Which Tasks to Automate First: Score Them, Don't Guess

Every owner I talk to has a list of things they'd love to automate. The list is long. And it's almost always in the wrong order.

Most people start with the task they hate most. Or the one a friend automated last month. Or the one a software ad showed them on Tuesday. I get the pull. But I spent more than thirty years running technology, including CTO roles at Angie's List and Experian, and I've watched smart people pick the wrong first project again and again. Not because they were careless. Because they never put their options side by side and compared them the same way.

So this post answers one question: which tasks should you automate first? Not which tasks could be automated. With enough money and patience, almost anything can. Which ones should go first, in a small business, when your own hours are the scarcest thing you've got.

My answer is a simple score. Four factors, 1 to 5 each, total out of 20. I call it the Automation Fit Score, and it takes about two minutes per task.

Your biggest time drain is usually the wrong first pick

This is the part people push back on, so I'll say it plainly. The task that eats the most hours is often a bad first automation.

A task can take ten hours a week and still be a poor candidate. Maybe every case is different. Maybe a mistake would cost you a customer. Meanwhile a smaller drain with clear rules and low risk can be much easier to build and much easier to trust.

Here's an example from the course. Tasha runs a bakery and catering business in Indianapolis with 14 staff. The weekly staff schedule felt like her biggest headache. She rebuilds it in a spreadsheet every week. But when she scored it, the schedule came in at 11 out of 20. The rules were full of exceptions, and a mistake left the shop short-handed on a Saturday. Catering inquiry intake scored 17. That's where she started.

Your first automation has one job: work. You want something you can finish, test and trust in a few weeks. That early win teaches you how these tools behave, and it gives you the confidence to take on the messy stuff later.

One more thing before scoring. Some work should never be scored at all. Tasting consultations, hard client conversations, hiring, tax advice. That's relationship and judgment work, and it stays with a person. Decide that first. No number should talk you into handing off work that belongs with you.

The Automation Fit Score scores four factors from 1 to 5

The Automation Fit Score rates a task on Frequency, Time Cost, Rule Clarity and Error Risk. Error Risk is reverse-scored, so a safe task earns a 5 and a dangerous one earns a 1. Add the four numbers for a total out of 20.

Score	Frequency	Time Cost (per month)	Rule Clarity	Error Risk (reversed)
1	A few times a year	Under 30 minutes	Every case needs judgment	A mistake costs serious money or a key relationship
2	About monthly	30 min to 2 hours	Some steps repeat, most need thought	Hard to undo, upsets customers
3	About weekly	2 to 5 hours	Clear steps, regular exceptions	Noticeable but fixable with an apology
4	Several times a week	5 to 10 hours	Clear steps, rare exceptions	Minor and easy to catch
5	Daily or more	More than 10 hours	Same steps every time	Trivial or caught right away

Then read the total:

• 15 to 20: a strong automation candidate. This is a job for a platform like Zapier or Make, often with an AI step inside it.
• 10 to 14: an Assist candidate. AI drafts or prepares, and a person decides. Often this needs no automation platform at all, just ChatGPT or Claude and a good prompt.
• Under 10: keep it manual for now. Check again in six months.

That middle band saves people real money. I've seen owners build a complicated automation for a task that only needed a decent prompt and five minutes of their attention. A 12 is telling you something. Listen to it.

Scoring one task at a time, one factor at a time

Score each factor on its own. Don't decide how you feel about the task and then work backward to the numbers. Look at Frequency alone and write a number. Then Time Cost. Then Rule Clarity. Then Error Risk. If a task falls between two descriptions, take the lower number.

Each factor has a trap.

Frequency counts the doing, not the calendar. Omar runs a solo bookkeeping practice with 32 clients. His document requests happen "monthly," so his first instinct was a 2. But every month he sends requests to 32 clients, plus follow-ups. He's doing the task several times a week. That's a 4.

Time Cost includes the small pieces. Tasha first wrote that each catering inquiry took ten minutes. When she counted the back-and-forth, the calendar check and copying details into her order sheet, it was closer to twenty. Her honest total was about six hours a week. A clear 5.

Rule Clarity is the factor everyone overrates. You do a task so often that it feels automatic, so you assume the rules are clear. They usually aren't. You're making small judgment calls without noticing. Try this: imagine explaining the task to a smart temp on their first day. Every time you'd have to say "unless" or "it depends," knock off a point.

Error Risk means the realistic worst case. A reminder email sent twice is embarrassing, not dangerous. That's a 4 or 5. A quote with the wrong price could cost you a customer. That's a 2 or 3. Score the task as if nobody checks the output. That gives you an honest read on the danger and tells you how much review the automation will need.

Here's how Tasha's list came out:

Task	Freq	Time	Rules	Risk	Total
Catering inquiry intake	5	5	4	3	17
Invoice reminders	4	3	5	4	16
Weekly social posts	4	3	2	4	13 (Assist)
Weekly staff schedule	3	4	2	2	11

Notice the social posts. Frequent, low risk, but the rules are fuzzy. Every post is a little different. So AI drafts and Tasha approves. That's Assist, not Automate.

When you're done, keep a ranked shortlist of three. Not one, because your first pick might turn out harder than it looked. Not ten, because you'll spread yourself thin and finish nothing. Three gives you a first choice, a backup, and usually one Assist task you can start this week with a plain AI assistant.

If two tasks tie, use these tiebreakers in order. Pick the one where you can see the result clearly. Then pick the one that touches fewer apps. Then pick the one that frees up time for work you actually want to do.

Where the score can fool you

The score is a guide, not a judge. Here's where it goes wrong.

You fudge the risk because you like the task. This is the big one. If you catch yourself nudging Error Risk up to get a favorite task over 15, stop. Ask what you don't want to admit about that task.

Your gut disagrees with the total. Don't ignore that. Don't just overrule the number either. Go back and find the one factor your gut is fighting. It's usually Rule Clarity or Error Risk. Either you scored it carelessly, or your gut is reacting to something the score doesn't measure, like how much you enjoy the work. Fix the first. Write down the second.

You score alone. Ask someone who knows your work to score the same task without seeing your numbers. Where you're two or more points apart on any factor, talk it through. Those conversations turn up steps and risks neither of you had named.

The score doesn't make the decision. You do. It just makes sure you're comparing a reminder email and a scheduling headache with the same four questions, instead of with whatever mood you're in that afternoon. AI is not replacing you. It's amplifying you. And that starts with pointing it at the right work.

Key takeaways

• Your biggest time drain is often a poor first automation, because messy rules and high stakes make it hard to build and hard to trust.
• The Automation Fit Score rates Frequency, Time Cost, Rule Clarity and Error Risk from 1 to 5, with Error Risk reversed, for a total out of 20.
• 15 or more means automate, 10 to 14 means AI assists while a person decides, and under 10 means keep it manual for now.
• Score one factor at a time, count the doing rather than the calendar, and never inflate the risk score to favor a task you like.
• Decide what stays human before you score anything; relationship and judgment work never gets a number.

Frequently asked questions

Which tasks should a small business automate first?

Start with tasks that happen often, take real time each month, follow clear rules, and are low-risk if something goes wrong. Those four traits make a first automation easy to build and easy to trust. The biggest time drain is not always the best first pick; a task with lots of exceptions or costly mistakes is better saved for later, once you have experience with the tools.

What is an Automation Fit Score?

An Automation Fit Score is a simple way to rank tasks for automation. You score each task from 1 to 5 on Frequency, Time Cost, Rule Clarity and Error Risk, where Error Risk is reversed so safer tasks score higher. The total is out of 20. A score of 15 or more is a strong automation candidate, 10 to 14 suits AI-assisted work with a person deciding, and under 10 stays manual.

Should I automate the task I hate the most?

Not automatically. The task you hate most may happen rarely, have unclear rules, or carry real risk if it goes wrong. Dislike is a fine reason to look at a task, but not a reliable reason to automate it first. Score it the same way you score everything else, on frequency, time, rule clarity and risk, and let the comparison decide what goes first.

Do I need Zapier or Make for every task I want AI to help with?

No. Tasks that score in the middle range, roughly 10 to 14 out of 20 on an Automation Fit Score, often need only a general AI assistant like ChatGPT or Claude and a clear prompt. A person reviews the result and decides. Automation platforms like Zapier or Make make the most sense for frequent, rule-based tasks that move information between apps.

What tasks should a small business never automate?

Keep relationship, judgment and high-stakes work with a person. That includes hard client conversations, hiring decisions, pricing exceptions, and legal, tax or medical advice. AI can still help prepare notes or drafts around that work, but a human makes the call and owns the conversation. Decide which tasks stay human before you score anything for automation.

Where this goes next

Scoring is one piece of Module 1 in my self-study class, AI Automation for Small Businesses. It walks you through a one-week Time-Drain Inventory, sorting tasks into what to automate and what to keep human, and ending with a ranked shortlist of three.

See the course