Automate the Most Boring Tasks, Not the Most Important Ones

When a company starts thinking about automation, temptation quickly appears:

if you can automate something, maybe it’s best to automate as much as possible.

Customer service.

Sales.

Recruitment.

Decisions.

Contact with contractors.

Creating offers.

Responding to complaints.

After all, AI can write, analyze, classify, summarize and take increasingly complex actions.

Technically, you can do a lot.

Except the good question isn’t:

“What can we automate?”

Better sounds:

“What should people no longer do manually?”

It’s a subtle difference, but it changes the whole approach.

The greatest value of automation is not always in replacing humans on the most important tasks.

It often means giving people time to finally focus on those tasks.

A bad approach to automation starts with spectacular things

Companies like to automate what looks good in a presentation.

“AI will independently respond to customers.”

“The system will make decisions.”

“A salesperson will no longer be needed for first contact.”

“The entire process will be autonomous.”

Sounds ambitious.

Meanwhile, an employee next door every day:

• copies the order number from one system to another,

• searches for the customer’s e-mail address,

• manually marks the case status,

• translates messages in a separate application,

• looking for a ready answer in an old document,

• pastes data to a spreadsheet,

• checks five inboxes,

• remembers who he was supposed to reply to,

• sends the same document for the twentieth time.

This is where the best automation usually lies.

It’s not glamorous.

But it can save the company dozens of hours a month.

First, remove the work that no one likes

A good candidate for automation is a task that meets several conditions:

• repeats itself often

• looks almost always the same,

• is based on clear rules,

• does not require much responsibility,

• the error can be easily detected or fixed,

• man does not add much value to it.

If an employee says:

“The worst thing is that every day I have to…”

it’s worth listening carefully.

This sentence is often followed by a process that should be simplified.

”I have to rewrite data every day”

Automate.

”I have to check every day to see if anyone has replied”

Automate reminders and statuses.

”I have to copy a message to the translator every day”

Automate translation.

”I have to send the same instruction every day”

Prepare a template or automated message.

”I have to open five apps every day”

Connect your data or communications in one place.

”Every day I have to decide whether to accept the complaint for PLN 20”

The problem here may not be automation.

It may be the lack of a clear decision rule.

Automation should not take away from what a person is best at

Man has a huge advantage where:

• empathy,

• responsibility,

• sense of the situation,

• negotiations,

• creativity,

• interpretation of incomplete information,

• building relationships,

• understanding the context,

• making unique decisions.

If the system takes over these elements, automation becomes risky.

Let’s imagine a complaint.

Customer writes:

“This is the second time the package has arrived damaged. The product was a gift for my son’s birthday and I am really disappointed.”

The system can:

• recognize the topic,

• find the order number,

• check history,

• mark the call as urgent,

• prepare a proposed answer.

This is a great use of automation.

But should the system decide on its own how to resolve the entire situation?

Not necessarily.

There may be context that is not visible in the data. Maybe the customer is a long-time customer. Maybe the product has a high value. Maybe the company has made a mistake before. Maybe the best solution would be a gesture that goes beyond the standard procedure.

This is where a human can do something that the system shouldn’t do on its own.

The best model: the machine prepares, the human decides

In many processes you don’t have to choose between:

100% handmade

a

100% automatic.

The best option is in the middle.

The system does the boring 80% of the work, and the human leaves the most important 20%.

Example.

The client sends a request for quotation.

Automation can:

1. read the message,

2. recognize the company,

3. find previous history,

4. classify the topic,

5. draw products and quantities,

6. prepare data in the appropriate format,

7. create a draft of your answer.

Merchant:

8. checks the context,

9. changes the offer,

10. decides on the price,

11. adds your own message,

12. sends.

That’s a huge difference.

The salesperson does not spend time copying and organizing data.

He spends it where his experience has value.

Automate information transfer

One of the most absurd sources of work in companies is the manual transfer of information.

The data already exists.

But humans have to move them from place A to place B.

Examples:

• order from store to sheet,

• customer data from the form to CRM,

• message to the ticketing system,

• invoice number into an email,

• shipment status into a customer reply,

• outcome of the conversation into a report.

In such tasks, a human is an expensive API.

And it’s usually not the best use of his time.

If the information is structured and the flow should always be the same, it is worth automating it first.

Automate reminders

People are good at solving problems.

They are much worse at remembering hundreds of small things.

“Call me back tomorrow.”

“Return to the client in three days.”

“Check if accounting has responded.”

“Remind yourself of the supplier.”

“Write after delivery.”

“Reply if the customer doesn’t respond.”

These are perfect tasks for the system.

A person shouldn’t keep the entire company’s calendar in his or her head.

Automation can keep an eye on the deadline.

A person can decide what to do when that deadline comes.

Automate searching, not thinking

An employee receives a customer’s question.

Before he answers, he must find:

• order number,

• product,

• date of purchase,

• previous conversation,

• shipping status,

• applicable procedure.

Five minutes does not go to solving the problem.

It goes to preparing to solve the problem.

The system can do this automatically.

After opening the conversation, it should immediately show everything the employee needs.

This is one of the best examples of automation: we don’t tell people what to think - we simply stop forcing them to look for information.

Automate the language, not the relationship

A customer from Germany writes in German.

The employee knows the product, process and company, but does not know German.

The classic process looks like this:

1. copying messages,

2. opening the translator,

3. translation,

4. writing answers,

5. re-translation,

6. copying,

7. sending.

The human adds no value to stages 1, 2, 5 and 6.

They can be removed.

Automatic message translation does not replace a consultant.

On the contrary.

It allows him to use knowledge and experience without having to perform mechanical work.

The customer continues to talk to the human.

Language simply ceases to be an obstacle.

Automate replies to repetitive, non-difficult conversations

Not every answer needs to be written from scratch.

If hundreds of customers ask:

“How to make a return?”

there is no point in creating every sentence again.

You can use ready answers for live chat and prepare a template:

Good morning, of course. You can report your return within 14 days. The instructions can be found here: […]

The employee selects a template, checks the context and sends it.

This is completely different from the situation:

“The product caused a problem and I am very dissatisfied with the way I was treated.”

In such a conversation, copying a ready-made text may only make the situation worse.

A good rule:

the more predictable the question, the more automation makes sense.

The more emotional and unusual the situation, the greater the value of the person.

Automate classification

A company may receive hundreds of messages.

Someone needs to determine:

• is it a sale,

• complaint,

• return,

• question about delivery,

• invoice,

• cooperation,

• spam.

This is important, but the classification itself rarely requires the employee’s experience.

The system can automatically tag the conversation, assign it to the appropriate category or forward it to the appropriate team.

This keeps people on the case.

Not sorting things out.

Automate data collection

Before the conversation, the salesperson needs:

• company name,

• NIP,

• country,

• type of product,

• number of pieces,

• implementation date.

However, you don’t have to ask every question manually.

The form may collect basic data in advance.

The system can sort them out.

Only later does a human join the conversation.

The same works for:

• complaints,

• returns,

• reservations,

• service requests,

• quote requests.

Automation should set the table.

He doesn’t have to lead the entire meeting.

Don’t automate responsibility

This is one of the most important boundaries.

The system can recommend.

The system can prepare.

The system can classify.

The system can perform actions according to previously established rules.

But when the decision has a significant impact on:

• money,

• customer relationship,

• reputation,

• security,

• terms of cooperation,

• unusual exception,

there should be a person who is responsible for it.

“AI decided so” is not a good business process.

The company is still responsible for the result.

Don’t automate everything that’s easy to measure

This is a common trap.

Easy to measure:

• number of answers,

• number of calls handled,

• reaction time,

• number of telephones.

Much more difficult to measure:

• did the customer feel taken seriously,

• whether the salesperson understood the real need,

• whether the complaint was resolved reasonably,

• whether the customer trusted the company.

The system can therefore optimize the indicator very well and at the same time worsen the actual effect.

Example:

the bot reduces the average first response time to three seconds.

Great result.

But then the client spends five minutes trying to reach the person.

It looks better in the report.

In fact, worse.

Don’t automate a problem you don’t understand yet

If the process is chaotic, automation can only accelerate the chaos.

Example.

The company has a complicated complaints procedure:

1. the employee copies the data to the spreadsheet,

2. sends an email to the manager,

3. the manager forwards the case to the warehouse,

4. the warehouse responds by e-mail,

5. the employee updates the sheet,

6. the customer gets the message.

You can automate all six steps.

But maybe the right question is:

why does this process have six steps?

Perhaps three is enough.

Therefore, the order should look like this:

remove → simplify → standardize → only then automate.

The best automation sometimes involves removing a task

The company is preparing a report.

The employee spends two hours on it every Friday.

Natural idea:

“Let’s automate the report.”

But first, it’s worth asking:

who reads it?

If the answer is:

“I don’t really know.”

the best automation costs zero.

Stop doing it.

The same applies to:

• meetings,

• approvals,

• copying data “just in case”,

• reports,

• additional emails,

• manual checks performed twice.

Don’t rush something just because it exists.

Check if it should exist.

How to find the best places for automation?

You can perform a simple audit.

For a week, ask your team to write down tasks that trigger one of these thoughts:

“Same thing again.”

“I need to transcribe this.”

“I have to remember this.”

“I have to find it.”

“I need to check five places.”

“I need to pass this information on to someone.”

“But I did it yesterday.”

These are the candidates.

Then rate each task in five areas.

Question	If the answer is yes
Is it repeated often?	good candidate
Does it have clear rules?	good candidate
Does it require little human judgment?	good candidate
Is the error easy to detect?	the risk is lower
Does it take a lot of time in total?	priority increases

The best automations are often found in the top right corner of this table: frequent, predictable, time-consuming, and boring.

Example: employee’s day before and after automation

Let’s imagine a person operating an online store.

Before

8:05 - Checks email.

8:12 - Opens Instagram.

8:18 - checks WhatsApp.

8:25 - Answers a question about delivery.

8:28 - copies the German message to the translator.

8:32 - looking for order number.

8:36 - asks the warehouse about availability.

8:44 - Returns to the previous client.

9:02 - He remembers the message from yesterday.

A large part of the job is not serving customers.

It involves organizing opportunities to serve customers.

After

Messages go to one inbox.

The system recognizes the language.

The translation is done automatically.

The most common answers are available as templates.

Conversations have status.

The employee sees which ones require a response.

Result?

The employee continues to serve customers.

Only a much smaller part of the day is spent on activities that do not require a human.

And that’s exactly the point.

AI should empower people, not just reduce their numbers

The most primitive question about AI is:

“How many employees can we replace with this?”

Much more interesting:

“How much more can our current team do if we stop wasting their time?”

Perhaps the salesperson can conduct more valuable conversations instead of preparing data.

Instead of translating messages, a consultant can solve problems faster.

Instead of preparing reports, the manager can improve the process.

Instead of answering the same question for the twentieth time, the owner can work on developing the company.

Good automation does not always remove humans from the process.

He often removes from the process everything that bothers a person.

Checklist: what to automate first?

First search:

• manual data rewriting,

• copying information between applications,

• translation of repetitive messages,

• reminders and follow-ups,

• classifying simple reports,

• repetitive answers,

• searching for data before performing a task,

• manual notifications,

• checking multiple boxes,

• preparing the same documents,

• tasks based on very clear rules.

Leave the man:

• difficult conversations,

• unusual complaints,

• negotiations,

• important financial decisions,

• building relationships,

• exceptions to standard rules,

• situations requiring empathy,

• high-risk decisions.

Automation should free up attention

Employee time is important.

But his attention is even more valuable.

Every time a person stops working to copy data, find a message, check another application, or recall a procedure, he or she loses a piece of concentration.

Therefore, the greatest effect of automation is not always visible in the table:

“We saved 23 seconds.”

Sometimes it looks different:

an employee can work peacefully on an important matter for an hour because the system takes care of the rest itself.

This is harder to measure.

But this is often what changes the way we work the most.

Summary

Don’t automate the most important things just because it’s technically possible.

Start with the most boring ones.

From rewriting.

From copying.

From searching.

From reminding.

From translation.

From classifying.

From repetition.

From opening another app just to see if something has appeared.

If the system can do the job faster and more predictably, let it.

And leave the man where his presence really changes something.

When deciding.

When talking.

When there’s a problem.

With a client who doesn’t fit into any ready-made scenario.

With an idea that no one had ever had before.

The best automation does not mean fewer people in the company.

It makes people less likely to behave like machines.

FAQ

What is best to automate in your company?

It is best to start with frequently recurring tasks based on clear rules: transcribing data, notifications, reminders, classification, ready-made answers, translations and transferring information between systems.

What should not be automated?

Processes related to important decisions, negotiations, difficult complaints, customer relations and situations requiring empathy and interpretation of unusual contexts require caution.

Should AI respond to customers on its own?

In simple, predictable situations this may make sense. In cases of unusual, emotional or important decisions, it is worth leaving the possibility of a human taking over the conversation quickly.

How to start automation in a small company?

For a few days, write down your most repetitive and annoying activities. Then choose those that take a lot of time in total and can be described by a simple rule.

Does automation always require AI?

NO. Many of the best automations do not use artificial intelligence. These can be simple integrations, templates, rules, statuses, forms and automatic notifications.

How to assess whether a process is suitable for automation?

Check the frequency, repeatability, predictability, time cost and consequences of error. The more frequent, simpler and less risky the task, the better the candidate.

What to do before automating the process?

First, check if the process is needed. Then simplify and standardize it. Only then is it worth automating it.