There is a scene that is repeated in every Italian company , with a regularity that has something liturgical about it. An entrepreneur (a person who has built something with his own hands, who knows every customer by name, who knows exactly where he margins and where he loses) sits in front of a management system and waits. He waits for it to load. He waits for a consultant to explain how to extract a report that should have been trivial. He waits for the quote for a customization that in a rational world would take half a day. He waits for the invoice, which arrives on time. That one, yes, no one ever waits for it.
It’s a scene we at Oltrematica know well, because we see it every time a new customer tells us their story. And after years of building technology solutions for companies of all sizes, we feel it’s our duty to say something uncomfortable: many of you are paying very expensive rent to inhabit a house built with your own bricks and mortar.
The data is the company’s. The value, however, someone else’s.
Let’s stop for a second and think about what’s really inside that management system you use every day. There are 20 years of invoices. There are the buying patterns of your customers. There is the seasonality of your demand, the marginality of each individual product, the speed with which you pay and how quickly you get paid. There is, in essence, the DNA of your company.
And that DNA is locked inside a system that you did not design, that you cannot freely interrogate, and from which you cannot get out without a pain that, not surprisingly, the provider has made as sharp as possible.
It is not a bug. It is the business model.
Technology lock-in is not a side effect: it is the strategy. Every proprietary integration, every closed format, every API that doesn’t exist or costs extra, is another thread in the cocoon that keeps you immobile. And the cocoon, unlike silkworm cocoons, produces nothing of value to those in it.
Reading business data with artificial intelligence changes the rules of the game
Here is the part that excites us, and should excite you as well.
For the first time in the history of enterprise computing, there is a technology capable of doing something that until yesterday was unthinkable: making sense of your data without having to rewrite everything from scratch.Artificial intelligence (the real kind, not the chatbot tacked onto the homepage) can read, interpret, connect and reason about information that today sleeps buried in your archives.
Think about what this means concretely. It means being able to ask, in Italian, a system,“Which customers have reduced orders in the last three months and why might this have happened?” and getting an answer that cross-references sales data, communications history, market trends and seasonality. Without opening Excel. Without waiting for the consultant. Without the 47-page report that no one will ever read.
It means that the deep knowledge of your company, the knowledge that today is distributed among the owner’s memory, the salesperson’s Excel spreadsheets, the 2019 emails and that notebook that “only Mario knows where it is,” can finally become a coherent, queryable, living system.
The subtle danger: AI as a more elegant padlock
But beware. There is a huge pitfall at this moment in history, and we tell you without mincing words.
The big management vendors, the ones who have been selling you hard software by the truckload for years, are adding “AI features” to their products. An assistant here, an automated suggestion there, some predictive analytics in the dashboard. That sounds like progress. In fact, in most cases, it is the exact opposite.
Adding a piece of artificial intelligence to a closed, proprietary system does not free you. It shackles you more. Because now you depend not only on the management system for your data, but also on the management system’s AI for the intelligence that that data generates. Another layer of dependence. Another reason not to leave. Another cost you did not choose.
It’s as if the owner of a rented apartment installed a beautiful, custom-made kitchen for you, but it only runs on gas from his system. You cannot take it away. You cannot use it elsewhere. The only thing you can do is to keep paying rent. Maybe a little higher, because now you have the new kitchen.
We would deeply distrust those who add a little piece of AI to a system whose only structural purpose is to make sure that you can never leave. Authentic innovation does not strengthen chains. It breaks them.
What does an “AI-native” approach to enterprise data analysis really mean? The approach used by Oltrematica
When we talk about AI-native solutions, we are not talking about putting a chatbot on top of a database. We are talking about a radically different approach to enterprise technology. An approach that starts with three principles that should be obvious but clearly are not.
- The data are yours. Not the software provider’s. Not the cloud provider’s. Yours. And it must be exportable, readable, portable at any time, without penalty and without intermediaries.
- Intelligence must be modular. Today the best AI model for analyzing text is different from the optimal model for numerical predictions. In six months it will change again. An intelligent solution does not marry with a single AI vendor: it uses the best available at any given time, because the underlying technology is interchangeable.
- Complexity must be in the system, not in your day. If to get a piece of information you have to open four programs, cross-reference three Excel sheets and call the consultant, the system has failed. Not you. The system.
This is the approach we follow in every project we build at Oltrematica. We take our clients’ data, even data trapped in archaic formats and closed systems, and build on top of it a layer of intelligence that remains owned by the client, that the client can control, and that grows with his or her business instead of stifling it.
Agility in data analysis is not a luxury. It is survival.
We know what you are thinking. “Nice, but we are not a Silicon Valley startup. We are a company with 30 employees, and we can’t afford to experiment.”
We understand that. And we tell you: that is exactly the point.
You cannot afford experiments, it is true. But you cannot afford to stand still either. The world around you is accelerating in a way that is unprecedented. Your competitors (not the big ones, the small and hungry ones) are already using AI to respond to customers faster, to analyze data with more depth, to make better decisions with fewer resources. Not five years from now. Now.
The agility we are talking about is not the agility of management manuals. It is something simpler and more urgent: it is the ability to change direction when you need to, without having to ask permission from your software vendor. It is being able to try out a new idea in weeks instead of months. It is not having to choose between “let’s keep the system we have” and “let’s revolutionize everything and hope for the best.”
AI offers, for the first time, a third way: start with what you have, extract real value from it, and gradually build something better. Without a big bang. Without traumatic migrations. Without the eighteen-month project that becomes thirty-six and then “let’s see.”
A matter of trust, not technology.
In the end, the decision in front of you is not a technical one. It is a decision of trust.
Do you trust anyone more who tells you “stay here, we’ll add AI, don’t worry about anything”-knowing that that “don’t worry” means “don’t look at how much you’re paying and what you’re giving up”?
Or do you trust more those who tell you “the data is yours, the intelligence we build on it is yours, and the day you want to switch providers you can do so without anyone holding you hostage”?
We are not asking you to throw away your management system tomorrow morning. We are asking you for something much simpler: start demanding that your data be accessible. Demand APIs. Demand export. Demand to know what format your company’s 20 years of work is stored in. If the answer is an embarrassed silence or a mind-boggling quote, you already have the answer you needed.
The future does not belong to the biggest software. It belongs to the smartest companies. And “smart” does not mean having the biggest budget. It means knowing that your data has tremendous value, and refusing to leave that value in the hands of those who have every interest in not letting you use it.
Your data is your data. It is time to treat it as such.
If you want to understand with us what can be done, concretely, with the data you already have, without revolutions and unrealistic promises, let’s talk about it. That’s what we do every day.
Some FAQs on the AI-native approach to enterprise data analysis
Here are some questions that happen to come to us from business owners and IT managers when we explain to them the AI-Native approach developed by Oltrematica for enterprise data analysis.
We have grouped them into a series that FAQ that we hope will resolve any possible doubts.
But my management works. Why should I change anything?
If it really works, you should not change anything. The right question, however, is a different one: are you using your management software or are you suffering from it? If you need an outside consultant to get a report, if every customization takes weeks and four-figure quotes, if you cannot freely export your data-then the management software works for the vendor, not for you. The point is not to throw it away tomorrow, but to start using it strategically with tools that are yours.
I have no budget to replace the entire management system. What do I do?
No one is asking you to do this, and we ourselves advise against it. The AI-native approach is not a big bang: it is a gradual path. You start by connecting the data you already have-even inside closed systems-to an external AI layer that reads it, interprets it, and returns value to you. The management system remains in place, but it stops being the only access point to your company’s information. The costs of an initial exploratory project are a fraction of those of a migration, and time is measured in weeks, not years.
My data is inside a very old management system, in proprietary formats. Can something still be done?
Yes, and it is a more common situation than you think. Modern AI is remarkably capable of working with heterogeneous data: relational databases, exported CSV files, PDFs, even hand-filled Excel sheets. The first step is always an analysis phase where we map what you have, where it is, and in what format. From there we build an extraction and normalization plan. The data doesn’t need to be perfect: it needs to be accessible. And if they are not, that is the first problem to solve-and it is your right to solve it.
My management vendor just added AI functionality. Isn’t that enough?
Depends. Ask yourself three questions: can I use this AI with data that sit outside the management system? Can I take away results and trained models if I switch providers? Can I choose a different AI engine if a better one comes out? If the answer to all three is no, you are not getting innovation. You are getting a more elegant lock. AI added to a closed system is to make that system more indispensable, not to make you more autonomous.
What exactly is an “AI-native” solution?
It is a solution designed from the outset to leverage artificial intelligence as a core component, not as an ancillary feature added after the fact. In practice it means three things: data is stored in open and accessible formats; AI models are modular and replaceable (today you can use one vendor, tomorrow another, without rewriting anything); and the interface is designed for natural interaction, where you ask a question and get an answer, not where you navigate through twenty menus to find a number.
What about security? Where does my data go?
This is a sacrosanct concern, and we take it very seriously. In a well-designed solution, the data stays in your infrastructure or in dedicated, certified cloud environments. AI models can be configured to work locally or with providers that guarantee non-retention of data (i.e.: they process the request and retain nothing). No data should transit on shared platforms without your explicit consent. Data sovereignty is non-negotiable and is the first requirement of any architecture we build.
How long does it take to see concrete results?
An initial exploratory project-what we call a Proof of Concept-typically takes four to eight weeks. During that time, key data sources are connected, an initial layer of intelligence is built on a specific use case, and the system is put in the hands of those who will need to use it to validate its real-world usefulness. It is not a presentation prototype: it is a working tool on real data. If it works, you extend it. If it doesn’t work, you have invested little and learned a lot.
How do I know if my data is “accessible”?
A simple test: ask your vendor to export all your data in an open standard format (CSV, JSON, SQL dump). Ask for documentation on available APIs. Ask how much it costs to do this. If you get quick, clear answers at no or reasonable cost, you are in good hands. If you get silence, vague answers or disproportionate quotes, that vendor is protecting its own business model, not your interests.
Am I not in danger of depending on you in the same way that I depend on the current management?
It is the smartest question you can ask, and the answer is structural: everything we build uses open formats, documented code, and standard architectures. The data remains yours, the code is yours, the documentation is yours. If you decided to work with someone else tomorrow, you could do so by taking everything away. This is not a risk for us: it is an incentive to continue to earn your trust.
Where do we start, concretely?
From a conversation. You tell us how you work, what systems you use, where you feel the biggest frictions, and what you wish you could do that you can’t do today. We tell you what is feasible, with what timeframe, and with what investment-without selling pre-packaged solutions. Every company has its own data, its own processes, its own priorities. The starting point is never the same, and anyone who tells you otherwise is probably selling a product, not solving a problem.