The revenue test every AI project should pass

Most of the AI spend we see inside small and mid-sized companies fails one simple test: nobody can say how it makes the business more money. We run every Imperium engagement through a filter we say out loud at every mastermind, and Archie has said it as bluntly as it can be said: "It's not AI for AI's sake ... AI which directly makes your business more money." This article is that filter written down, with the formula, the sequence, and the failure cases, all sourced from what we have actually said and seen on the record.

What is the formula?

At the operating-system mastermind I wrote our internal metric on the board: revenue growth is speed of execution multiplied by the quality of decisions. Data feeds both sides. Better data makes better decisions, faster execution produces more data, and the loop compounds. It sounds simple because it is. The value of any AI project is whether it moves one of those two variables. If it speeds up execution, it passes. If it sharpens a decision, it passes. If it does neither, it is a hobby with a subscription fee.

The counter-example we warn about is the dashboard. Archie's phrasing at the revenue session was direct: the one thing you don't want to do is build a dashboard for the sake of it. A dashboard that nobody acts on moves neither variable. We have seen teams spend two weeks and thousands in credits on a beautiful screen that changed no decision and sped up no work.

Why do you question before you automate?

The sequence matters as much as the formula. The four steps we quote, from Elon Musk's engineering algorithm, are: question the requirements, delete what is unnecessary, simplify what remains, automate last. Archie keeps it on his phone as a screensaver, and the point we land in every telling is that most people start at step four.

"What's the point of automating something if it doesn't need to be in your company?"

Archie, What is an AI Operating System?, Imperium mastermind

In practice this deletes a surprising amount of work before anything gets built. A weekly report nobody reads should be deleted, and deleting it costs nothing. Automating it would have cost real money and preserved a dead process forever.

What does passing the test look like?

The proof we point to is what clients say in the room, in their own words. A business owner who runs his company on one of these systems told our mastermind audience what happened to his admin load: contracts read, reconciliation done, accounts watched, competitors scraped, and in his words, "literally five people's job just disappeared." That is his claim about his own firm, said publicly at the session, and it is the shape of result the test selects for.

The same owner also described a compliance deadline that would normally have eaten a full day: the system scraped the new rules, drafted the emails to everyone in his organisation, produced the documents, asked "should I send this?", and finished in 20 minutes. His summary was the sentence every owner wants to say: all the admin is gone.

The measure we track underneath these stories is revenue per head, and we describe the same coin from two sides on the recordings. My framing, echoing a line one of our attendees used first: we are not here to cut people out of a company, we are here to make every single person more efficient. Archie states the harder economic version: the biggest lever is the ability to scale without adding headcount, and the companies which reduce their headcount massively and have the highest revenue per head hold the longest competitive advantage. Where a role is pure manual input, he means exactly what he says. Where a role carries judgement, the busywork goes, the person moves up the value chain, and the ratio improves either way.

What does failing the test look like?

Failing usually looks like hiring or buying on fear. A consultancy owner who joined one of Archie's recorded conversations described the pattern he sees constantly: business owners get FOMO from what they see on social media, then go and hire the first technical person they can find at 150 to 180 thousand dollars a year in his Australian market, with no commercial lens and no implementation strategy. His verdict on the recording was that the hire usually costs more than the salary.

The same fear drives tool-buying. A subscription to something impressive feels like progress, and it quietly fails the test because nobody asked which decision it sharpens or which execution it speeds up. The test is boring on purpose. It is one question asked before any money moves: where does this show up in revenue?

How fast does the window close?

Faster than most owners think, which is why we push speed so hard. My own summary of a busy month, said on the podcast: we pivoted our software company in four weeks, created an event in one week, and had about 25 pieces of content lined up to record in four days. Speed, speed, speed. That pace is the only durable advantage we have found. Systems themselves are becoming commonplace. Archie tells audiences plainly that in three to six months a lot of businesses will have these systems in place, and then the question becomes what actually differentiates you.

The honest version of the pitch is that the revenue test never stops applying. It filters the first project, the tenth, and the way we charge. Archie has said on the record why we structure engagements around results: we are ingrained in your business, our offer is not "we will automate this", it is being directly correlated to you making more money. A test that strict has to run both ways, and it does. It is the same one we ask you to hold us to.