The processes most worth automating first

The first automation an organization builds does two jobs: it takes work off people, and it decides what everybody believes automation is worth. Choosing it badly is expensive in a way that does not show up in the project report.

What makes a good first candidate

High volume, low variation. Daily beats monthly, and hundreds beats dozens. Frequency is where the value accumulates, and it is also what makes the exceptions worth designing for.

A clear right answer. Work where correctness is checkable. This lets you measure whether it is working without a person reviewing every output.

Stable inputs. The same fields, arriving in roughly the same way, from a system nobody is about to replace.

A visible current cost. Someone can say how long it takes now and who does it. Without that baseline, the result is unarguable in either direction.

A willing owner. A business manager who wants this and will help work through the exceptions. Automation imposed on an unwilling team fails on adoption regardless of how well it is built.

The four that look attractive and are not

A process about to change. Automating something that a system replacement or a policy change will alter within the year is work done twice.

A process nobody can describe the same way twice. If three people give three accounts of how it works, you do not have a process — you have a set of habits. Map it first; sometimes the mapping is the whole benefit.

The most painful process. Pain usually correlates with complexity and exception volume, not with value. It is a good second or third project, once the platform and the patterns exist.

Something that only one person understands. Tempting, because the risk is obvious. Difficult, because there is no documented process to automate and no second person to validate against.

Where to look

Four places, in order of how often they pay:

• Data moving between systems by hand. Someone exports, reformats and imports. Common, boring, and immediately valuable.
• Reading documents to get a handful of fields out. Invoices, tickets, statements, forms.
• Routine checks and reconciliations. Comparing two sources and flagging differences.
• Status chasing. Producing the same report or the same set of reminders on a schedule.

Build the first one properly

The first automation sets the pattern. Built as software — with reusable components, configuration separated from logic, error handling and logging — the second costs a fraction of the first.

Recorded quickly to prove a point, it becomes something nobody can extend and everybody is afraid to change. The pilot that worked and then went nowhere is almost always this.