Which processes can be automated in a company
Not every process is worth automating, but far more are than people think. Concrete criteria, examples by area, and how to decide where to start.
June 25, 2026 · Diego Osvaldo Díaz · 3 min read · Updated on
Any process can be automated if it meets three conditions: it happens often, it follows rules someone can explain, and it works with information held in some system or document. What isn't worth automating is anything that changes unpredictably or depends on professional judgement in every case.
The three criteria, in detail
Frequency
A task done twenty times a day is a candidate; one done twice a year almost never is. The effort of automating is paid back by repetition, not by complexity.
Explainable rules
If someone can describe how they decide what to do in each situation, that logic can be moved into a system. If the answer is “it depends, you just know”, that criterion has to be made explicit first — and sometimes that exercise alone improves the process.
Accessible information
The system needs to be able to query the data. If the information lives only in someone's head or in loose notes, there is a tidying step first.
Examples by area
Administration and accounting
- Receiving and reading invoices and receipts arriving by email or WhatsApp.
- Entering data into the accounting system without manual typing.
- Tracking due dates with advance alerts.
- Reconciling bank movements against internal records.
- Automatic filing of documentation by client and period.
Sales
- Capturing inquiries from web, WhatsApp, and email into a single funnel.
- Automatic CRM logging, without relying on data-entry discipline.
- Follow-up reminders per opportunity.
- Generating proposals from templates.
Customer service
- Answering frequent questions around the clock.
- Classifying inquiries by topic and urgency.
- Routing to the right area with the full history.
- Tracking open cases through to resolution.
Operations
- Syncing data between systems that don't talk to each other.
- Recurring reports currently assembled by hand.
- Stock control and restocking alerts.
- Traceability of every stage of an internal workflow.
What is NOT worth automating
This list matters as much as the previous one, and it rarely appears in a sales proposal.
- Processes that change unpredictably: you automate and have to rebuild every month.
- Decisions requiring professional judgement or negotiation in each case.
- Low-volume tasks: the effort is never recovered.
- Processes nobody can explain: they have to be understood first.
- Workflows about to change: better to wait until they settle.
How to prioritise when there are several candidates
There are almost always more automatable processes than capacity to tackle them at once. The simplest way to order them is impact against effort:
| Criterion | Low effort | High effort |
|---|---|---|
| High impact | Start here | Plan as a project |
| Low impact | Do if there's spare time | Drop |
The high-impact, low-effort quadrant is the one to attack first, and not only for the return: it's the one that builds the internal confidence needed to take on the bigger projects.
How to spot them in your operation
There's a practical way to find them without hiring anyone. For one week, ask each area to note down the tasks they did more than five times. By the end of the week you'll have the candidate list, ordered by real frequency rather than perception.
Then, for each one: can someone explain the rule? Is the information in a system? If both answers are yes, it's automatable.
A process can be automated if it's frequent, follows explainable rules, and works with accessible information. It isn't worth automating the unpredictable, the low-volume, or what nobody can explain. To prioritise, cross impact against effort and start with high impact and low effort.
Frequently asked questions
Which processes should be automated first?
High-impact, low-effort ones: frequent tasks with clear rules and available data. They tend to sit in administration and customer service, where repetitive volume is highest.
Are there processes that shouldn't be automated?
Yes: those that change unpredictably, those requiring professional judgement in each case, low-volume ones, and those nobody can explain. In the last case, understanding them is the prior work.
Do we have to change the systems we use?
Generally no. Most automations are built by connecting existing systems. If one doesn't allow integration, alternatives are assessed during the mapping stage.
How do I identify candidates in my company?
For one week, have each area note the tasks they repeated more than five times. That list, ordered by real frequency, is the best starting point.
Does automation eliminate jobs?
It automates tasks, not entire roles. Typically the team stops copying data and moves to supervising exceptions and higher-value work.