What we would tell a company starting from zero in January
Fourteen months of posts compressed into the order we would actually do things, for a Swiss SME that has done nothing yet.
Suppose a company of thirty people has done nothing with AI and wants to start in January. Here is the whole thing in order, based on what has actually worked in the companies we have watched.
Week one: name the person
One person decides which tools are used and what may go into them, with a spending limit above which they ask. Not necessarily the most technical person — whoever best understands how work actually flows through the company.
Everything else is blocked on this, which is why it is first.
Week two: find out what is already happening
Ask, without consequences, what people already use. The answer is never nothing. You are not looking for wrongdoing; you are looking for the tools you are already exposed to and did not know about.
Week three: one page, three sentences
Which tools are approved. Two or three concrete examples of what must never go into them, drawn from your business. Who to ask. Send it to everyone once.
Month two: pick one boring process
Not the impressive one. Something repetitive, where a mistake is cheap and somebody would notice it. Write down how it currently works before changing anything — that step is skipped most often and is where the value usually turns out to be.
Month three: measure something
Three numbers: what you spent, how many people used it, one process that changed. Write them down with the date. Repeat quarterly.
What is deliberately not on this list
Choosing a model. Comparing providers. Any compliance work. A strategy document. A budget beyond a few hundred francs a month.
None of those produces value in the first quarter, and companies that begin with them generally still have not started by the summer. The order above is unglamorous and it works, which after fourteen months of writing is the most useful thing we can say.