Which business processes to automate and which you should eliminate first

Before automating, review what you're trying to solve

A company can invest in software, integrate systems and automate tasks without seeing any real improvement in its operations. The reason is simple: technology speeds up a procedure, but it doesn't, by itself, fix the problems that created it.

Think of an organization that spends several hours a week preparing a report. To reduce the workload, it automates data collection and document generation, and what used to take four hours now takes fifteen minutes. It looks like a good investment, until it becomes clear that nobody asked whether the report was still needed, who used it or what decisions were based on it. If it no longer served any purpose, the company paid to automate something it could have eliminated. That's why, before asking how to automate a process, it's worth asking why it exists.

Which processes are good candidates for automation

The best opportunities tend to lie in frequent, repetitive activities with clear rules, where the necessary information is available and exceptions can be handled without complications. Transferring data between two systems every day, generating standardized documents, updating certain records or sending notifications triggered by defined events are typical examples. By contrast, a task that changes constantly, has many exceptions or requires complex decisions tends to be harder and more expensive to automate.

To evaluate an opportunity, it helps to look at five factors together:

• Frequency: how often the activity is performed.
• Time: how much effort each run requires.
• Rules: how predictable the procedure is.
• Errors: what the consequences of an incorrect run would be.
• Cost: how much it takes to implement and maintain the solution.

None of them is enough on its own. A task may be repeated hundreds of times, but if each run takes a few seconds and automation requires a large investment, the benefit may be small. Conversely, an infrequent procedure can justify a technology solution if it consumes many hours or leads to costly mistakes.

Three examples across different operations

Retail: stock queries between branches

In a retail chain, a salesperson needs to know whether a product is available at another branch. They send a WhatsApp message, wait for a reply and sometimes have to ask for a second check. The first idea might be to implement an automated assistant to answer stock queries, but first it's worth understanding why the salesperson can't see that information directly.

The system may already have a lookup function nobody uses, permissions may be misconfigured or the data may not be updated properly. In that case, improving access to the information and its quality solves the problem without building anything new. If, on the other hand, the information is spread across systems that don't talk to each other, an integration may make sense. The decision depends on the cause of the problem, not on whether it's technically possible to automate it.

Administration: manually compiled reports

An administrative team prepares a weekly report that combines data from several spreadsheets, which involves copying information, checking figures and organizing results. Before automating it, the first step is to identify which parts of the report are actually used. Management may only need five indicators, while the team keeps producing a twenty-page document out of habit. Once the content has been reduced, a tool the company already has may be enough to consolidate the data and update those indicators. The biggest improvement comes from narrowing the scope of the work, and automation takes care of what remains necessary.

Logistics: duplicate data between warehouse and admin

At a distribution company, warehouse staff record order picking in a spreadsheet, and the admin team later re-enters part of that data into the sales system. The procedure takes time and multiplies the chances of error. Automating the transfer is one option, but first it's worth analyzing whether both records are needed and what information each team requires. If a single source of data can be established, one of the two tasks disappears; if both systems need to coexist, an integration avoids the double entry. Sometimes the best solution isn't automating two steps, but eliminating one.

Eliminate, simplify or automate

When an activity consumes too much capacity, there are three possible paths. Eliminating it makes sense when it no longer serves a necessary purpose or its result can be obtained another way. Simplifying it means keeping it but reducing its complexity, whether by removing redundant steps or approvals, reorganizing responsibilities or improving access to information. Automating it means using technology to carry out, fully or partially, something that is still needed.

The order matters: if a task can be eliminated, it doesn't need automation, and if it can be significantly simplified, the procedure should be reviewed before it's moved into a tool. Some activities also depend on a person's judgment; in those cases, technology can help prepare information or reduce peripheral tasks without replacing the decision.

Buy software, integrate tools or build your own solution

Once an opportunity has been identified, another decision comes up: which technology to use. A new system isn't always necessary. Many companies already have tools that can configure workflows, generate reports, set rules or connect applications, and it's worth reviewing those capabilities before investing.

An integration is usually enough when the problem is moving information between tools, and an off-the-shelf solution may be the right choice if it addresses a common need in the industry. Custom development makes more sense when there are specific requirements that available alternatives don't cover and the expected benefit justifies both the investment and the ongoing maintenance. AI tools can also help classify documents, extract data or prepare drafts, provided there are appropriate controls over the quality of the results and the data being used. In every case, the technology choice should respond to the operational problem, not to an interest in adopting a particular tool.

How to calculate whether an automation is worth it

As an illustrative example, take a manual task that takes twelve minutes and is repeated 300 times a month. The calculation (12 × 300 ÷ 60) comes to 60 hours a month of capacity consumed. If automation reduces manual involvement to two minutes per run, the task would take up ten hours, freeing around 50 hours a month.

That figure alone isn't enough to approve the investment, though. Implementation, maintenance, supervision, exceptions and potential errors also need to be factored in. And recovering 50 hours doesn't mean saving 50 hours of wages: the financial benefit depends on how that capacity is used, whether to handle more transactions, reduce backlogs, avoid overtime or absorb growth.

When automation isn't worth it

There are situations where keeping an activity manual is the most sensible option: when the task is performed rarely, changes constantly or requires decisions that are hard to translate into rules, when the data isn't reliable, when maintenance costs more than the benefit, or when an error could have serious consequences.

In other cases, the problem lies not in how the task is carried out but in how the work is organized. Automating an unnecessary approval speeds up the response, but it keeps a step that perhaps shouldn't exist at all.

The best automation starts before choosing the technology

Automation can help a company handle more transactions, make fewer errors and spend less time on repetitive tasks, but those results depend on choosing the right work to automate. The starting point is understanding how the process works, which activities are necessary and where capacity is being consumed that could be recovered.

At DeepWork, we analyze operations and processes to identify what is worth eliminating, simplifying or automating, prioritizing the changes with the greatest impact. The DeepWork Assessment is the first step to understanding what problem your company needs to solve before adopting new technology.

Learn about the DeepWork Assessment