Before an action is handed to software, somebody has to be able to say how it would be taken back, and by whom. This is the list I ask a company to write, and what goes in each column.

Sort the actions by how long you have to change your mind

Not by how important they are. Two actions can matter exactly the same amount to the business and still need completely different amounts of supervision, because one of them can be put back all afternoon and the other cannot be put back at all.

Some actions are undoable. A draft is saved, a status is corrected, a note is added to a record. If it is wrong, somebody sets it back and the day carries on.

Some are expensive to undo. Stock is committed, a slot in a delivery round is taken, a credit hold is lifted. It can be reversed, but reversing it costs a phone call, or somebody’s afternoon, or a customer noticing.

And some cannot be undone. A message reaches a customer, money moves, a document goes to someone outside the company. Nothing puts those back.

The third group is usually the shortest of the three, which is why writing it down is an afternoon rather than a project.

Four columns

A filled-in row looks like this

Take releasing an order that is on credit hold. What it touches: the hold flag on that order, and the picking queue it drops into. How it is undone: put the hold back. Who can undo it: whoever is covering the credit desk that day. The window: until the picking list prints, because after that the order is physically being assembled and putting the hold back means walking into the warehouse.

Four short answers, and the last one is the reason this action can be automated at all: there is a clean point where the cost changes, and it can be written down.

The rows in the third group are shorter and less comfortable. Sending a customer the confirmation that their order has shipped: what it touches: their inbox. How it is undone: it is not. Who can undo it: nobody. The window: there is none. Nothing about that row is difficult to fill in. What is difficult is deciding, once it is written down in front of you, whether anything should be allowed to send it without a person looking.

What the finished list is for

Two things. It tells you which actions can run without anybody watching, because the worst case is a correction someone can make. And for everything else, it tells you what has to be true before it runs: who approves it, and how long they have to change their mind.

That second part is why the approval belongs in the same table as the action, with the name and the window as fields — a permission rather than a screen. An approval that lives only in a screen cannot be queried afterwards. What it leaves behind is the order in which things were clicked.

If you write nothing else, write the third group: the actions nothing puts back. It is short, and it is the only one where being wrong cannot be corrected afterwards.

My first sale was seven dollars, on the last day of 2022. Everything I run today starts there, and most of what I know about products I learned on calls, not at a desk.

Seven dollars is a real number

People treat a first sale as a formality. It is not. Seven dollars means one stranger, somewhere, decided that a thing I made was worth more than the money in their pocket. Nothing I had built before had ever been tested that way. Every product opinion I hold now traces back to that test, and to the two years of repeating it.

I sold while I studied

For the next two years I sold while I studied. Between December 2022 and November 2024 I delivered three hundred Looker Studio dashboards for clients, one commission at a time, and collected 103 verified client reviews. The clients were in the United States, in Europe, in South America and in Korea. Small businesses, and companies of considerable size.

At the busiest, five meetings a day. Not five calls: five conversations where somebody explains their business to you and you have to understand it well enough to be useful by the end of the hour.

The lesson

Here is the part I would tell anyone starting. In three hundred jobs, almost nobody asked me for a chart. They asked me a question about their own business and hoped a chart would answer it. The ones who left happiest were not the ones with the prettiest dashboards. They were the ones who could finally say a sentence out loud about their own company and know it was true.

That is why the interpretation goes next to the chart. A number nobody can read is not information, it is decoration, and I had to sell three hundred times before I stopped confusing the two.

The prices went up

The prices went up as the work got better. I am not going to put figures on that, because the figures are not the point and they would be the only thing anyone remembered. The point is the direction, and the reason for it: I stopped selling a file and started selling the answer.

What it turned into

Out of those three hundred commissions came Kodalogic, founded in Madrid in September 2024: six finished dashboards, distilled from what repeated in all of them, with the interpretation written next to each chart. And out of the habit of asking what a business actually needed to know came Hellomatik, founded on 28 July 2025, and the method we published with a DOI.

Isaac Correa, Madrid. ORCID 0009-0003-0657-4419.

My role is the one I gave myself: I think, and I direct the product. This is the order I trust, and where it came from.

Data, then media, then product

In January 2025 I sat down to explain what I believed about building, and it came out as three words in an order: data, then media, then product. Not because data is fashionable, but because the other two are guesses without it.

Data is what the business already knows and cannot say out loud. Media is how you find out whether anyone outside the building cares. Product is what you are allowed to build once those two have answered. Do it in another order and you will build something beautiful for nobody, which is the most expensive mistake available.

Two months, screen by screen

The part nobody photographs: two months designing screen by screen. Not wireframes in a workshop, actual screens, one at a time, each one argued over until it earned its place.

A screen is a decision made visible. If you cannot say what decision a screen helps someone make, the screen is decoration, and decoration in software is not free: someone maintains it forever.

From phone calls to a map of a company

Hellomatik started as something small: a system to answer the phone in a restaurant. A voice, a brain and a pair of hands. The voice takes the call, the brain decides what the call means, and the hands do the thing.

The hands were the problem. To do the thing, a system has to know what things exist, which rules the business follows and who is allowed to approve what. That question turned out to be much bigger than a restaurant, and answering it properly is what became the method we published with a DOI.

HAL 9000, and why it is a warning

People ask what the goal looks like, and the honest answer is that the picture in my head is HAL 9000: a system that understands the whole company and can act inside it.

I say that with some humility, because HAL is a cautionary tale, and the reason it is one is the same reason our work is mostly about limits. A system that understands everything and answers to nobody is the failure mode, not the goal. That is why every action carries whether it can be undone and whose approval it needs, and why the permission table is the part I would show a sceptic first.

What we run today

The report states the scope rather than implying it: four systems in production, among them a logistics one with 24 rules over live warehouse and ERP data, with over 30 million operations and 460,000 locations, and a B2B sales one that recalculates 29,000 customers nightly over 14 million sales lines. A method with its scope written down is a method.

Isaac Correa, Madrid. ORCID 0009-0003-0657-4419.

There is a moment in almost every automation project that nobody puts in the brochure. It happens in week one, it has nothing to do with technology, and it is usually the most valuable thing that will happen all quarter.

You ask two people the same question — when does an order need a second approval? — and you get two different answers. Both people are competent. Both have been doing the job for years. Both are certain. And the company has been running on that disagreement for as long as anyone can remember, absorbing it the way a body absorbs a limp.

Nobody was hiding anything. The rule simply lived in people’s heads, where rules do not have to agree with each other.

Why this surfaces now and not before

Software has been eating business processes for thirty years without forcing this reckoning, because software was always told exactly what to do. A form either has an approval field or it does not. The ambiguity stayed comfortably upstream, in the humans deciding which form to open.

Ask a system to decide, and the ambiguity has nowhere left to hide. Not because the technology is smarter, but because you now have to state the rule out loud to a thing that cannot read a room.

The unglamorous inventory

So the work starts somewhere nobody expects: making a list.

Which things exist — orders, customers, invoices, shipments — and which system tells the truth about each one. That last part sounds like bookkeeping and is not. The same order lives in three systems in most companies, and the three do not match. Until you name which one is authoritative, every clever thing you build downstream inherits the argument.

Then the rules. Not new policy — the policy people already follow, written so that something other than a person can evaluate it. A discount past a certain depth stops being the sales rep’s call. An account in arrears waits for someone senior before anything leaves the warehouse.

Then, and only then, the interesting part.

The part that decides whether it is safe

Every action a system can take is tagged with two facts: whether it can be walked back, and whose signature it needs.

Drafting a reply costs nothing and can be thrown away. Moving money is final the moment it happens. Once that distinction is written down rather than assumed, you can hand a system genuine independence over the reversible half of the work without ever having the frightening conversation about the other half.

This is the step that turns a security review from a confrontation into a reading. You are not asking anyone to trust a piece of software’s judgment. You are showing them a table of permissions that the business itself signed off, and pointing out that the software cannot go around it.

What the company gets even if the project stops

Here is the case for doing the inventory regardless of what you think about AI.

At the end of it, a company that had its operating logic distributed across a few dozen heads has it written in one place. It survives someone leaving. It can be argued with, corrected and versioned. New people can read it instead of absorbing it slowly over their first year.

That has been valuable since long before anyone used the word agent, and it will still be valuable when the current generation of tools is embarrassing to look at.

The rest — the automation, the agents, the part that demos well — sits on top of it. It is real, and it works, and it is not the foundation. The foundation is the boring list.

The full method behind this is published as a technical report with a DOI rather than a brochure: how we build a company’s ontology, deposited at 10.5281/zenodo.22647735.

Isaac Correa, Madrid. ORCID 0009-0003-0657-4419.

These are the four questions I put to a company’s own people in the first week, before a line of software is written. They are deliberately small — the impressive questions are the easy ones to fake an answer to. Work through them with your team and mark, honestly, which description is nearer the truth today. Nothing is sent anywhere; the page keeps your answers in your own browser and nowhere else.

What to do with the answers

Ask the questions separately, of people in different departments, and write down what they say word for word. If the answers agree, you are further along than most companies and the technical work will be quick. If they do not, you have just found the project — and the disagreement is worth more than any of the individual answers, because somebody has been absorbing it personally for years.

When the answers do agree, the next list is not another question but an inventory: which actions can be undone, and who would undo them.

None of this needs me, or software, or a budget. It needs an afternoon and somebody willing to write down an inconvenient answer. The longer version of why is here, the vocabulary is here, and the method these come from is published with a DOI.

In September 2026 we published the method Hellomatik uses as a technical report with a DOI, deposited in two independent repositories and indexed under my ORCID. This page says what is in it and where to read it.

Why a DOI and not a whitepaper

A technical report can be cited, checked and taken apart. A DOI puts a fixed, dated version of the text in a repository that is not ours, so anyone can quote it and it does not depend on us keeping a link alive.

The people we want reading it are operations directors, systems people, the person in a company who is quietly responsible for the ERP. They read papers.

What is in it

The report goes layer by layer. It starts with the data of the company and which system each piece comes from. Then the rules the company already decides by, written so that an agent can apply them rather than guess. Then the actions: what can be done, with what parameters, leaving what record. And last, permissions: who, or which agent, can go how far, and which actions wait for a signature.

It also describes what we have running today: four systems in production, including a logistics one with 24 rules over a live warehouse and a B2B sales one that recalculates 29,000 customers. A method with its scope written down is a method.

The story of how the method came about is here.

Most advice about innovation says the same thing: be creative. I think that is the wrong instruction, and I run two companies on the opposite one. This is the short version, with the two rules I actually use.

Creativity is not something you look for

Nobody innovates by sitting down to be creative. Creativity is not an input you can decide to have. What you can decide to have is the ability to make better connections between things, and that ability comes from one place only: having lived things that do not resemble each other.

That is why I trust people with two careers that do not touch more than I trust anyone who has spent a whole life inside one sector. The specialist knows his field better than I ever will. But the person who has sold, built and studied unrelated things has more material to connect, and connections are where the new comes from.

Rule one: copy, and copy well

My advice to anyone starting out has nothing romantic in it. Copy others, as much and as well as you can, until you can find creativity where there is none. Copying well is not cheating. It is how you learn what already works, and it is the only honest way to notice the gap that nobody has filled yet.

Kodalogic, the dashboard company I founded in September 2024, came out of exactly that. We looked at three hundred jobs done for real clients, found the pattern that repeated in all of them, and kept the best of each. The product was not designed. It was distilled. The businesses we had looked at sold whole catalogues of them. We sold a handful, and they sold.

Rule two: do not look for gold, make it

The second rule follows from the first. Once you have copied well enough to see the gap, you stop looking for an opportunity somebody else left lying around. You make one. Hellomatik, the company I founded in Madrid on 28 July 2025, started as a system to answer restaurant phone calls and became a method for writing down how a whole company decides, published as a technical report with a DOI. Nobody left that lying around. It was made, one connection at a time, between dashboards, phone calls and the question of what a company actually is.

The people this works with

There is a kind of person this works with, and I have a one-line definition: people who see greys where the rest of us see blacks. People who find the opportunity where others see the setback. They do not solve problems and they do not look for solutions. They create them.

Isaac Correa, Madrid. ORCID 0009-0003-0657-4419. The whole story, with the dates.

I wrote this in November 2025, while building Hellomatik in Madrid. It is where I set down what I think the next tool is, and what it will always need from us. The machine of the future is the tool that can take actions in the real world when you write what you need, in plain language.

Why it arrives

It is reasonable to assume that, at some point in the evolution of technology, running a company will stop being a manual, fragmented activity spread across many separate tools. The historical pattern has always been the same: when a task is repetitive, can be expressed as rules and is necessary for an organisation to keep going, sooner or later it gets automated.

We can imagine, then, that creating a company will come down to switching on a central system: a platform that takes over the functions several human departments perform today. That platform will not just store data or execute orders, as today’s programs do. It will learn patterns, infer intentions and make operational decisions within set limits.

If a company needs to publish information, adjust its public presence, answer customers or coordinate suppliers, the system will not wait for detailed instructions. It will act. It will recognise what is worth doing given the circumstances and carry it out with the constancy of a machine and the flexibility of an experienced professional.

In practical terms, it will be like hiring not a tool but an administrative mind. An entity that coordinates, maintains and corrects the internal workings of a business. A company managing another company from the inside, without a body, without a timetable, without wear. Its only job will be to operate.

What that changes

If the whole management of a business can be concentrated in one platform, reachable from anywhere and able to carry out tasks on its own initiative, the barrier to starting a company drops sharply. In other times, starting an economic activity demanded capital, contacts, time and specialised knowledge. If much of that complexity is delegated to intelligent operating systems, anyone with a basic idea could try to become an independent worker.

That has a direct consequence: the number of independents rises. Not because of a cultural change, but a structural one. When costs fall, participation rises. We have seen it before in economic history, every time technology simplified production or distribution.

The same simplification has a side effect. If everyone has the same tools and the same automated technical capabilities, the differentiator is no longer manual skill, nor even specialised knowledge. It is the ability to shape something that others consider valuable. In other words: judgment, taste, and the ability to generate interest.

Companies will then look less like human organisations and more like systems of services interacting with each other. Many will exist only as digital structures that produce content, negotiate with other automated entities and keep themselves running without constant human presence. In that scenario, the main task of the individual will not be to do the work, but to decide what is worth doing. That is not a loss of humanity. It is a reconfiguration: human activity moves from execution to intention.

What stays human

The value of a company is the people who work in it. That will always be true, however far the technology goes. What is also true is that a company’s clients value it not for its headcount but for its ability to meet their needs. A company that can understand and fulfil what its clients need has value regardless of how many people run it. So the machine of the future is the one that lets a legal person depend a little less on physical persons for its existence. And all of it works only with a human component in its decisions.

From organisation to being

I would call a being any entity, or person, able to make decisions with a purpose behind them, knowing why. Today’s AI agents and chat products do not have that component. They execute without knowing why. They are programs, like all programs, that now meet needs in a simpler and more efficient way. Nobody has given them enough to decide what is best at each step. My bet is that a company can gradually evolve from an organisation into a being in that sense: an entity that decides with a purpose, and knows the purpose.

What the machine needs

First, memory to feed on context: the tools it is connected to, what they do, its personality, the situation, the knowledge of the company. That artificial brain must not be limited to what we give it at the start. It has to create records and memory of its own experience, so that it does its job better each time.

The most important thing inside that brain is a goal and a few basic principles to base its decisions on. The goal matters as much as the principles, because faced with a decision it will choose one option or another according to the values it was given. Most people, when they build an AI system, see it as a program that is a bit easier to operate. That is the wrong layer. Its real power is in giving it a purpose and rules for making decisions.

This matters most when it deals with humans, because human interaction is unpredictable. Think of sitting an exam in a subject you have never studied, with a three-hundred-page book in front of you. You can do it, but the result is imperfect. The next day you already sense which pages to look at. After enough mornings of the same exam you answer without opening the book. And faced with a completely new exam, you chain ideas from the previous ones into answers you had never given. That is what memory will do for this machine: turn knowledge into instinct.

Making a decision is the most human thing there is. To make one, a person weighs feelings, knowledge, needs, principles and experience. A company, to me, is like a child: if you look after it, it grows, and it needs you every day. Anyone who owns a small business knows that feeling: the wish to step away for a few days and come back to find the clients answered and the work done. What I am building is what lets that child grow on its own. Put without the metaphor: the difference between having a business and being tied to it.

This essay is deposited as a working paper with its own DOI, 10.5281/zenodo.22671771, and is also on Academia.edu.

Isaac Correa, Madrid. The method Hellomatik uses to write down a company’s operating model is published as a separate technical report with a DOI. ORCID 0009-0003-0657-4419.

Read the working paper (PDF) · DOI 10.5281/zenodo.22671771

My first sale was seven dollars, on the last day of 2022. Everything I run today starts there: Hellomatik, the company I lead in Madrid, Kodalogic, the company I founded before it, and a method with a DOI.

November 2022: the first applause

A team I was part of won a special recognition for the most innovative solution at an entrepreneurship competition. I was twenty, and it was the first time a room applauded something I had built. I barely remember what we presented.

The record does. The ceremony was streamed on 30 November 2022 and the video is still up: it runs for three hours and thirty-nine minutes, and I come on at 3:28:19. I give the timestamp because the point is not the prize. The point is that you can go and check it. That habit — say only things another person can verify on their own — is the one thing I have kept from then until now.

31 December 2022: seven dollars

On the last day of that year a client in the United States paid me seven dollars for an analytics dashboard on a freelance platform. I did the job as if it were worth a thousand, got paid, and closed the year looking at that number. It was the first of many.

2023 and 2024: three hundred dashboards from a desk

For the next two years I sold while I studied. I managed clients from half the world from my desk, with the meetings in the afternoon: the United States above all, half of Europe, South America, South Korea. Small businesses and companies of considerable size. Some days there were five meetings, and the deliveries went out in hours.

The prices went up as the work got better, and the numbers at the end are the ones I keep: three hundred dashboards delivered, and 103 verified client reviews.

The question that founded a company

By the three hundredth dashboard I could see the pattern. Three hundred clients had asked for the same thing: the same questions to their data, the same report, the same blank canvas to start from. The question I asked was not about the product. It was about access: why can only the people who talk to me have this?

I founded Kodalogic in Madrid in September 2024. Out of three hundred jobs for real clients, the catalogue is six finished dashboards, distilled from what repeated in all of them, with the interpretation written next to each chart. Everything Kodalogic has sold came from appearing in search results, and that took articles and patience: I wrote 58 of them, all signed. The first sale came from the United States, and that client still pays today.

In January 2025 the two of us were invited back to a university to explain the company to a hundred students. That is where I first put into words the order I still work by: first you are a data company, then a media company, and only at the end a product company.

From a phone call to a company’s ontology

The first idea for the next company was small: a system that answered restaurant phone calls. Then the idea grew. A dashboard gives a company its data. Why not talk to that data? Why not give a company a voice, and a brain? And the question that made it mine: why not give it hands too?

Before a single line of code, I spent two months alone with the idea. The first month deciding what it had to be and how it would be explained. The second turning that into a design, screen by screen, of a product that did not exist yet. Months later, with the tool running and in use, I compared the design with the real thing. Almost nothing had moved. When an idea is that clear, you can build the whole tool before you build it.

Hellomatik was founded on 28 July 2025 in Madrid. I was twenty-two. The name is a contraction of “hola” and “automatizaciones”. My role is the one I gave myself: I think, and I run the product. The reference I keep in mind is HAL 9000, an entity that understands a whole organisation and helps the people inside it, and I keep it as a reminder of how far there is still to go, not of how far we have come.

The first client was a large clinic. We gave the agents the whole content of the company, then a goal and an intention, so they would not just answer but resolve. Those agents are still running, and the clinic is still a client.

The idea kept evolving, from answering a call to writing down how a company decides: its data, its rules, its actions and who may approve what. In September 2026 we published that method as a technical report with a DOI, How we build a company’s ontology.

What selling it taught me

You do not lead with the technology. Nobody buys “we built an AI company that automates with your data”. What works is naming the client’s own problem: if your CRM is a mess, we fixed that for several companies, and we can look at yours. A product is aimed at one specific person, or it is aimed at no one.

How I think about innovation

Nobody innovates by sitting down to be creative. What you can decide to have is the ability to make better connections between things, and that comes from having lived things that do not resemble each other. So my advice to anyone starting out has nothing romantic in it: copy others, as much and as well as you can, until you can find creativity where there is none. Then you stop looking for an opportunity somebody else left lying around, and you make one.

There is a kind of person this works with, and I have a one-line definition: people who see greys where the rest of us see blacks. People who find the opportunity where others see the setback. They do not solve problems and they do not look for solutions. They create them.

Isaac Correa, Madrid. ORCID 0009-0003-0657-4419.