The bridge, reassembled

Editorial note. This chapter preserves the earlier concise English edition. A full translation of the expanded Spanish edition published on July 21, 2026 is in progress.

The book held a single idea and carried it to the end. At the close it is worth reassembling it: seeing at a glance how each part holds up the others, because it is that interlock —and not any loose claim— that constitutes the argument.

The arc, in one page

It began with the problem: organizations arrive at artificial intelligence from different starting points —the one that believes it has nothing, the one of disconnected islands, the mature but fragmented one—, and almost all of them stumble in the same place, because they treat as technical a problem that is sociotechnical. Between those who read people and those who build the system, a translation gap opens up, and that is where the project is lost. → The bridge thesis.

It continued with the method: closing that gap is not achieved with an intermediary who coordinates, but with an integrated competence that reads and builds without the meaning being lost in the handover. That competence is ordered into four links —sociology, software, data, AI governance— exercised as a loop without handoffs; it is spoken in a vocabulary of its own sharpened for the work; and it starts with a diagnosis that reads the organization before touching the technology, and that has the courage to say “not here”.

It rested on foundations: eight currents that do not frequent one another —from the Tavistock coal mine to human-centered AI— and that, nonetheless, converge on the same point. That disciplines so distant arrive at the same thing is what makes the thesis a finding and not an opinion. → Authors and currents. And it proved itself against the new: when software stops suggesting and moves to deciding, the thesis does not break, it shifts one rung upward. → The agentic extension.

And it came down to applications: to small business, where poorly applied AI sinks precisely the one that needs it most, and that is why the bridge begins by guarding against the harm; to the State, where what is at stake is not money but rights; and to a measurement instrument that makes rule number one —measure maturity before buying— falsifiable.

The thesis, in its mature form

With everything reassembled, the thesis can be stated more strongly than at the beginning. It is not only that technology adoption is a human problem before a technical one; it is that that structure of the problem does not age with the technology, it sharpens with it. Each rung an organization climbs —from the spreadsheet to the system, from the system to prediction, from prediction to the agent that decides on its own— makes the omission of the human reading more expensive, not cheaper. At the bottom, a poorly understood system is avoided: people go back to their spreadsheet. At the top, a process no one understood is not executed more slowly, it is executed on its own, at scale, and with no witness. That is why the competence to read the human fabric and build the system at once is not an asset that AI renders obsolete: it is, precisely, what its autonomy renders indispensable.

Hence value is never measured in the model, but in the person: an hour recovered, a decision that comes out better, a gap that is closed. When the metric shifts from the artifact to the person, it becomes impossible to confuse activity with result —and almost every project that does not pay off lives in that confusion—.

And now what

The book does not close on a conclusion the reader contemplates, but on one that interpellates them, different depending on where they read from.

If you run a small business, the first thing is not to choose a tool but to measure where you start from and resist the pressure to “get on board with AI” without a diagnosis that says where it pays off and where it only destroys value. The cleanest opportunity is, almost always, in what looks like backwardness: there is order to be built and little inheritance to dismantle.

If you lead a public body, the bar is higher because the cost of error is paid in rights. The question is not which system to buy, but which institutional capability to strengthen so that AI frees the public servant rather than replacing them, and so that every decision affecting a person is auditable.

And if you are someone who builds —a professional, a team—, the invitation is to cross the boundary the market keeps comfortable: not to stay at diagnosing without building, nor at building without reading, but to exercise both readings as a single practice. It is scarce by structure, and that is why it is worth it.

How the pieces fit (dependency map)

No chapter of the book stands alone; nor must you read them all with equal weight. This table is not an index: it shows what needs what so the argument does not hang in the air.

If you want to understand…	Start with…	Go deeper in…	Take to the field in…
Why projects fail	Thesis	Authors	SMEs or State
How work is done in practice	Four links	Diagnosis, Concepts	Distribuidora Norte (case throughout the book)
Whether to buy or scale AI	Diagnosis	IMIA	Verdict it pays off / not yet / destroys value
Whether the intervention was worth it	Metrics	Hygiene	Five-axis dashboard (prioritize 2–3)
What changes when AI decides alone	Agentic extension	D6 gate in IMIA	Judgment in the design in Concepts

Self-assessment (fifteen questions)

Answer honestly about your organization or a concrete project. There is no score: these are signals of where the method prescribes reading before building.

Can you name the declared problem and the real problem in one sentence each?
Who loses power if this works? Did you map it or are you discovering it in production?
Does the key data point mean the same thing to operations, sales and systems?
Is there a business owner who can answer “what is an active customer?” without hesitation?
Does the resistance you see carry a valid operational reason or only habit?
Is the pilot used by innovators or by the majority who need it to work without friction?
Is there a written AI use policy, even a minimal one?
Could you export your data if the vendor raised prices forty percent?
Did you measure a baseline before promising a percentage improvement?
Does the dashboard answer a Monday decision or only decorate a meeting?
Who maintains the glossary when a process changes on the floor?
Is there a second lead if the person driving the change leaves?
Would automating this duplicate effort —part in the tool, part by hand?
If an agent decided alone tomorrow, would you know who is accountable for its rules?
Is the value you seek measured in the person —hour, decision, autonomy— or in the model?

If many answers are “no” or “I don’t know”, that is not failure: it is where the diagnostic method begins. Distribuidora Norte, on its second pass, failed several of these before unifying delivery; that is why the book insists on measuring before scaling.

A book that knows itself unfinished

One final honesty remains, the same one with which the book began. A good part of what holds up this argument is one’s own construction —the diagnostic method, the concepts, the agentic extension, the maturity instrument— and it is presented as such, not as a finding borrowed from any author. The instrument still awaits its first field data: today it measures configuration with a design scoring, not calibrated, and saying so is part of the rigor. The practice the book describes scales slowly, because it rests on a judgment that transfers as a craft, not as a manual; admitting it is also part of the rigor.

None of that weakens the thesis. It strengthens it, because it puts it where the brand demands it be: in the facts, not in the adjectives. This is a living book —every claim that carries data carries its source, and it grows with each revision—. The bridge is not a finished idea the reader receives: it is a practice that awaits them on the other side.