Which business processes should you automate first?

Those recurring tasks are a good starting point when you want to automate business processes. You can see where the work is, which data is needed and where things go wrong. But not every manual task is ready to be handed over.

The best first step is usually a small, clearly defined process that happens often enough and has a manageable set of exceptions. That lets you check whether the solution works before making a larger part of your business depend on it.

Find where information keeps getting copied

Follow an enquiry, order or internal request from start to finish. Where does someone move information without adding anything to it? Where do people pass on a status that's already recorded somewhere? And where does a colleague have to wait for someone to read an email?

That often reveals useful candidates: copying customer details, assigning a task, sending a reminder or preparing a document using information you already have.

For each candidate, note how often it happens, roughly how long it takes and what happens when someone makes a mistake. Use a normal working week as your reference. An irritating task that comes up once a year can attract lots of attention, while a small daily task adds up to far more work.

Are the rules clear enough?

A process is easier to automate if you can explain what starts it, which information it needs and what should happen next. For example: once an enquiry is complete, the colleague responsible gets a task with the necessary details.

But what counts as complete? And who's responsible if two teams are involved? If colleagues give different answers to those questions, you need to agree on the rules first. Otherwise, you're simply building that uncertainty into the software.

Work through the exceptions too. An unknown customer, missing data or an unusual order doesn't have to stop the entire automation. You can set those cases aside for a person to review. Just make sure someone sees them and knows what needs to happen.

Compare usefulness and consequences

A simple assessment helps you decide what to tackle first. You could give each candidate a low, medium or high score on four points:

• Repetition: how often does the process happen?
• Manual work: how much copying, checking or chasing does it involve?
• Clarity: are the rules and exceptions understood?
• Consequences: what happens if the automation makes the wrong choice?

A frequent administrative handover with clear rules can be a comfortable first step. An automated decision with serious consequences for a customer needs more checks and testing, even if it could save a lot of time.

A score like this helps the conversation; it isn't a calculation that automatically gives you the right answer. Consider who uses the process and who has time to test the new way of working too.

When does AI add something?

You don't automatically need AI to pass on a completed form. If the data and rules are fixed, a straightforward integration or rule-based automation may be enough.

AI can be useful when you want to organise free text, produce an initial summary or categorise messages, for example. Its output needs different checks from a fixed rule. Test a range of examples and agree on when a person takes over.

One possible first version lets AI make a suggestion that a colleague confirms. That helps you see where the system is useful and where it still falls short. You don't have to hand over the entire decision to benefit from it.

Make one workflow work from start to finish

You might choose the handover of a complete enquiry to the right colleague. Work through that route, including confirmation, error messages and a manual fallback. What if the receiving system is unavailable? Where can someone see that an enquiry has got stuck?

Ask the people who do the work to run through real situations. Check whether the automation creates extra work to put things right. Saving time at the start helps little if a colleague has to correct everything later.

At BORING, we look at the case to decide which tools fit. If an existing solution supports the process well, we consider it. If it doesn't exist, we can build the solution you need from scratch.

Measure the difference from the old process

Before starting, record how many manual steps there are, roughly how much time they take and which errors keep happening. After launch, compare that with the same kinds of work. That shows whether the process has actually become simpler.

Want to weigh the investment against time saved and costs? Our article on measuring marketing automation ROI explains the return calculation in more detail.

For now, start with one question for your team: which information do we keep entering somewhere else? Take one of those handovers and map out what really happens. You'll have a concrete process to improve, including the exceptions you might otherwise only discover during development.