The bridge between people and systems.
The risk is not automating; it's automating without understanding people.
Digital transformation fails with a frequency the industry would rather not look at — and almost never for lack of technology, but because the technology was installed on top of an organization no one had read. The problem is not artificial intelligence; it's the lack of social intelligence to know where and how to apply it. My work is to inhabit that crossing —between society and technology, between people and systems— and to walk it end to end: reading the human fabric and building the technical system, without delegating either of the two.
Three ways of naming the same pair. The left half is the one almost no one reads; the right, the one almost everyone buys. The bridge is having both —and dropping neither.
The thesis. Every organization is a sociotechnical system: you cannot change its technology without first reading its human fabric. Whoever can read both —and build in both— is the bridge that transformation needs and almost never has.
The methodFour links, a single loop
Not four loose specialties: four stretches of one path, from social intelligence to artificial and back to a person's decision. The difference from a multidisciplinary team is that there are no handoffs: the four disciplines are exercised as a single practice, so the hypothesis that comes from reading the person is the same one that models the data and designs the AI, and it does not degrade at the edges. That continuity is the product.
gonzalo@flores:~$ ./method --loop
[1] sociology ▸ [2] software eng. ▸ [3] data eng. ▸ [4] AI governance
reads the org builds the reliable data value back to
(human problem) system (the (in its context) the decision
hypothesis (from data to
becomes a tool) the person)
▲ │
└─────────── value returns to the person ◀─────────────┘
and reopens the loop — no handoffs Read the organization
Reads it as a system of actors and networks: traces the real processes —not the manual's—, their inputs and outputs, how information flows and with what codes and languages. Maps culture, incentives and power, and separates the stated problem from the real one to rule where AI pays off and where it only destroys value.
Build the system
Turns that reading into auditable, maintainable software: secure backends, versioned APIs, MCP servers, with the discipline of a regulated environment.
Data that doesn't lie
Guarantees that data exists, is reliable and meaningful in its context. Without quality data there is no AI: it only automates errors faster.
Return value to the decision
Designs AI that amplifies the person's judgment rather than replacing it. Closes the loop: from data to the human decision, with auditable governance controls. When AI acts on its own, judgment doesn't vanish: it moves to the design and audit of the rules the agent decides by.
Those four links are the method —how I think—. As a service they come in three planes: platform (build), governance (govern) and adoption (enable). Each plane rests on the full loop; none is a loose box. You can see the three on the home page.
What I define myself againstThe bridge, not the easy way out
Each pair is a fork in the road, not a matter of degree. On the left, technological solutionism: the premise that every problem yields to more technology. On the right, the sociotechnical premise I hold: technology only works when it's designed together with the organization that will use it. These aren't two styles of the same craft, but two rival theories of what makes a system work.
Concepts of the paradigm
The market hole between those who understand people but don't build, and those who build but don't read people. That's where transformation projects die; that's where the bridge works.
A technically flawless datum that still doesn't say what people think it does, because its meaning depends on a human context that got lost. Almost always, the footprint of a poorly incentivized process.
The liability that accrues when technology is installed without resolving the social subsystem. It doesn't show up in the code; it's collected in zero adoption, sabotage and rejection.
Good AI gives the person decision-making power back; bad AI takes it away. Especially in the public sector: it amplifies the worker, it doesn't replace them.
From the spreadsheet to agents that act on their own, adoption is climbed one rung at a time. The higher the rung, the more costly it is to have skipped the human reading: that's why agentic AI doesn't make the paradigm obsolete—it makes it more urgent.
The same tool amplifies whatever it finds: over order and judgment it multiplies value; over disorder, with no judgment to filter it, it multiplies harm. The bridge is what sets the sign.
SMEs and the public sector
I measure maturity first: a sociotechnical diagnosis that detects where AI pays off and where it only destroys value, without large investments. The evidence demands prudence: in an experiment with 640 entrepreneurs in Kenya, generative AI improved high performers (~+15%) but worsened low performers (~−8%) —without human judgment to filter it, AI widens the gap instead of closing it.
I don't install systems: I redesign state action from the citizen's point of view —mapping civic processes, data governance and auditable AI—. In the public sector, what cannot be audited cannot be used. AI frees the civil servant's capacity; it doesn't replace them.
In Latin America, AI penetration does not reach 4% —against more than 20% in Europe— and within each country the distance between large firms and SMEs is enormous (in Brazil, 41% vs. 11%). That gap is the opportunity and the mission at once. Source: CEPAL, 2024.
What it rests onFoundations
Not a consultant's hunch: a framework supported by research with decades of backing.
- Sociotechnical systems (Trist & Bamforth, 1951; Mumford, ETHICS method) — the social and technical subsystems are optimized together or fail together.
- Actor-network theory (Latour, 2005; Callon) — technology stabilizes when it manages to translate and align the interests of the actors.
- Human-Centered AI (Shneiderman, 2020; Dignum, 2019; O'Neil, 2016) — AI must amplify and give control to people, not isolate them.
- Contextual data quality (Wang & Strong, 1996; Redman, 2008) — quality is fitness for use; without social context, data lies.
- AI readiness (Jöhnk et al., 2021) — readiness is strategy, culture, talent and governance, not just infrastructure.
The full development —thesis, evidence, method and foundations— lives in the digital book The Sociotechnical Bridge; its core is distilled in the essay An organization's digital transformation is not a technical problem.
Social intelligence for your artificial intelligence
Sociotechnical diagnosis, data platform, or AI governance and adoption. For companies, agencies and institutions.