What to Automate First: A Simple Scoring Framework

The first thing most people automate is the thing that annoyed them most recently. Not the thing that costs them the most hours. Those are almost never the same thing.

We see this in nearly every audit. Someone spent a weekend wiring up an automation that renames files in their downloads folder, while three afternoons a week still go to writing the same lead-qualification email by hand. The renamer was fun to build. The email is where the hours were.

So before we build anything for a client, we score everything. The formula is short enough to fit on an index card:

Frequency × Friction ÷ Judgment. Highest score gets automated first.

That's it. Here's how each variable works, and then a worked example so you can run it on your own week.

The three variables

Frequency is how many times per week the task actually happens. Not how often it feels like it happens — count it. Check your sent folder, your calendar, your invoice history. People routinely misjudge this by 2× in both directions.

Friction is a 1–10 score combining two things: minutes per occurrence, and dread. Dread matters because a task you avoid gets done late, and late has its own cost. A 10-minute task you procrastinate on for three days is not a 10-minute task. Rough guide: 1–3 for tasks under 5 minutes you don't mind, 4–6 for 10–20 minute tasks, 7–10 for anything over 20 minutes or anything you actively avoid.

Judgment is a 1–10 score for how much human taste the task needs. Copying an invoice number into a spreadsheet is a 1. Deciding whether to fire a client is a 10. This one goes in the denominator: the more judgment a task needs, the lower it scores, because automating it badly costs more than doing it by hand.

A worked example

Here are five real tasks from a recent audit of a solo marketing consultant, scored honestly:

• Answering pricing and availability emails. Frequency 10/week, friction 3, judgment 2. Score: 15.
• Chasing late invoices. Frequency 3/week, friction 8 (only 15 minutes each, but she dreaded them enough to skip weeks), judgment 2. Score: 12.
• Researching leads before intro calls. Frequency 4/week, friction 6 (about 25 minutes each), judgment 3. Score: 8.
• Categorizing expenses for the bookkeeper. Frequency 1/week, friction 5, judgment 1. Score: 5.
• Writing proposals. Frequency 2/week, friction 9, judgment 7. Score: 2.6.

Two things worth noticing. First, the winner is boring. Pricing emails are not anyone's idea of an exciting automation project, but 10 occurrences a week at even 8 minutes each is 80 minutes, every week, forever. Second, proposals — the task she wanted automated, because it hurt the most — scored last. High friction, sure. But at 2 per week with heavy judgment, it's the wrong first build. It became the fourth build, and even then only the research-and-first-draft half of it.

The framework's real job is to protect you from your own instincts. Annoyance is a bad prioritization signal. Arithmetic is a decent one.

Where agents change the math

The formula predates AI agents, and for years it quietly told everyone the same thing: automate the low-judgment stuff, leave the rest alone. What's changed is the denominator.

Tasks that used to score judgment 7 or 8 — drafting a client update, triaging an inbox, summarizing a discovery call, doing pre-call research — can now be scored more like a 3 or 4, if you keep a human approval step. The agent does the 25 minutes of assembly; you do the 90 seconds of taste. You're not automating the judgment. You're shrinking the task down to only the judgment.

That's the distinction that matters when re-scoring your list, and it's the difference between an automation and an agent — we wrote up the full breakdown in agents vs. automations. When you re-run the numbers with that in mind, tasks like proposal drafting jump from the bottom of the list to the middle. They still shouldn't be first. But they're no longer off the table.

How to run this on your own business

1. List every task you did more than twice in the last two weeks. Aim for 15–25 items. Your sent folder and calendar are the honest record; your memory is not.
2. Score each: frequency per week, friction 1–10, judgment 1–10.
3. Compute frequency × friction ÷ judgment. Sort descending.
4. Build the top item. Just the top one. Run it for two weeks before touching number two.

One warning: resist the urge to adjust scores until your favorite project wins. If the arithmetic says pricing emails and your heart says proposals, the arithmetic is right. The proposals will still be there in six weeks, and by then you'll have 8 hours a month back to spend on them.

If you want the longer version of this exercise — the one where we go looking for the hours you've stopped noticing — start with the automation audit. The scoring framework is step two. Step one is admitting how you actually spend your week.