Choose one task with an owner
A task that comes back often and follows rules someone can explain.
- For example: putting an enquiry from your site into your customer list.
- Or sending a colleague a message as soon as a file is complete.
- One person owns the outcome and sorts it out when something goes wrong.
- Start with something small whose effect you can see, not a whole department.
When nobody can explain it
Does nobody know why a step exists, or do the rules change every week? Then simplify the work first and write it down. An automation only repeats a messy process faster.
Write the task on one page
Not only how it goes right, but also what happens when something is missing.
- Start: what sets the task off, and how do you spot a repeated request?
- Details: which fields are required, and which system is right?
- Decisions: which rules are fixed, and where does a person decide?
- Done: what shows it worked, and who hears about it when it does not?
Access and examples
Also write down which programs can be connected and who can grant access. Show a supplier real examples, but take out names and other sensitive details.
A small example
An invented example to show the idea, not a client project.
- An installation company (fictional) receives enquiries through the form on its site.
- The automation checks that the required fields are filled in and looks for an existing customer.
- It puts a draft in the customer list and gives the right colleague a task to check it.
- If something is missing, the enquiry goes to a list for checking. Nothing is guessed.
- Sending a quote or naming a price stays a decision for a person.
Plan for failure first
Connections sometimes drop and accounts lose access. Settle that before you scale up.
- How does the owner see that a task did not go through?
- What may run again without anything happening twice?
- How do you pause the automation and do the work by hand?
- Keep enough to look into a fault, but no personal data you do not need.
When AI reads or writes
Let AI help with sorting or a first draft, and test it on real examples. An answer that looks right is not proof that it is right. Important decisions stay with a person.
Measure a trial against your starting point
Hoped-for time savings are not a promised return. Only a trial shows what it gives you.
- Write down now how often the task comes up, how long it takes and which mistakes happen.
- Run a short trial with one task.
- Then compare: what got finished, what went to a person, and how much upkeep did it take?
Checklist: are you ready?
Does the task come back every week? Can someone explain the steps in plain words? Is there one owner who decides and fixes things? Do you know what should happen when details are missing? Do you know how long the task takes now? Yes to all of them? Then the task is ready to automate.
What does the first step cost?
Start small. Add the next task only when the first one works well.
- One task that runs by itself: One automation, €1,950€2,359.50 one-off plus €99€119.79 a month, excl. 21% VAT.
- The same customer questions again and again: AI assistant, €2,450€2,964.50 one-off plus €149€180.29 a month, excl. 21% VAT. It says it is an AI and hands doubts to a person.
- Quotes, planning and admin together: Automation package, €4,950€5,989.50 one-off plus €299€361.79 a month, excl. 21% VAT.
A test version you can view at every step.
Subscriptions can be cancelled monthly after any stated minimum term.
More about automation