The world just renamed AI. Whatever you call it, it is already working for you.

The short version

• On 29 September 2026 President Trump signed an executive order directing US federal agencies to use “super intelligence” and “SI” instead of “artificial intelligence” and “AI” in official communications, websites, reports and policy files. Politico and the BBC both reported it.
• Read that as a verdict, not a vocabulary lesson. No one renames a technology that failed. You do not stand up at the United Nations and rebrand a thing no one is using.
• The industry moved with it. Elon Musk pledged to rename SpaceXAI to SpaceXSI, Salesforce renamed AIForce to SIForce, Satya Nadella posted about “Super Intelligence,” and the chief executives of Google, OpenAI, Meta, Anthropic and Nvidia signed the White House Accord on Super Intelligence.
• Be honest about the word. Experts cited by PolitiFact point out that “superintelligence” already meant a hypothetical system better than humans at everything, and no system has achieved that. The name overshoots. What is underneath it is still real, and it still works.
• The real story is in McKinsey’s August 2026 survey: 40% of firms above $1bn revenue report scaling AI agents, up from 27%. Smaller firms stayed flat at 22%. The gap is growing, and it has nothing to do with vocabulary.

On 29 September 2026 the United States government changed what it calls the technology everyone has been arguing about. An executive order directs federal agencies to use “super intelligence” and “SI” in place of “artificial intelligence” and “AI” in official communications, websites, reports and policy files, to the extent permitted by law. Politico and the BBC reported the signing. Nine days later, on 8 October, a Truth Social post called anyone still saying artificial intelligence “THE ENEMY!”

There has been a lot of eye-rolling about this. Here is a different reading, and it is the one that matters to anyone running a business.

No one renames something that does not matter. Governments do not hold signing ceremonies for technologies that fizzled. Firms do not rewrite their product names, their websites and their marketing in a fortnight for a trend they expect to die. Whatever you think of the wording, the last three weeks are a loud, public admission from the most powerful institutions in the world that this technology arrived, that it works, and that it is now part of how things get done.

That is the good news hiding inside an argument about labels. The arguing is over who owns the word. The opportunity was never in the word.

What happened, in order

The sequence is worth knowing, because the coverage has collapsed it into a single headline. On 29 September 2026 came the order itself, reported by Politico and the BBC. Then came the speech at the United Nations General Assembly in New York, he said the United States and, he hoped, the world would switch their documents to the more accurate term super intelligence rather than artificial intelligence. So it’s super intelligence… and he called it the new world of super intelligence, or SI.

On 8 October 2026 the Truth Social post did most of the spreading: “The White House considers anyone that uses the term, Artificial Intelligence, as opposed to the highly accepted new and more accurate term, Super Intelligence, THE ENEMY!”

Then the machinery moved. The White House updated its AI website at ai.gov, removing AI references and using SI in their place. Elon Musk pledged to rename SpaceXAI to SpaceXSI. Microsoft chief executive Satya Nadella referred to “Super Intelligence” in a post. Salesforce renamed AIForce to SIForce. And the chief executives of Google, OpenAI, Meta, Anthropic and Nvidia signed the White House Accord on Super Intelligence.

Sit with that list for a second. These are not research labs hedging a bet. These are the firms that build and sell this technology, and within days of a government changing its vocabulary, they changed theirs too. Firms follow value. They moved because being seen to be at the front of this is worth more than being seen to be cautious about the word.

Why a rename is not a small thing

Words like this are expensive to change. Every manual, every website, every policy document, every product name has to be touched. No one pays that cost for a hobby.

So the plain reading of October 2026 is this: the argument about whether this technology is real has finished, and the people who were slowest to accept that are now the loudest about naming it. The debate has moved on from does it work to what do we call it, and that second question is the good kind of problem to have.

For a business owner, that shift is permission. You no longer have to justify experimenting. You do not have to defend the idea to a sceptical board or a wary ops manager. It is settled: this is a technology with a federal government behind it and every major technology company signed up to it. What is left is not belief. It is implementation.

Whatever you call it, the question is no longer whether this technology is real. Nobody argues about the name of something that does not work. The question is what your system knows, and whether it finishes the job without anyone chasing it.

One honest caveat about the word

Being in favour of this technology does not mean pretending it is more than it is. There is a fair criticism of the rename, and it deserves a hearing.

“Superintelligence” already had a job. It long meant a hypothetical future system that exceeds human capability across the board, not better at one task, better at everything at once. PolitiFact reported experts cautioning that the rename is misleading for this reason, and noted the detail underneath it: no system has achieved that. Not one.

So the word written into official US government files describes something that does not exist yet, while the technology that genuinely does exist, useful, uneven, occasionally wrong in confident ways, has had its old name retired by decree. That is worth saying out loud, because overselling is how trust gets burned.

But notice what the criticism is not. No one is saying the technology does not work. The complaint is that the new name flatters it. That is a much smaller problem than the one we were all having two years ago, which was whether any of it would ever amount to anything.

Three of those four are aspirations. One describes work you can hand over today and measure tomorrow, and it is the one whose name changed.

The proof is not in the name

Here is the number worth more attention than any headline from the last fortnight. McKinsey’s State of AI Global Survey, published 25 August 2026 with 1,719 respondents across 97 countries, found that 40% of firms above $1bn in revenue report scaling AI agents, up from 27% a year earlier. Smaller firms were flat at 22%.

Two things are true in those numbers. The first is how fast the large firms moved: thirteen percentage points in a year. That is not hype. That is thousands of real workflows handed over, measured, and kept. The second is the gap. Big firms are pulling away from smaller ones, and the distance is widening.

Read together with this month’s renaming, a picture forms that has nothing to do with labels. The institutions with the most to lose, governments signing orders, chief executives signing accords, billion-dollar firms renaming products, are behaving like people who already know the answer. And the ones quietly getting on with it are already thirteen points ahead.

The chance in front of a smaller business right now is that the gap is made of ordinary things, not secrets. It is not a better model. Everyone can buy the same models. It is three unglamorous questions, answered before anything is switched on:

• What does it know? A system that cannot see your order history, your policies or your client emails is a clever stranger. Connecting it to what your business knows is the work, and it is the part that makes it useful.
• What is it allowed to do? Every workflow needs a line. Below the line it acts alone. Above the line a person decides. Without that line, teams either freeze or get burned, and the project quietly stops.
• Does it finish? Not “does it answer.” Does the work reach a conclusion without someone chasing it? This is where most attempts fail, silently, in month two.

None of these three care whether you call it AI, SI, or whatever is decided next year. They are questions about wiring, permission and follow-through. And they are all answerable this month, in a business of any size.

What this actually feels like

It is easy to stay at the level of presidents and accords, so bring it down to a Tuesday.

The invoicing that gets chased without you asking. The client email answered before you have sat down. The diary that reorganises itself around the work that matters. The report that exists on Monday morning because someone, something, built it overnight.

That is what the last three weeks of headlines are about. Not the acronym. The fact that this is normal enough now to be renamed, regulated and argued about by governments. Ten years ago most of it was science fiction. This month it got a federal vocabulary.

The part no order can deliver is the part that matters most to an owner: the hours back. Not because the software is clever, but because it knows your business, has been told what it may decide, and finishes. Time back with your family on a weeknight. A question answered at midnight because someone was curious. A life with less grinding admin in it. That is the whole point of all of it, and no name change touches it.

Where Twinthos comes in

Twinthos answers those three questions than arguing about names. A digital employee is connected to the email, diary and accounting systems your business already runs on, taught the rules that live in your colleagues’ heads, and given a written Authority Line per workflow: below it, it acts on its own; above it, a person decides, with a record of what happened and why.

That line is the whole product. It is what turns software that might be right into someone you can leave alone, and it is why the value was never in the model. Models are a commodity and get better and cheaper every quarter. What does not commoditise is knowing what your business knows, being told what it may do, and finishing.

Digital employees free humans to be human. The point was never to replace your team’s judgment; it was to take the volume-shaped work, the chasing, confirming, routing and entering, so the people you pay to think can spend their week thinking. Whatever the technology is called next year, that deal stays the same, and it is the best deal on offer.

If you want to know which parts of your week are volume-shaped, measured in hours in your actual business, the £999 Operations Audit is where that starts. You keep the diagnosis whether or not you go further.

Next step

Start with the £999 Operations AuditBook a discovery call

Twinthos is a managed digital employee service, not software you install. It works across the email, diary and accounting systems your business already runs. Below the Authority Line it acts on its own; above the line a person decides.

Common questions

What does SI stand for?

SI stands for “super intelligence.” From 29 September 2026 an executive order directs US federal agencies to use it in place of AI in official communications, websites, reports and policy documents. Politico and the BBC reported the signing.

Did the US government officially rename AI to SI?

It changed how federal agencies refer to the technology. The executive order was signed on 29 September 2026 and ai.gov was updated to remove AI references in favour of SI. The direction covers government documents and communications, not how private companies or individuals speak.

What did Trump actually say about the term change?

At the UN General Assembly in New York he said the United States and, he hoped, the world would switch their documents to the more accurate term super intelligence rather than artificial intelligence. So it’s super intelligence… and he called it the new world of super intelligence, or SI. On 8 October 2026 he wrote on Truth Social that the White House considers anyone using the term artificial intelligence “THE ENEMY!”

Is renaming AI to SI the same as achieving superintelligence?

No. Superintelligence has long meant a hypothetical system exceeding human capability across the board, and experts cited by PolitiFact caution that no system has reached that level. The rename says something flattering about the technology. It does not change what any system can do today, which is already a great deal.

Have companies adopted the new term?

Several have. Elon Musk pledged to rename SpaceXAI to SpaceXSI, Satya Nadella referred to “Super Intelligence” in a post, Salesforce renamed AIForce to SIForce, and the chief executives of Google, OpenAI, Meta, Anthropic and Nvidia signed the White House Accord on Super Intelligence.

Does any of this change what my business should do?

Very little, and that is good news. The questions that decide whether this pays off are what the system knows, what it is authorised to do, and whether it finishes the work. McKinsey found in August 2026 that 40% of organisations above $1bn revenue are scaling agents, up from 27%, while smaller organisations stayed flat at 22%. None of that turns on the name, it turns on getting those three answers written down.

Sources

1. Trump signs executive order renaming AI to super intelligence, Politico, 29 September 2026 Primary report of the order and its “to the extent permitted by law” scope.
2. BBC News coverage, 2026 Second independent report of the signing and its effect on federal agency language.
3. Donald Trump on the new world of super intelligence, The Verge, 2026 Source for the UN General Assembly quotation in New York.
4. Trump declares users of the term artificial intelligence the enemy, Politico, 8 October 2026 Source for the Truth Social post quoted above.
5. Tech leaders adopting Trump’s super intelligence term, Yahoo Tech, 2026 Source for Nadella, the SIForce rename and the Accord signatories.
6. Trump’s AI rebrand catching on with Elon Musk, Yahoo Tech, 2026 Source for the ai.gov change and the SpaceXSI pledge.
7. Trump’s AI super intelligence rename, fact-checked, PolitiFact, 7 October 2026 Source for the expert caution that the rename is misleading.
8. The state of AI, McKinsey, 25 August 2026 1,719 respondents, 97 countries; the 40%, 27% and flat-22% agent scaling figures.