, look closely at two answers. Can you write it down? Ask two people who do the task to describe the steps separately. If their lists don’t match, agree on one process and run it by hand for a few weeks. Otherwise, you’ll automate whichever version the person building it happens to follow.

Then ask what happens if it’s done wrong? That answer can change which task I’d choose, even when the totals are close.

Three candidates, scored

These are my estimates for three familiar agency workflows, based on how the work looks at KORTX. Score them for your own team; your answers may be different.

Daily checks come out on top. Someone looks across the campaigns each day: Are they spending? Are they on pace? Are they hitting the client’s targets? Much of that work follows rules the team already uses.

The 3 in the last column assumes you keep the existing human review while you evaluate the monitor. A false alarm costs someone time; a missed alert can let a real problem continue. Compare what the monitor flags with what people catch before relying on it.

Ad group builds need a closer review. The steps are clear and the work is frequent, but a mistake that gets through to launch can spend real money in the wrong place. That’s why we check each build against the approved plan before anything goes live.

My starting rule is to automate a check before you automate an action. It gives the team a chance to see how the automation behaves while a person still decides what to do.

Pacing problems, zero spend and missing campaign connections are a few examples of what our monitors check across every campaign. When one trips, it opens a ticket with an owner and a deadline. We’re still tuning the alerts so that what fires is worth a person’s time.

The one I’d hold off on

The monthly client recap is a tempting first project. It takes hours, and AI tools can produce a fluent draft in seconds.

But a useful recap depends on what the account manager knows: what this client cares about this month, which number needs explaining, and what decision the report should help them make. Those choices change from month to month, and the client sees the result directly.

That’s why I’d score the steps, not the whole job. Pulling numbers from five platforms into one table is easier to define and verify. Writing what those numbers mean for this client takes more judgment. Automating the data collection gives the account manager more time for the explanation.

Where the scorecard falls short

Use the total to start a discussion. A few things still need a closer look:

• The scores are judgment calls. If two people score a task differently, ask why. They may be following different processes or dealing with different exceptions.
• A total can hide a dealbreaker. A task can score 13 and still have a 1 in the last column. That needs a review step before a mistake can reach the client or spend money.
• Automation needs upkeep. Platform settings change and data feeds break. Someone needs to own the workflow and notice when it stops working. I’ll cover that in a later post.
• Hours saved isn’t the only payoff. Catching mistakes matters too. Record how long the task takes now and how often it goes wrong so you can judge what improved.

You can choose a first project before deciding whether it needs AI. Many daily checks follow a fixed rule: if spend is zero, flag it. Use a rule or spreadsheet formula where it does the job. Consider AI for steps that require interpreting language, with a way to review its output. A separate post looks at where that line falls.

Try this this week

Bring three recurring tasks to your next ops meeting and score them with the people who do the work. Start with the highest-scoring candidate that has no 1 in the last column and someone willing to own it.

For that task, write down the steps, the time it takes, how often it goes wrong, and who will check the result. If people disagree on the steps, settle those first. That’s enough to choose a first project.