Insight
What Should a Business Automate First?
Choose the first workflow by volume, operational risk, repeatability and available controls.
Where to start
Not the thing that annoys you most.
The task people want to automate first is almost always the one causing the most irritation. That is rarely the one that should go first, because irritating usually means messy, and messy is exactly what automation makes worse.
A good first automation has four properties: it happens often, it is done the same way every time, the rules can be written down, and being wrong is recoverable. Score any candidate task against those four and the right starting point usually picks itself.
High frequency matters because it determines payback. Consistency matters because a process performed three different ways by three people has no single version to automate, only a decision nobody has made yet. Writable rules matter because if you cannot explain the decision to a new joiner in a paragraph, you cannot hand it to a system either. And recoverability matters because your first automation is where you learn how this works, and learning is cheaper when mistakes are undoable.
The order that works
Stabilise, standardise, then automate.
This sequence is not a preference. It is the difference between an automation that survives and one that quietly produces wrong answers at scale.
1
Stabilise
Does the process produce the same result on a Tuesday as on a Friday, with a different person doing it? If not, that is the work. Automating variation just produces variation faster.
2
Standardise
Is there one agreed way, written down clearly enough that a new joiner could follow it? Writing it down is where most of the real disagreements surface, which is precisely why it is valuable.
3
Automate
Only now. By this point the build is quick and boring, because every decision it has to make has already been made by a person, on purpose.
The uncomfortable implication is that a good automation partner will sometimes tell you not to automate yet. A process that fails the first two gates needs process work, not software, and selling you software instead is how automation projects end up as expensive disappointments.
A practical first pass
Map one path from enquiry to money.
You do not need a full process audit to find the first candidate. You need one path, followed honestly from end to end.
Pick the journey that matters most commercially, usually from the moment an enquiry arrives to the moment it becomes revenue. Write down every step, every person who touches it, and every system it passes through. Then mark three things: where it waits, where information gets retyped from one place into another, and where things fall out of the process entirely.
Waiting is where customers lose patience. Retyping is where errors are born, because a person copying a value between two screens will eventually copy it wrong. And the things that fall out are usually invisible in your reporting, because a report can only count things that happened. Enquiries that were never answered do not appear anywhere, which is why the leak persists.
In most businesses the biggest single win is not making a step faster. It is making a leak visible for the first time.
Common questions
Straight answers.
What if everything feels equally urgent?
Then start with whichever is highest volume and most repetitive, not whichever is loudest. Volume gives you payback and repetition gives you a process stable enough to automate safely. The loudest problem is often a judgement-heavy one that should stay with a person.
Should we fix the process first or automate it first?
Fix it first, always. Automation applied to an unstable process runs the instability faster and with nobody watching. A person doing a messy job catches most of their own mistakes. A system doing a messy job catches none of them and does it hundreds of times before anyone notices.
How long before we see anything?
For a single well-chosen process, expect a working result inside a quarter and a measured number at the three-month mark. If there is nothing measurable by then, something is wrong with the choice of process and it should be diagnosed immediately rather than waiting.
Start with the workflow
See what can be automated.
Tell Astra where work slows down. We will map the system, its controls and the right human hand-offs.
Discuss Your Workflow