Where do you start with AI in your business?

You read everywhere that you should be doing something with AI. Your accountant says so, your trade association says so, and on LinkedIn there is someone who claims to have automated half their admin away with a chatbot. Meanwhile you have no idea where you should begin, and no appetite for wasting three months on a trial that delivers nothing.

That is a fair reflex. Most failed AI projects at small and medium-sized businesses (SMEs) start with the wrong question: which tool shall we pick? That question comes much later. You start with the work.

Start with the work that comes back every week

AI is good at work that happens often, follows a recognisable pattern, and where it is no problem if a person still takes a quick look. That is almost never the exciting work. It is the dull work that costs your business half an hour every day and that nobody writes down because it “comes with the job”.

Walk around with a notepad for a week and note where people sigh. In practice, the same kinds of work tend to come up:

• Sorting incoming email and forwarding it to the right person, at an installation company with a shared info mailbox.
• Retyping packing slips, price lists and order confirmations from suppliers into the system, at a wholesaler.
• Answering the same fifteen questions customers ask about delivery times, warranty and opening hours.
• Typing timesheets and job sheets from paper into the system, usually on Friday afternoon.
• Writing out quotes that are ninety percent the same as the previous one.
• Preparing receipts and invoices for the books, at an accounting firm for dozens of clients at once.

Four questions that identify a good first project

Once you have a few candidates, run them past these four questions. If one is left that scores well on all four, you have your first project.

1. Does this happen often? Something that happens once a quarter is rarely worth automating, however annoying it is.
2. Can you explain how it works now? If nobody in the business can write down the rules, a system cannot learn them either. Then the process itself is the problem, not the lack of AI.
3. What happens when it goes wrong? Choose something where a mistake is visible and fixable. A wrongly sorted email gets noticed. So does a wrongly sent invoice, but that one is more expensive.
4. Is there someone who wants it? Without a colleague who puts time into it and benefits from it, every project gets bogged down. Technology is rarely the problem; ownership is.

Starting small means really small

A good first project is finished in weeks, not quarters. It touches one department, at most two systems, and at the end you can answer a simple question: has this work become easier or not. With us, that usually starts with a half-day core session (four hours), in which we follow one process from start to finish at your office or in your workshop and count the actions. That costs €596 excluding VAT and travel expenses. After that you know whether there is something to gain, even if the answer is no.

Build work after that is charged per hour: €110 for straightforward work such as setting up integrations and configuring processes, €165 for complex development. We say up front what we think will be needed and where the uncertainty lies.

What disappoints at the start

Two things almost always disappoint. The first is your data. A system that has to read in orders cannot guess that three product numbers with a typo are the same product. Cleaning that up is dull manual work up front, it never appears in any demo, and it is often the biggest part of the first weeks.

The second is the time saving itself. Saving half an hour a day sounds like half a day a week, but it does not work that way. Those minutes are scattered and disappear into the day. The real gain is usually somewhere else: fewer mistakes, less dependence on that one colleague who knows everything, and work that is simply done the same day instead of on Friday.

If you cannot explain how something works now, automating is not the next step. Writing it down is.

When you need us and when you do not

If you mainly want to answer your email faster, take meeting notes or have texts written, you do not need an agency. A subscription to an AI assistant and an afternoon of practice with your team will get you further than we can. The same applies if your software package already has an AI feature you are not using yet: switch that on first.

We are useful as soon as systems need to be connected that do not talk to each other by themselves, as soon as custom-built work is needed that no existing tool provides, or as soon as you get stuck on the question of which process to tackle first. And if after the core session we think you are better off with a standard package, we say so. That is a short conversation, but an honest one.