What we sell is trust. When a leader asks me what our firm does, that is the whole answer, and the rest of this page is what the answer means. You are putting artificial intelligence to work in your company, or you are about to, and what we offer is the ability to trust it, shown with evidence rather than asserted. There are seven things we can let you come to trust: that the AI itself is trustworthy, because its governance travels inside it; that it is aligned with what your business is trying to do; that it pays back; that it sits inside your risk and compliance limits; that it is the right tool for the job in front of it; that you could change vendors without starting over; and that a named person owns every agent that runs in your name. Each of those is a page of its own. This page is about the third, because the third is the one that cannot be bought later. Whether AI paid back is a question about before and after, and almost nobody measures the before.
In the last years of my career I ran AI governance for a global enterprise of more than 120,000 people, and a great many AI proposals crossed my desk. I learned to tell within a page where each one had started, and nearly all of them had started in the same place, which was the middle. There was a data lake already being built, or a model already chosen, or a pilot already a month old, and the proposal was asking for the next thing. What was almost never on the page was the first thing: who the part of the business this served actually was, what it was for, how it decided, and what a good year looked like to it. I would ask, and the answer was always that the business knew all of that. It did. It had just never written any of it down anywhere a machine could read it, and the machine was about to be hired. The firms we work with now are a thousandth of that size, and what we see in them is the same pattern running faster, because the pilot is the owner's own idea and there is no committee standing between the idea and the login.
A leader who has decided to put AI to work usually starts where the technology is: at the data lake, at the model, at the pilot. I do not say that as a criticism. The technology is where the vendors are, where the demonstrations are and where the budget line already sits, and a reasonable person goes where the path is paved. The work begins in the middle of the stream, and the AI that results does not know what matters to the business, because the organization never told it who it is, what it holds, who decides or what good looks like. The leader reads the result as a technology failure and buys a better tool, which fails for the same reason. I watched that cycle run more than once, in good companies, with serious people, and the second tool was always better than the first and did not help.
The sentence I have settled on, after watching it enough times, is the one this page and the book that follows it rest on: leaders dive into AI in the middle and get nothing, because they skip the work of knowing what the organization is.
We onboard AI the way we onboard software, when it needs onboarding the way a new colleague is onboarded. A new colleague's first week is not spent at the data warehouse. It is spent learning who the company is, what it is trying to do this year, who to ask and who decides, and what the people around them mean when they say that something was done well. Nobody hands a new hire a database and a login and calls that onboarding. We do exactly that with AI, and then we are surprised that it does not know what matters.
I spent a good while looking for the framework that would have caught this, and I want to be fair to the ones I found, because many of them are good. The project playbooks, the responsible AI frameworks, the data readiness checklists, the vendors' own adoption guides, the regulators' guidance: I read them with one question in mind, which was where each of them begins. Every one of them switches on after the decision to build. They assume a project exists, a use case has been chosen and someone has decided what the AI is for, and they govern from that point forward, often very well. None of them asks who the organization is, why the project exists or what work it serves, because by the time they are opened those questions have been answered somewhere, by someone, and the framework takes the answers on trust.
So there is an open lane before the decision to build, and nobody is standing in it. I call it the Inception Void. Inception is the moment before a thing is decided, and the lane is a void because the methods, the tools and the professions all begin one step later, so that the one question whose answer governs everything downstream is the one question no one is assigned to ask. A company does not fall into it on purpose. It falls in because every path it can see starts on the far side.
The work that closes the void I call Step Zero, and the name is deliberate, because it has to happen before the first step of anything else. Step Zero captures who the organization is, what it holds, who decides and what good looks like. Who it is means its mission, its values and the standard it holds itself to when nobody senior is in the room. What it holds means its processes, its systems, its people, and the agents and tools already running, found rather than assumed. Who decides means the real decision rights, which are rarely the ones on the organization chart. What good looks like means the goals, written down, each with the number that would tell you it had been met.
And it begins with the financial ledger, first, before any of the rest, because the ledger is the one record every company already keeps that says what the business actually did. Mission and goals mapped to actual performance, frozen and dated before AI touches the business, is what I call the pre-AI baseline, and it is the thing the third trust depends on. Nobody measures the business before AI, so nobody can show what AI changed. An organization that adopts AI without a dated baseline cannot tell an investment that worked from one that only felt busy, and I have sat in the meeting where a leader tried, with a vendor's slide in one hand and a quarter's numbers in the other and no way to connect them. A dated baseline does not promise that the AI will pay back. It makes the question answerable, which is more than anyone had before.
I should say which Step Zero I mean, because the name has grown. In my first book, Step Zero was Culture Mapping: before you govern an agent, you write down what trust means here, what the organization values and what it refuses. That is still true, and it is still the first thing we ask. Later I widened it to Context Mapping, because culture is the who of an organization and an agent also needs the why and the what, the purpose and the actual work. And in the method we run now it widened once more and put the ledger first, so that the context an agent inherits arrives with a measured starting point attached. Those are three sizes of one idea, nested inside one another, and when I write Step Zero on this page I mean the largest of them.
The method has five steps, and I will name them and not teach them, because the teaching is the engagement and the names are enough to see the shape. Step Zero is the first, inception with the ledger. Inventory and mapping runs alongside it: every agent, tool, vendor and process is found and placed on the framework. Alignment comes third, connecting each of those to a goal the organization wrote down and finding the gaps, including the spend that serves no goal. Execution is the fourth, a dated plan with owners, controls and human checkpoints. The fifth is the living twin, the picture kept current and re-measured against the baseline.
The picture is the part I would want to see first if I were the one deciding, so it is the part you see first. On day one you see the finished twin, which is one picture of your business with every agent, vendor and goal placed on the map, and most of it is drawn as outlines, because most of it has not been captured yet. I call that the ghost twin. Each step turns a part of it solid. The ledger makes the baseline solid. The inventory makes the agents and the vendors solid. Alignment draws the lines between them and the goals, and the lines that cannot be drawn are the gaps. Nothing in the method is hidden from you, since you are looking at the whole of the end state from the first morning, and the engagement is the work of filling it in. The picture's full name is the Trusted Living Twin. I use the short name everywhere else, because what matters about it is not the name but that it is yours, that it is kept current, and that you could take a complete copy of it on any day you chose.
Most of the leaders we talk to have already started. There is a pilot, or three, and a vendor contract, and an agent that somebody in sales connected to the customer system over lunch, and the honest fear is that Step Zero means starting over. A company that has already begun does not go back to the beginning. It backs up one step and runs Step Zero while the inventory of what it has already built is made alongside, which is why the second step runs concurrently with the first instead of after it. What you have built is not wasted. It is unplaced, and placing it is most of the work. The agents that turn out to serve a goal keep running. The ones that turn out to serve none become a decision you get to make on purpose, which is more than you have today.
On the map a page ago, five pillars across and three signatures down, Step Zero is the band underneath the grid, the ground the whole picture stands on, and it is drawn before anything is placed above it.
We have found that the void is easiest to see on a company that is not yours. Hollister Reed is a fictional demonstration firm, a regional accounting practice built for the purpose, with the ad hoc AI use most firms of its size have and the gaps that go with it, and its twin is the first thing we show a leader, from the ghost on the first morning to the picture at the end. The engagement that builds a twin like it is run by Mortar Cloud LLC, the firm I founded, and the engagement itself is described on the firm's own site. I still do not know why the first question is the one nobody is assigned to ask, when it is the only question whose answer a machine cannot supply for itself, and I have stopped expecting a framework to answer that for me. What we can do is show you what closing the Inception Void looks like on a firm where nothing is at stake, and let you decide whether it looks like yours.
The twin is here: the Hollister Reed demonstration. When you have seen it, a screen on the same site draws your own ghost twin from five questions, and that is the next step: take the five-question screen.