Superintelligence is the word of the fall, and for small businesses it mostly brings confusion. In two weeks, it made its way into a speech at the UN, a White House executive order, and the news you read over your morning coffee. A business owner then doesn’t know whether to change anything. The short answer: on Monday morning, you’ll open the same app as last Friday, and you should decide based on what it can do today.
The question I get most often in consultations right now has three words: shouldn’t we wait? I understand it. If you don’t want to buy something that will be outdated in a year, you wait. But with AI, you pay for waiting in a different currency than with a new phone.
What changes on Monday morning?
Nothing changes in the tools your business uses. Executive Order 14434 of September 29, 2026, requires US federal agencies to write “Super Intelligence” instead of “Artificial Intelligence” in their documents. But under Section 3 of the same text, the definition doesn’t change. The new name refers to the same technologies that US law has so far called artificial intelligence.
The word superintelligence now carries two meanings: a technical term for a machine smarter than people, and an official name for today’s AI. I go through both in detail in the article on what superintelligence is. For a small business’s decisions, it’s enough to know that a machine smarter than people doesn’t exist yet, and the official name is just a new label on today’s AI. And in Europe, nothing gets renamed: the AI Act (Regulation (EU) 2024/1689) still applies.
The assistant that drafted a reply to a customer for you on Friday will draft it exactly the same way on Monday. Someone in the news just calls it something else now.
What AI can really handle in a small business today
Today’s AI reliably handles repetitive chunks of work with text and data, as long as a person oversees it. In the small businesses I set up systems for, it’s most often this:
- answers to the same customer questions, over and over,
- meeting transcripts and summaries,
- sorting email and drafting replies for approval,
- invoices, quotes, and contracts from a template,
- translations and rewriting texts in a different tone.
Language model
A program that learned from an enormous amount of text and uses it to predict how an answer should continue. It’s behind ChatGPT, Claude, and most business assistants. It writes quickly and fluently, but it knows nothing for certain. That’s why its work gets checked.
Nothing on that list needs a programmer. I describe three inexpensive first steps in the article on AI for small businesses: a chatbot, one automation, and invoicing. What today’s AI can’t do is take over decisions. It doesn’t know which loyal customer to let off a late fee, and when it isn’t sure, it can be wrong in a confident tone.
Why waiting for superintelligence doesn’t pay
Waiting has a cost nobody invoices you for: every month you put it off is a month of manual work a machine could already be doing. Even researchers don’t know when a machine smarter than people will arrive. Their estimates differ by decades, and I go through them in the article on when superintelligence arrives.
The next-version trap
Putting off a decision in the hope that the next version of a tool will be so much better that today’s isn’t worth it. New AI versions are announced constantly, so the trap never springs on its own. A business can stand in it for years and feel the whole time that it’s deciding sensibly.
And repetitive work doesn’t change with a smarter tool. Today’s assistant answers “how much is shipping to Brno?” just as well as the one in the headlines. The sooner you hand off an answer like that, the sooner a person stops writing it.
The Monday hour
On Monday morning, open last week’s sent mail. Count the replies you wrote almost the same way: about price, about timing, about the process. Where you count more than five, you’ve found the first job for today’s AI. You don’t need superintelligence for it.
Three decisions for this month
For a small business, the big words boil down to three small decisions. None of them needs new technology.
Pick one routine and give it a month. One, not five, and one where you can say how much time it eats up today. I describe how to plan that month week by week in the article on the first thirty days with AI. If you’re unsure which one to let go of first, the order in which work gets handed to a system will help.
Don’t pay extra for a word. If a vendor offers you the same tool with a superintelligence label at a higher price, ask what extra work you’ll get. Executive Order 14434 changed nothing technically, and by its own text it creates no enforceable rights.
Check where AI writes to the outside world on your behalf. Under the EU AI Act, rules for labeling AI-generated content apply from August 2, 2026. Go through the texts, images, and chatbot replies created this way, and check whether the rules apply to you. Unlike the new American word, these rules actually apply in Europe.
What you can safely ignore
For this month’s decisions, you can put most superintelligence news out of your mind. It may matter for politics and big AI companies, but it doesn’t reach your inbox or your invoicing.
- Arrival dates. Whether you hear five years or forty, it won’t affect which routine you pick first.
- Company renamings. On October 4, Elon Musk wrote that he would rename his AI company SpaceXAI to SpaceXSI. Not a single button in your tools will move because of it.
- Big companies’ agreements with the White House. According to the Guardian, the September 29 document is voluntary and unenforced. For a business in the EU, the European act is what’s binding.
- Rewriting AI as SI. Customers and European rules alike still talk about artificial intelligence.
Who needs more than a calm approach
Companies that develop or sell AI themselves need more than this calm approach. For them, the big words are business reality: customers ask about them, and competitors put them in their offers. They have to follow them and be able to explain what their product can and can’t do.
The second exception is heavily regulated fields, such as healthcare or finance. There, what decides about AI is what a tool is allowed to do rather than what it can do, and a legal review has to come before any month-long trial. And if a business has no repetitive work, it has nothing to hand to a machine, whatever AI happens to be called.
The whole Monday in a nutshell
- Nothing changes in your tools on Monday morning; Executive Order 14434 renamed AI only in US government documents.
- Superintelligence as a machine smarter than people doesn’t exist yet, and holding off decisions until it does makes no sense.
- Today’s AI can handle repetitive work with text and data; decisions stay with people.
- This month, pick one routine, don’t pay for a new word, and check where AI writes to the outside world on your behalf.
- Leave the dates, renamings, and voluntary agreements to the news.
So Monday morning can look perfectly ordinary: coffee, an open inbox, and five almost identical replies from last week. Those five replies are a better place to start than any headline. If you’d like, I’ll pick the first routine with you and bring AI into your business, all the way to the point where it runs on its own. Otherwise, take your most common reply and hand it to the assistant you already have, this week.
