Where do you start with automation?

Most automation projects start with the wrong task. The most visible one gets picked, rather than the one that hurts most.

Below is a simple method for finding the work that is genuinely worth automating.

Measure first: a three-column table

For one week, note your team's repetitive tasks in three columns:

| Task | Frequency | Duration | |---|---|---| | Preparing the daily sales report | Daily | 25 min | | Entering order details into accounting | ~15 per day | 3 min | | Preparing a quote document | 8 per week | 20 min |

Frequency multiplied by duration gives you the annual hours. Automation decisions made without this table almost always go to the wrong task.

Then filter: four criteria

A high hour count is not enough on its own. The candidate also has to meet these four:

Are the rules clear? If your answer is "it depends on the situation," you first need to write down what that decision depends on. A rule that cannot be written cannot be automated.

Is the input structured? Pulling data from a form with fixed fields is easy; extracting it from free text is hard, though possible.

Is an error expensive? A wrong accounting entry and a mislabelled email are not the same thing. Risky work keeps a human in the loop.

Is it likely to change? Automating a process that will be completely different in six months is wasted effort.

The three levels of automation

Automation is not one thing. The same process can be automated at three different levels, and their costs differ enormously.

Level 1: notification. The work is still done by hand, but the system tells you when it needs doing. This is the cheapest level and it solves surprisingly many problems.

Level 2: preparation. The system prepares the work and a person checks and approves it. Drafting a quote or staging invoice data belongs here. This is the right level for risky processes.

Level 3: fully automatic. No human involvement. Reserve it for work where the rules are entirely clear and errors are cheap.

Most businesses want to jump straight to level three. Starting at level one delivers results far faster and reveals how the process actually works.

Three things to avoid in a first project

Starting with the biggest process. The first project is a learning project. Start with something small and clearly measurable; scale up once there is confidence.

Switching off the old method immediately. Run the new flow alongside the old one for a while. If the results match, retire the old one. That gives you a reversible transition.

Not designing for failure. Automation will stop one day. A connection will drop or a service will not respond. What happens at that moment has to be designed up front: an alert has to go out and a manual path has to stay open. Work that disappears silently is the biggest risk in automation.

What changes after automation

An often-overlooked point: automation does not remove the work, it changes its nature.

The person who used to enter data no longer enters data, but they do check that the system is working and handle the exceptions. That new job also has to be assigned to someone. If it is not, the automation ends up unowned and gets abandoned at the first serious error.

That is why we answer this question at the start of every automation project: who looks after this when it breaks?

Choosing tools

Off-the-shelf automation tools are fast and cheap for simple connections. Their limits are equally clear: as the number of steps grows, as conditional logic enters, and as volume rises, they become both harder to manage and more expensive per month.

A rough dividing line: if you are connecting two systems with a simple rule, an off-the-shelf tool is enough. If your own business rules are involved and volume is high, custom development is cheaper and more stable over time.

The technical side of that distinction is covered in what is API integration.

Automation is not always the answer

Sometimes the right answer is to eliminate the process rather than automate it.

An example we run into often: "We prepare this report by hand every day, let's automate it." When we ask, it turns out nobody reads the report. Removing it saves the same time and costs nothing.

Before starting any automation project, ask of every step: is this step actually necessary?

What to expect

A well-chosen first automation project usually delivers:

• The repetitive work leaving the team's day.
• Fewer errors from manual data entry.
• Data staying in one place and staying current.

These are concrete gains, but they cannot be expressed as a percentage, because they come out differently in every business. The table you measured will give you your own number.

---

If you have your measurement table, we can look at it together. See business process automation.