Where a CEO starts transforming their company with AI: the 5 axes of the first quarter

Last night I had a conversation with the CEO of a company of around two hundred employees. He said it to my face with the honesty of someone who has already read the articles, attended three conferences and had coffee with two big consultancies: "I know we have to get started with AI across every area, but I don't even know where to begin". He was right in his silence. What he'd spent the whole year without was a usable answer.

If you're the CEO of a company of between one hundred and five hundred employees and you're at that same point, this article is for you. It isn't about tools. It isn't about pilots. It's about something more uncomfortable: the decision you have to make before spending the first euro.

The CEO's silence in July

Last night's conversation wasn't exceptional. It's the archetypal conversation I've been having with mid-market CEOs for months. Three different ways of saying the same thing: "I know we have to get started, but I don't know what to start first". "I've tried a pilot, but I don't know whether it's working". "My team brings me a new proposal every week, they all sound reasonable, I don't know which one to sign".

The mistake isn't the CEO's. The mistake is whoever answered before I did. Almost always the answer was a tool — "start with Copilot, try ChatGPT, set up an AI committee". All those answers are a consequence, not a cause. It's as if an owner asked where to start renovating their company and the answer were "buy new furniture".

The CEO's question is honest. The "buy a tool" answer is the one that's wrong.

Why this isn't an IT project

There are three reasons why your company's CTO, however good, can't kick off the transformation you need.

The first is that AI transformation isn't technology, it's operations. The technology part is almost solved: the models exist, the platforms exist, the tools can be bought with a credit card. What doesn't exist in your company is the redesign of how finance, operations, sales, HR and customer service work when an AI agent enters the flow. That redesign is operational, not technical. The CTO has no mandate to touch finance operations, and the CFO has no mandate to touch sales operations. You need a cross-functional mandate that doesn't exist in your org chart today.

The second reason is scope. The CTO runs a perimeter: applications, infrastructure, cybersecurity, product teams. Important, but limited. The five critical areas of your AI transformation are spread out: three live outside the CTO's perimeter. If you delegate the kickoff to them, they'll hand you a correct plan for their perimeter and a hole in the other four.

The third reason is the commercial model of the big consultancies. The firms circling your leadership team sell pilots. A pilot is an isolated slice, with an external team, on a six-month horizon, with a success metric defined by the consultancy. What they don't sell you, and won't sell you, is the transfer of knowledge to your company when they leave. If you end the first year with three promising pilots and your organization still depends on the consultancy to maintain them, what you've bought is dependency disguised as transformation. If you're in the middle of an evaluation, the five criteria for choosing your AI partner in 2026 help you separate the partner who transfers from the one who retains.

These three reasons aren't a criticism of the CTO or the consultancies. They're an observation that the CEO's question needs a role that today doesn't exist in most mid-market companies.

The 5 axes you must inventory in the first quarter

Before approving a single pilot, before buying a single license, before signing a single consulting contract, your company needs an honest inventory of five axes. The inventory is ninety days of work by an accountable role with a mandate from the CEO. It isn't a PowerPoint: it's an operational snapshot of what you have and what you're missing.

Axis 1 · Real processes (not the documented ones)

The vast majority of the companies I know have flowcharts in their SharePoint that don't match how the work actually gets done. The commercial proposal to the customer, per the flowchart, goes out five days after the lead is created. In practice, it goes out in twelve days after six emails between three people and an Excel that they update on Fridays. The difference between those two realities is the space where your AI will have measurable impact or where it will fail in silence.

The Axis 1 inventory is direct observation of the five to seven critical areas of your business. Not "tell me how you do it" meetings: observation of a full day of the real process, honest documentation, and the difference between what the flowchart says and what actually happens. That difference is your opportunity map.

Axis 2 · Corporate intelligence

There's a type of knowledge in your company that lives in the heads of five or six senior people and in spreadsheets they share with no one. Why we always sell more in March to logistics customers. Why we review supplier X's invoice by hand when all the others are automatic. How sales decides which accounts the seniors handle and which the juniors handle. Each answer to one of those questions is an asset your AI can use, scale and preserve — but only if it's codified.

The Axis 2 inventory is mapping where that knowledge lives, what would happen if the person who holds it left tomorrow, and what can start being codified first. There's a technical name for this that's circulating in the international conversation — context engineering — but the name doesn't matter. What matters is the CEO's question: how much of my business depends on people I haven't documented?

Axis 3 · Models and tools already active with no signed policy

When we run this inventory at a client, in seventy percent of cases we discover something the CEO wasn't expecting: there are already five, eight, twelve AI tools in use inside the company, with no signed policy, no audit, no inventory. Copilots installed by tech teams, individual ChatGPT subscriptions paid by managers on their corporate card, Notion plugins with AI enabled, email assistants contracted without going through procurement. If the question is "where do we start", part of the answer is "by recognizing what's already running without you knowing it".

This is probably the inventory that surprises the CEO most in their first two weeks. And it's the only point from which you can sign a coherent policy: you can't regulate what you haven't inventoried.

Axis 4 · Governance, compliance and security

On August 2, 2026, new obligations of the European Artificial Intelligence Regulation — the EU AI Act — take effect that affect companies like yours. If your Axis 3 inventory tells you that you have eight AI tools active with no signed policy and no risk classification, that date is relevant. Not to panic — to have an inventory.

The Axis 4 inventory is the crossover between what you've discovered in Axis 3, what you're planning to do with AI over the next twelve months, and the EU AI Act obligations that apply to your sector and your specific use. It isn't legal work in the abstract — it's joint work by legal, IT, operations and leadership. And it's done better in the initial ninety days than after an incident.

Axis 5 · Internalized transfer

This is the axis no consultancy will suggest you inventory, because it's the axis that most limits their commercial model. The question of this axis is very simple: when the provider who signed this work leaves, what stays in my company?

If the answer is "a system that works but that only the provider understands", you've bought dependency. If the answer is "a system my team understands, maintains and evolves", you've bought capability. The difference, over a three-year horizon, is usually several hundred thousand euros.

The Axis 5 inventory isn't a contractual promise. It's a metric observable month to month: how much of the work the provider signed in month 6 is done by your team in month 12 without support. We call it time-to-independence. The sooner it passes to your team, the more value the provider has delivered.

Why this inventory isn't run by your CTO or your consultancy

Let's recap. The five-axis inventory touches operations, finance, legal, IT and HR. It needs direct observation of processes across five areas. It requires authority to look inside non-inventoried spending. It has to deliver a policy that leadership can sign. And, above all, it has to transfer knowledge to your organization within twelve or eighteen months so it doesn't stay tied to whoever signed the work. It's the same principle that underpins the split of control across eight axes that we apply in larger-scale projects: who decides, who operates and what you take with you the day you change provider.

Your CTO runs a technical perimeter — they have no cross-area mandate. Your CFO has neither the time nor the role for this. Your external consultancy doesn't want the knowledge to stay in your company. And an intern or a marketing manager with an interest in AI has no authority to sit down with your leadership team.

What you need is a new role with a mandate from the CEO and a defined horizon: twelve to eighteen months of operation, a transfer metric, a planned exit.

AI Officer: the role that delivers the inventory and operates the transformation

At onext we call it the AI Officer. It's an external figure with a direct mandate from the CEO for twelve to eighteen months. In the first ninety days they deliver the five-axis inventory. Over the following nine to fifteen months they operate the transformation: process redesign, codification of corporate intelligence, model and tool policy, governance framework and, above all, internalized transfer to your company's team. Their primary metric isn't time-to-deploy — it's time-to-independence.

It isn't the same as an internal Chief AI Officer. An internal CAIO makes sense when your company has already been through the kickoff phase and needs a permanent leader. The external AI Officer is the previous phase: the one who builds the system and leaves. Almost always, when their mandate ends, they've hired or promoted internally the permanent CAIO who replaces them.

Nor is it a mini pilot. A pilot solves a narrow problem in six months. The AI Officer redesigns your company's operations so that AI enters every area with method, not through chaotic experimentation.

And it isn't a consultancy. The consultancy bills hours and stays until you decide to end it. The AI Officer has a finite mandate, a transfer metric, and an exit date planned from day one.

Your company, in the next chapter

Your company's next chapter with AI isn't bought. It's signed. It's signed by a CEO who, instead of choosing a tool, decides under what mandate and with whom at the wheel to run the inventory of their company's five axes before the first euro is spent.

What last night's CEO lacked wasn't information. He had all the information in the world. What he lacked was a structure to order the decision. I offered him this framework; it did him more good in thirty minutes than three months of conferences.

If the conversation he had with me last night resembles the conversation you've been having with yourself, let's talk. Thirty minutes with your leadership team are enough to see whether the five axes are more or less far off than they look from your seat.

Jordi García · Tech Lead at onext · June 16, 2026.