A guide for the CEO · 2026

AI in your company: the honest guide for the CEO.

Not a trends report or a list of fashionable tools. An honest guide for the CEO who wants to transform their company with AI — for real. The questions that matter and the answers worth having clear, before you sign anything. No jargon, no hype, no sales agenda.

01 — The starting point

Almost every company already uses AI. Almost none has a strategy.

In two years we've gone from AI being a curiosity to it being on almost every leadership table. But the pace of adoption hasn't come with judgment.

When we sit down with the leadership of a large company, the picture repeats with almost uncomfortable precision: there are ChatGPT licences scattered around, someone on the team "experimenting", a pilot here, an automation there. And a year later, no real impact on the business. Lots of activity, little direction.

~8%
of large companies were using AI seriously two years ago
45-55%
use it today — the fastest technological leap we've ever seen
85%
stay on the surface: loose prompting, no governance, no measurement

Figures from Nalum's AI adoption report · Spain and Mexico, 2026

We call it strategic chaos: the phase where a company already uses AI in many places, but in a disorganised way, with no plan connecting those uses to the business. It's not a tools problem. It's a judgment problem. And you don't fix it by buying more software.

02 — The framing mistake

AI is not a tool you add to your company.

It's the operating system of your company. And that shift in framing changes everything.

The most expensive mistake we see isn't technical. It's about framing. Most executives think of AI as one more tool: something you buy, install in a department, and that promises to save a few hours. Selling AI as "time savings" gets it wrong, because it leaves untouched what actually matters — how the company is structured.

Real transformation isn't doing the same thing a bit faster. It's redesigning how the organisation works, starting from the fact that there's now a new lever. Which teams make sense. Which processes are redundant. Where the fat was and where the real muscle is. That isn't solved with a licence: it's solved with a leadership decision.

"The true transformation from AI isn't productivity gains. It's redesigning the structure of the company."

— Nalum's thesis on AI

That's why the first question is never "which AI tool do I buy?". It's "what do I want my company to be in two years, and how does AI help me get there?". If you start with the tool, you end up in the chaos. If you start with the business, the tool arrives on its own and in its right place.

03 — Before you talk to anyone

The 5 questions a CEO should answer first.

Before hiring someone — before even talking to a provider — there are five questions worth being clear on. If you can't answer them, that's exactly your first job. And probably the sign that you need someone to think with you, not sell you something.

01

What business problem do I want to solve — not what technology do I want to use?

"Implementing AI" is not a goal. "Cutting waste in perishables by 20%" or "responding to an inbound lead in under a minute" are. If your goal has the word "AI" inside it, it's not a goal yet.

The test If you remove the word "AI" from your goal and it stops making sense, you don't have a goal. You have a trend.
02

Is my data in shape?

AI is built on data. If your information lives in silos, with empty fields and no single source of truth, your first project isn't an AI project — it's a data clean-up. It's not glamorous, but it's where almost everything is won or lost.

The test Could you answer today, with data and not gut feeling, which is your most profitable product per store? If you hesitate, start there.
03

Who owns this inside my company?

AI transformation touches structure, processes and people. That can't be led by "the IT person" in their spare time. It needs an owner with weight on the leadership team and the mandate to make uncomfortable decisions.

The test If no one on your leadership team loses sleep over this, it won't happen. AI without an owner stays stuck in pilots.
04

Am I willing to change how the company works, or do I just want a patch?

This is the honest question. If you're only looking to reassure the board with a flashy project, there are cheaper options. If you genuinely want to move the needle, be ready to redesign, not just to install.

The test An AI project that makes no one uncomfortable almost never changes anything. Real transformation reorganises.
05

How will I know if it worked?

Define the metric before you start, not after. If in 90 days you can't say which number moved and by how much, the problem wasn't the technology — it was that you never defined what winning meant.

The test Write down today, in one sentence, what would have to happen in 90 days to call it a success. If you can't, that's the first deliverable.
04 — Who to do it with

The market offers you two options. A third one is missing.

When a CEO goes out looking for help with AI, they almost always run into two types of provider. And the problem is that neither one is complete.

Option A

The big consultancy

Sells strategy with a known name. But whoever signs your proposal isn't who executes, the cycle is slow and bureaucratic, and the cost is enormous. You end up with an immaculate PowerPoint and a junior team learning on your dime.

The risk Strategy on paper, far from real operations.
Option B

The technical agency

Knows technology and installs tools fast. But doesn't understand your business. They build you the bot without asking whether the bot was the problem. You end up with loose pieces that don't talk to each other or to your strategy.

The risk Technology with no business judgment behind it.
The third option

Someone who has already sat in your chair.

There's a third type of provider, less common: people who have built, scaled and sold real companies, and who have operated from inside large corporations. They don't give you textbook theory — they give you what they'd do if it were their company and their money. Strategy and operations in the same person. A founder's judgment, not a consultant's.

Yes — this is what Nalum does. But the judgment holds whoever you choose.

05 — Red flags

5 things that, if you see them, should make you run.

They promise results before looking at your company

Anyone guaranteeing a "30% efficiency gain" without having seen your data, your processes or your people is guessing or lying. What works in one place can be a disaster in yours.

They talk about AI, not about your business

If the first meeting is more about models and tools than about your customers, your margins and your problems, they have it backwards. AI is the how. Your business is the what.

They jump straight to "building" with no discovery

Anyone wanting to start developing before understanding the problem is selling you a generic solution. The right order is: understand, diagnose, decide, then build.

They leave you dependent forever

If the system is a black box, they give you no access or documentation, and every minor change costs another invoice — they're not helping you, they're tying you down. A good partner leaves you capable, not hooked.

They agree with everything you say

If you're paying for expert opinion and all you get is validation, you're wasting your money. The value is in someone telling you, respectfully, where you're getting it wrong — not in applause.

06 — How to start well

You don't need the perfect plan. You need to start well.

Paralysis is as expensive as improvisation. The companies that win aren't the ones waiting for the perfect AI strategy — they're the ones that take the first step well and learn fast. Here's what that first step looks like.

Before spending anything

Define the problem, not the solution

Write down which process you want to improve, which metric you want to move, and where you are today. Without this, any provider will sell you their favourite solution, fit or not.

Week 1

Take an honest snapshot of your data

What state is your information in? If it's messy, that's the first job. Better to know before promising miracles than after paying for them.

Before committing big

Pay for a diagnosis, not a demo

A free demo is a sales pitch. A paid diagnosis aligns incentives: the provider proves judgment and you get a real map before committing capital. If someone skips this step, be suspicious.

The first 4-8 weeks

Start small, with a measurable result

A small, low-risk pilot, in a specific area, with a number you can move in weeks. You validate, you learn, and only then you scale. Never roll out across the whole company what you haven't tested in one corner.

07 — What to actually expect

A document you can read and decide on. Not a show.

If you do it right, what you should receive at the end of a good diagnosis isn't an 80-slide deck to impress the board. It's something far more useful and far rarer: clarity.

At Nalum we call it the CEO Memo: a written document, founder to founder, with no consultant language. What we'd do if we were sitting in your chair. We describe it not to sell it to you, but because it's a good standard to demand from anyone you hire.

What it should include

An honest diagnosis.
Ordered priorities.
Concrete recommendations.
Decisions you can make when you close it.
The plan for the next 90 days.

What it shouldn't have

Decorative slides.
Made-up frameworks.
Generic benchmarks.
Consultant language.
Disclaimers to cover themselves.

The difference between describing and giving an opinion is the difference between a consultancy and a partner. A consultancy hands you an analysis and leaves the decision to you. A good partner tells you what they'd do — and why. Demand the second.

08 — The next step
— Nalum

If you recognise your company in this guide and you're not sure of the next step, let's talk.

We're founders, not career consultants. We've built and sold companies, and we work from that experience: with our own judgment, honesty, and the candour of people who have sat in your chair. We help leadership teams see clearly where they are, prioritise where it makes sense to start, and launch the first pilots without building unnecessary structures or depending blindly on large providers. We only take 4-6 projects a year.

We work in
Spain · Portugal · LATAM
Email
Web