AI business automation: where should a company start?
Start with one task your team repeats every day, with a clear input and a clear result: supplier quotes that become a comparison sheet, invoices that become entries in your system, WhatsApp messages that become logged orders. Automate that task from start to finish, measure the time it gives back, then move to the next one. Starting with a general tool, or a plan for the whole company, takes months before anyone feels a difference.
- PDF from A
- Excel sheet from B
- WhatsApp photo from C
| Item | A | B | C |
|---|---|---|---|
| Printer paper (box) | 24 | 27 | 25 |
| Packing boxes (×100) | 62 | 58 | 60 |
| Cleaning supplies (case) | 41 | 44 | 38 |
| 24-inch monitor | 139 | 132 | 135 |
Why one tool is not enough
Most searches start with a tool: convert a PDF to Excel, a WhatsApp bot, invoice-reading software. Each of these does one step. Inside a company the work is a chain: a file arrives by email or WhatsApp, someone reads it, checks it against something, someone approves it, and then it is entered into the system.
If the tool only does the reading, your employee still does every other step by hand, and the time saved is small. Automating the work means running the whole chain, with a person reviewing at the right point.
How to pick the first task
A good first task ticks most of these:
- It repeats every day or every week, not once a year.
- It has a clear input: a file, a photo, a message.
- It has a result you can check: a sheet, an entry, an order.
- It takes hours you can count today.
- A mistake in it can be caught before it costs money.
What it looks like, step by step
- A file or message arrives
- It is readWhere a single tool stops
- It is checked
- A person reviews unclear cases
- Entered in your system
- We sit with the person who does the work today and write down the steps as they really are, not as they are supposed to be.
- We collect real examples: the files or messages from the past few weeks.
- We build the whole chain: receiving, reading, checking, then entering the result in Excel or your system.
- We add a review point: unclear cases go to a person, the rest goes through.
- We run it on your real work and measure the time before and after.
What it saves
- The hours spent copying from one file into another.
- Typing mistakes that only show up after the money is paid.
- Work stopping when the one person who knows how to do it is away.
How much that is worth depends on your hours and salaries. The calculator on the offers page works it out from your own numbers.
The limits of AI today
- It makes mistakes sometimes. That is why we never run a workflow without a review step for unclear cases.
- It does not replace judgement: choosing a key supplier or approving an exception stays with your people.
- The result is only as good as the input. A dark photo or illegible handwriting needs a person.
- If every employee does the task differently, automation exposes that first, and you may need to agree on one way of working before it runs.
Related guides
Common questions
What companies usually ask before they start.
Have a task that repeats every day?
We build the whole workflow around how your team works today, in a two-week Sprint, at a fixed price we agree on before we start. You don't need a meeting to start: tell us about the task and we reply within two working days.