Preface

There is a scene that repeats, almost without variation. An organization —a small business, a city government, a whole division of a large company— decides to take the leap: at last, technology; at last, artificial intelligence. There is enthusiasm, there is budget, sometimes there is even a good vendor. And a few months later the enthusiasm has cooled, the budget is spent, and the system that was going to change everything is still there, switched on, with no one using it. Nobody wanted it to end that way. It was not bad luck or clumsiness on people’s part. It was a misunderstanding about what kind of problem was being solved.

This is a book about that misunderstanding and about how not to commit it. It holds a single idea, carried to its consequences: that transforming an organization with technology —and today, with artificial intelligence— is first of all a human problem, and only afterwards a technical one. Whoever treats it as a technical problem fails in an almost predictable way. Whoever learns to read the human fabric and to build the system occupies a position that is rare, valuable and —worth saying from the start— deeply satisfying to inhabit. The pages that follow explain why, and how.

Why this book: the point of the reflection

It is easy to talk about technology adoption in the cold language of productivity: rates, efficiencies, return on investment, competitive advantage. All of that is true, and this book does not dodge it: there are pages devoted to the economic gap —measurable, and ever wider— that artificial intelligence is opening between the organizations that absorb it and those that stand watching. The economic dimension is real and serious; ignoring it would be naive. But staying only there is to miss what matters most.

Because behind every “project that did not pay off” there are people. There is the employee who learned to do the job one way and is told, overnight, that now it will be done another —and who suspects, not without reason, that the underlying plan is to make them redundant—. There is the owner of the small business who bet years of savings on a system they were promised and cannot understand why their own people avoid it. There is the public servant who genuinely wanted to serve the citizen better and ended up with a digital procedure slower than the paper one. Technology never touches only technology: it touches the way people spend their days, earn their living and feel —or stop feeling— that they are good for something. Designing those systems badly is not an abstract technical error; it impoverishes, a little, the lives whose working hours those systems are part of.

Artificial intelligence raises the stakes. It is not one more tool among others: it lands exactly where human judgment used to be —the decision, the criterion, the craft—. That is why it stirs, at once, so much hope and so much fear; and why it deserves something better than fashionable enthusiasm or reflexive rejection. It deserves someone willing to ask, case by case: does this amplify what this person knows how to do, or replace them without replacing what they contributed? Does it make them more capable, or merely more dispensable? A model does not answer that question. It is answered by someone who sat down to understand the real work of real people, and who takes responsibility for the answer.

There is, besides, a simpler and more personal reason, and it is better said plainly. Solving a real problem —one that was making someone’s life harder, and that after your intervention stopped doing so— is one of the most human and most deeply satisfying experiences there is. It is not about “putting a model into production”: it is about easing a task, freeing up an hour, making a decision a little fairer because it now rests on data that does not lie. That is the ultimate point of this craft, and the reason it is worth doing well before doing fast. The bridge this book describes is not only a scarce market position: it is a way of working that refuses to treat people as a detail to be resolved after the technology. People first. Technique, with all its rigor, comes to serve them —not the other way around—.

Where one starts from

This book takes no starting point for granted, because there is not just one. Organizations arrive at this conversation from very different places, and it is worth naming them, because every reader will recognize themselves in one.

At one extreme is the organization that feels it has almost nothing. And I say “feels”, because pure zero hardly exists anymore: today it is practically impossible to operate without some information system. There is a phone where the real work is coordinated, a spreadsheet someone guards jealously, the tax system, a chat group that is, in fact, the true operating system of the organization. What is missing is not technology: it is design. Those systems grew on their own, improvised, living in people’s heads and in conversations that leave no trace. The challenge here is not to digitize for the sake of it, but to build a first backbone without crushing the informal thing that —against all odds— was working.

In the middle is the organization of islands: systems bought over the years that do not talk to one another. Billing on one side, the stock spreadsheet on another, a customer system nobody updates, the master file only one person understands and on which, secretly, the whole operation depends. The data exists, but fragmented, and each area has its own version of the truth. The challenge is no longer to have; it is to integrate, to get those islands to form a coherent continent.

And at the other extreme is the mature but fragmented organization: it already has solid systems, ordered data, capable teams. What it needs is not to buy more tools —it has plenty— but to orchestrate what it already owns and take the leap: turn that data into better decisions, add artificial intelligence where it genuinely adds judgment, govern all of it with confidence. It is, paradoxically, a leap as delicate as the first, because here the risk is not scarcity but inertia: many systems, many vested interests, much “we have always done it this way”.

What unites these three points —and all those in between— is that in none of them is the decisive question technical. It is not “what do I buy?” nor “which model do I use?”. It is always the same sociotechnical question: what does this group of people do, why do they do it this way, and what will really change in their work if this moves? The bridge serves across the whole spectrum. What changes from one end to the other is not the principle but where the emphasis falls: build, integrate or orchestrate. The method this book proposes adapts to all three.

And there is a second axis, as important as the first: not only where one starts from, but toward where one leaps. Because technology adoption is not a switch that turns on all at once, but a ladder climbed step by step. At the bottom is the everyday office work —the spreadsheet, the document, the email—. One step up, the systems that order the operation: an ERP that integrates administration, a CRM that gathers the customer relationship in one place. Higher still, predictive artificial intelligence, which uses accumulated data to anticipate —demand, a customer’s churn, a risk— instead of merely recording the past. And on today’s step, what has come to be called the agentic revolution: software that no longer only suggests, but executes tasks end to end, and that pushes a new wave of productivity built on automating digital work.

Each step is a leap to a capability the organization did not have, and in all of them the temptation is the same: to believe the leap can be bought. It cannot. The higher the step, the higher the risk of climbing it without first reading the human fabric —and the agentic wave takes it to the extreme, because an agent automating a process nobody understood does not make the errors faster: it makes them on its own, at scale, with no one watching. The promise of productivity is real; so is the condition for collecting it: that each step be climbed on top of an organization someone took the trouble to understand. The ladder is climbed with method, not in a single jump.

What this book is

It is not a tool manual or a catalog of success stories. It is the ordered, argued exposition of a paradigm: a central thesis —the sociotechnical bridge—, the method that puts it into practice by articulating four disciplines (sociology of organizations, software engineering, data engineering, and AI architecture and governance), the vocabulary with which that method thinks, the academic foundations that hold it up, and the concrete applications to small business and to the State. One idea, unfolded carefully to the end.

It is, besides, a living book. It is under construction and gains depth with each revision: some chapters are already developed in detail, with their apparatus of citations; others are still working versions that will be expanded. Each chapter honestly declares its degree of maturity. It is published in the open precisely because its value lies in being improved in plain sight, and not in pretending to be finished. A book that argues one must read people could not, itself, hide from the reader.

Who it is for

For whoever decides on technology in an organization and suspects, deep down, that the problem was not in the tool. For the technical professional who senses that the best system they ever built goes unused for reasons that appear nowhere in the code. For the public servant, the small-business owner, the sociology or engineering student looking for a framework that does not split people from data as if they were two worlds. And for whoever wants to examine the soundness of a way of thinking rather than a résumé: the book is, also, that thinking put to the test.

The promise: rigor, not hype

A single rule governs these pages: every claim that carries data carries its source, and only what could be verified is published. Where a number is an estimate or a design decision, it is said in so many words. There are no filler figures or “studies say” without a reference. The sources are gathered, in author-date style, in the Bibliography. This is not academic pedantry: it is a matter of respect. The cost of a single invented citation is the credibility of everything else, and someone who writes about trust cannot afford to betray it in a footnote. That is why rigor is not an ornament of the argument: it is part of the argument.

How to read it

It can be read straight through —the order of the parts follows that of the reasoning— or chapter by chapter, entering wherever one’s own problem presses hardest:

• The sociotechnical problem sets out the thesis and its proof. If only one chapter is to be read, let it be this one.
• The method brings the thesis down to a concrete procedure: the four links, the proper vocabulary and the diagnostic protocol.
• Foundations lays out the currents and authors the thesis rests on, without hiding their tensions, and its extension to AI with agency.
• Applications takes it to the field: small business, the public sector and the maturity instrument, IMIA (AI Maturity Index).

The search box (the / key) runs across the whole book and jumps to the exact passage. The previous/next navigation follows the thread of the argument, for those who prefer to go hand in hand.

On the origin

This paradigm was not built in the abstract, at a clean desk. It came out of an integration uncommon in the market: that of the sociology of organizations with more than a decade of software and data engineering in productive, regulated environments, where systems have to truly work and be accountable. That biography explains where the framework comes from, but it is not its subject: the book sets out the paradigm, not its author. Its validation rests on two planes —the academic foundations it cites and the practice it was born from (systems in production, MCP servers, the IMIA maturity model)—, and not on a personal track record.

What follows is the thread of the argument, from the beginning. If all goes well, by the end of the book the scene from the first paragraph —the switched-on system no one uses— will have stopped looking like bad luck and become what it always was: a problem that could have been read in time.