Is it worth automating this process? How to know before spending money

Some tasks are exhausting just to think about: copying data from a form into a spreadsheet, checking whether someone paid, forwarding the same information, chasing an approval, or remembering to follow up with a customer three days later.

When a task keeps repeating, the first reaction is often to automate it.

Sometimes it should be. But not always.

A task can be boring and still not justify an automation. The opposite can also happen: something that takes just a few minutes each time can be happening so many times that, at the end of the month, becomes hours of work, errors and forgotten follow-ups.

The useful question is not only whether something can be automated. These days, there is a way to automate almost anything.

The question is, is it worth automating in your case?

To answer it you don't need to know about APIs, AI agents or integrations. You need to understand the process you already have.

The short answer

A good candidate usually has several of these signs:

It is repeated frequently, it follows quite clear steps, it moves information between tools, it consumes time, it generates errors or delays and it will probably continue to exist for long enough to justify the effort to automate it. If the process changes every week, it happens very rarely, it depends a lot on the human criterion or it is not even clear how it should be done, it is usually better to fix it before automating it.

1. Don't start with the tool

A common mistake is to start with the tool: AI, a WhatsApp connection, a bot, Make, Zapier or a request to make everything automatic.

The tool comes later.

First describe what is happening today, as if you were to explain it to a new person who will have to do the job tomorrow.

For example:

CURRENT PROCESS

A request arrives through the website. Someone opens the message, copies the name and phone number into a spreadsheet, notifies a salesperson through another channel, and creates a reminder to follow up two days later.

Now we can ask useful questions: How often does this happen? Are those steps always followed? What happens if information is missing? Who decides which salesperson gets the request? Where is the follow-up recorded?

That's better than choosing a tool before you understand the problem.

2. The first question: how many times does it really happen?

Frequency matters because an automation has a cost of creation, testing and maintenance.

If a task takes ten minutes but occurs once every three months, it is probably not urgent. If it takes three minutes and happens 80 times a week, the story changes.

Do a simple calculation:

MONTHLY TIME SPENT MANUALLY

Minutes per task × times per month × people involved. Then divide by 60 to estimate how many hours that task takes.

It's not a perfect formula. It doesn't capture interruptions, mistakes, or the time to re-concentrate, but it helps to stop deciding by feeling.

EXAMPLE

A receptionist takes four minutes to record a form submission, copy it into a spreadsheet and notify the person responsible. At 15 requests a day over 22 working days, that is about 22 hours a month spent just moving information.

Now we have something to compare against the cost and maintenance of automating it.

3. Are the steps clear or does each case end up being different?

Automation works best when we can explain the rules.

For example:

• If the form arrives complete, save the data.
• If the selected service is A, assign the contact to Team A.
• If 48 hours go by without an answer, send a reminder.
• If the payment appears as confirmed, change the order status.

That doesn't mean everything should be rigid, but we do need to know what happens normally and what things are exceptions.

If the answer to What do we do next? is It depends five times in a row, the process still needs to be understood.

WARNING SIGN

If two people do the same task in completely different ways and no one knows what the right one is, automating now can turn disorder into a faster mess.

First we have to agree on a way of working, then we decide which part deserves automation.

4. Is the information already where automation can find it?

Many automations don't fail because of logic. They fail because the data goes wrong.

A process can be seen as simple until we discover that:

• names are written in five different ways;
• important information is hidden within free messages;
• no one uses the same order number;
• a sheet has columns that each person interprets differently;
• mandatory data are missing;
• the tool that holds the information does not offer a practical way to connect it.

Before automating, ask if the data needed for the process is sufficiently consistent.

Sometimes the first step toward automation is not building a workflow. It is fixing a form, defining required fields or stopping the same information from being saved in three places.

5. What happens when something does not go as expected?

Every process has exceptions.

The problem is not that exceptions exist. It is failing to account for them.

Imagine an automation that marks an order as ready when a payment confirmation arrives. What if the amount is wrong? What if the customer paid twice, the order number does not match, or the service needs an additional review?

Good automation not only knows what to do when everything goes well. It also knows when to stop and ask for help.

AUTOMATING DOES NOT MEAN REMOVING PEOPLE

In many processes, the best solution handles routine cases automatically and leaves exceptions, sensitive decisions and situations needing context to a person.

That's usually safer than trying to force total automation.

6. Does the task need rules or criteria?

There is an important difference between repeating a decision and making a decision.

Moving data from one system to another, checking if a field is empty, sending a notice, or creating a task are usually easy-to-describe jobs with rules.

Deciding whether a customer is genuinely upset, whether an exception should be approved, whether a proposal addresses a complex need, or whether a situation is too risky may require human judgment.

AI can help with less structured information, such as sorting messages or preparing a draft. But that an AI can participate does not mean that it should have the final decision.

EXAMPLE

You can automatically classify a support message and route it to the right person. That does not mean automating the final reply to a customer reporting a serious problem.

Sometimes the right automation reaches the door of human decision and stops there.

7. Will this process continue to exist within six months?

Automating a task that is about to disappear or change completely can waste money.

Before you build, ask:

• Is the process stable or are we still making it up?
• Are we going to change the system soon?
• Is the volume growing or is it temporary?
• Does the rule depend on a specific promotion or situation?
• Will anyone be responsible for reviewing automation when the business changes?

Automation also ages. If you change a form, API, policy, or tool, you may need maintenance.

That is why the build cost is not the only question. You also need to consider how long the automation will remain useful.

8. Don’t automate a process just to stop seeing it

This is probably the most important rule.

If the process is poorly designed, automating it does not make it good.

For example, if a request goes through four approvals that no one knows why they exist, the first question should not be how to automate the four approvals.

It should be: do we really need four?

Perhaps the best automation is to remove two steps and automate one of the remaining steps.

SIMPLIFY FIRST

Before automating a workflow, look for duplicate steps, unnecessary approvals, unused data, avoidable messages, and tasks that exist only because two tools do not communicate.

Automating less, but doing it well, often creates more value than building a huge workflow around a process no one has questioned.

Three possible outcomes, and all three can be right

Reviewing a process should not always end with a yes.

A complete example: following up on requests

Suppose a small business receives requests through its website and WhatsApp.

Today this happens:

• someone reads the request;
• copies the data to a sheet;
• chooses who should handle it;
• sends an internal message;
• two days later, checks the spreadsheet for a response;
• if no one has replied, follows up again.

At first glance, it seems like a perfect candidate for automation, but you still have to ask.

WHAT MAY BE EASY TO AUTOMATE

• save form data automatically;
• assign requests according to a clear rule;
• notify the person responsible;
• create a reminder;
• record the date of the last contact.

PART THAT MAY NEED A PERSON

• decide what to do with an incomplete request;
• interpret a very ambiguous request;
• handle an upset customer;
• approve a commercial exception.

The result does not have to be a robot that does everything. It can be a process that removes repetitive work and brings a person in where their judgment adds value.

Where does artificial intelligence fit in?

Not every automation needs AI.

If the job is to save form data and notify someone, straightforward automation is probably simpler, more predictable and easier to maintain.

AI may make more sense when the input is less structured: a long email, a document, a conversation, an open-ended description, or a classification that does not fit a simple if A, then B rule.

Even then, AI does not need to run the entire process.

It can read, summarize, extract information or suggest a category, while a rule or a person decides what happens next.

A USEFUL QUESTION FOR 2026

Instead of asking How do we add AI here?, ask which part of the process needs to interpret ambiguous information and which part simply needs to follow rules.

You do not always need to build something from scratch

Another way to overspend is to build a solution of your own for something your current tool can already do.

Before developing, review three levels:

• Existing functions: perhaps the CRM, billing system or platform you already pay has an automation included.
• Integration: Maybe two existing tools only need to share information.
• Specific development: it makes sense when the rules, restrictions or experience you need do not fit well into existing solutions.

The right solution might be a five-minute rule, a connection between existing tools or a more complete system. Its size should follow the problem, not enthusiasm for technology.

Check this before requesting automation

Choose one specific task, not an entire sales or administration department.

Then answer yes or no:

• Does it happen at least several times a week?
• Can I describe the steps without improvising too much?
• Do most cases follow the same path?
• Do the necessary data already exist in accessible systems, forms or files?
• Does the task take time that could be used on something more valuable?
• Do errors or delays have any cost?
• Do I know what should happen when an exception occurs?
• Will the process probably continue to exist over the next few months?
• Is there a person responsible for checking that it's still working?

The more yes answers you have, the more reason to consider automation. But this is not a score: one important factor, such as an error that could seriously harm a customer, can change the design completely.

What to bring to a conversation about automation

You don't need to come up with a technical diagram.

Bring this:

• a real example of the task;
• how many times it happens;
• who does it;
• which tools you use;
• what information comes in;
• what result should come out;
• what exceptions you remember;
• what is most annoying or time consuming;
• what would happen if automation fails;
• what you'd like to stop doing manually.

That makes it much better to evaluate whether it is appropriate to automate, simplify, integrate tools or simply leave the process as it is.

Good automation starts with knowing what not to automate

Good automation is not measured by how many steps it completes on its own.

It is measured by whether it eliminates unnecessary work without creating a new problem.

Sometimes the best result is to save twenty hours a month. Sometimes it is to prevent a request from being lost. Sometimes it is to reduce errors. And sometimes the correct conclusion is not to build anything yet.

That's a good decision, too.

Start with a specific task.

Explain it as it happens today.

Then we can decide which part is genuinely worth automating.

Next step