How do I know what to automate in my business?

You run a 20-person firm, everyone keeps telling you to automate, and every tool demo promises the same thing. The problem is not motivation. It is picking the right target. Ask ten people what to automate in my business and you get ten answers, most of them a tool name instead of a task. The useful question is narrower: which specific jobs your team does by hand are worth handing to software, and which ones are not.

Quick Answer: Automate a task when it is repetitive, follows the same steps every time, runs on rules rather than judgment, and moves information between systems. The strongest candidates are high-frequency, low-decision jobs like data entry, approvals, and scheduled reports. Score each task on those four traits, then start with the highest score.

What makes a task worth automating?

A task is worth automating when it scores high on four traits: how often it runs, how rule-based it is, how much information it moves between systems, and how little human judgment it needs. The more a task looks like copy, paste, repeat, the better a candidate it is.

Which of the high scorers you build first is a separate, scoping call, and it is the core of what an implementation project works out before anyone touches a tool. Here is the test you can run on any task in your business in under a minute.

A quick way to sanity-check the score: the software vendor Zapier asked its power users how they decide, and the common rule was simple. If a task takes more than 15 to 20 minutes a day, follows the same steps, and uses software from start to finish, it is worth automating. Fifteen minutes a day is about 60 hours a year for one task. Most firms have a dozen of them.

If a task scores high on all four traits, it belongs on the list. Which one to build first is a decision about hours and risk, not about software.

What are the best tasks to automate first?

The best first tasks are the ones where a person reads information in one place and types it into another. That single pattern, moving data between systems, is the most common and most reliable win in almost every business. Approvals and scheduled reports come next.

There is a useful reframe here. You are not automating a job. You are automating a task inside a job. McKinsey's research on automation found that fewer than 5 percent of jobs can be fully automated, but in about 60 percent of occupations, at least one-third of the work could be automated with current technology. The same research names collecting and processing data as the categories machines already do faster than people, including back-office work in accounting and paralegal roles. So the paralegal keeps the job. The retyping goes to software.

The best candidates cluster in a handful of everyday workflows:

When workers were asked which tasks they most wanted off their plate, data collection and entry topped the list, followed by approvals and status updates. That matches what we see in the field. Most of these map to a small set of operations and admin solutions rather than anything exotic. For a deeper look at ordering, see which processes to automate first.

How do I find what to automate in my business?

To find what to automate in my business, track where the hours go for one week, then run each recurring task through the four-trait test. You do not need software or a consultant, just an honest log of the manual steps your team repeats, which most owners have never written down.

Here is a method that takes about a week:

1. Log the repeats. For five working days, have each person note any task they do more than once that follows the same steps. Keep it to one line each.
2. Circle the data movers. Mark every task where someone reads information in one tool and types it into another. These are almost always your best candidates.
3. Score each one. Run the survivors through the four-trait test above. Anything high on all four goes to the top.
4. Estimate the time. Multiply minutes per run by how often it runs. A five-minute task done 20 times a day is bigger than a two-hour task done once a month.
5. Pick one. Start with the single highest-scoring, highest-time task. One working automation beats five half-built ones.

This is not a small pile. More than 40 percent of workers say they spend at least a quarter of their week on manual, repetitive tasks, according to a Smartsheet survey, with email and data entry taking the most time. In the law-firm audits we run at Eleventh AI, each one has found more than ten automation opportunities. A structured look surfaces more than memory does. If you want the framework version of this, where to start with AI and automation walks through the first workflow in more detail.

What should you not automate yet?

Do not automate work that needs human judgment, a process that is still broken, or anything built on data you cannot trust. Some tasks pass the four-trait test on paper and still should not be your first move. Three cases come up again and again, and getting them wrong is how automation projects quietly fail.

Do not automate on top of data you cannot trust. If the same client shows up three different ways across your spreadsheets, automating a report on top of that just produces wrong numbers faster. For a 500-employee New York real estate company, the work that mattered first was building one source of truth for their data, not automating the reports sitting on top of the mess. Fix the foundation, then automate.

Do not automate a broken process. If a workflow is a tangle of workarounds, automating it locks the tangle in place. Simplify the steps first. Often that alone removes the task, and then there is nothing left to automate.

Do not automate judgment or relationship work. Client complaints, hiring calls, and pricing decisions stay with a person. Software handles the parts around them: the intake, the scheduling, the follow-up logging.

The honest "not yet" case: if a process changes every month or has no fixed rules, wait. You will spend more time rebuilding the automation than the task ever cost you. Automate the stable, boring, repeating work first. If you are unsure your operation is ready at all, is my business ready for AI is a fair gut-check before you spend anything.

How much does automating a process cost, and when does it pay off?

The cost of automating a single process depends far more on how many systems have to connect than on the software itself. The real driver of payback is the hours the task eats today. A task that costs one person an hour a day pays back fast. A task done twice a year rarely does.

Do the math before you build. Take the weekly time a task costs, multiply by the loaded hourly cost of the person doing it, and compare that to the one-time setup cost plus any monthly tool fees. If the task runs often, the weekly saving repeats while the build cost is paid once. That is why frequency sits at the top of the four-trait test.

---

Not sure which of your workflows to score first? That is the exact question our AI assessment answers. Six questions, about two minutes, and you get a personalized preview of where automation would pay off in your business before you spend anything on an audit or a build. Start assessment.

---

More practical guides on finding and scoping automation live on the Eleventh AI blog.

Frequently asked questions

How do I decide what to automate in my business first?

Score each recurring task on four traits: how often it runs, whether it follows fixed rules, whether it moves data between systems, and how little judgment it needs. Tasks that score high on all four, like copying form data into a CRM or sending scheduled reminders, are your first targets. Then weigh time cost: a five-minute task done 20 times a day beats a long task you touch once a month. Start with the single highest-scoring, highest-time task rather than trying to automate everything at once. One working automation is worth more than five half-finished ones.

What is usually the first thing a small business should automate?

For most small businesses, the first win is moving information between systems that a person currently retypes by hand. Client intake into a CRM, invoice reminders, and the weekly report that someone rebuilds from the same sources every Monday are the usual starting points. These qualify because they run often, follow the same steps, and need no judgment. Workers themselves rank data collection and entry as the tasks they most want automated. Pick the one that costs the most total hours per week, since frequent tasks pay back the setup cost fastest.

Is my business too small to bother automating anything?

What matters is not headcount but how much repeated, rule-based work your team does. A 15-person firm that retypes the same client details dozens of times a day has a clear case. A team with almost no fixed, recurring process has little to automate, and probably little to audit yet. The four-trait test tells you which side you are on. If your people repeat structured tasks that move data between tools every week, there are hours to reclaim, and that is usually true once a business has real operational workflows running.

How do I know if a task is too complex to automate?

Complexity is less about difficulty and more about consistency. A task is too complex to automate cleanly if the steps change based on the situation, if it needs a person's judgment, or if there are more exceptions than rules. A good test: could you write the task down as a set of if-this-then-that instructions a new hire could follow without asking questions? If yes, it is a candidate. If every case needs a decision, keep it with a person and automate the routine steps around it, like the scheduling and the record-keeping.

Does automating tasks mean I have to lay off staff?

Usually not. Automation typically removes tasks, not roles. McKinsey found fewer than 5 percent of jobs can be fully automated, while about 60 percent contain at least a third of work that could be. In practice that means a paralegal stops retyping intake forms and spends that time on client work, not that the paralegal is gone. Most owners we work with are trying to grow without adding headcount, so the saved hours go straight back into the work only people can do. The tasks that disappear are the ones nobody wanted to do anyway.

How long does it take to automate one process once I pick it?

It depends mostly on how many systems have to connect and how clean your data is. A single automation that moves data between two tools you already use can be running in days. Something that touches several systems, or that first needs your records cleaned up, takes longer to plan than to build. The slowest part is almost never the software. It is deciding exactly what the process should do, which is why mapping the workflow before building saves the most time. Start with one process, get it working, then move to the next.