Before
Individual test
Each participant writes «yes», «no» or «I do not have enough evidence», and notes one fact that supports it. Without consulting each other.
Book
Antes del próximo piloto
The four foundations a board must audit to reduce the risk and accelerate the value of AI
Héctor Enríquez
In preparation. Editorial cut-off date: 17 July 2026.
The book is written in Spanish.
The thesis
AI does not build the foundations that are missing. It amplifies the ones it finds.
The business value of AI is conditioned by the systems, the data and the controls it connects to. A capable tool can give a remarkable demonstration and still fail as an investment: the technology works and the business gets no value.
So this is not an argument against AI. It is an argument against investing as if the model acted alone.
The mechanism
The chain that turns capability into value
- Capable model
- Reliable context
- Integration with the process
- Decision or action
- Adoption
- Measurable result
When an AI project «did not deliver», the first link almost never broke. One of the other five did, and nobody was looking there.
Contents
Seven chapters
- 01The uncomfortable truth: productivity is not value
- 02Pilot purgatory: why initiatives do not scale
- 03The thesis: AI reveals the structure it finds
- 04The four foundations: how to judge whether your company can carry artificial intelligence
- 05The investment decision: sequence before platform
- 06When foundations fail in production: risk, responsibility and control
- 07The next wave: why agents punish weak foundations
Each chapter closes with a board decision and a working session with its question, its participants and its deliverable. It is a board decision framework, not a consulting methodology and not a return guarantee.
To use today
The five board questions
Each one is answered for a bounded use case, not for «the company» in the abstract, and each answer has to come with evidence, not adjectives.
- 01
Connection
For this use case, do the systems exchange information and actions with the latency, the controls and the reliability the decision demands?
«We have an API» and «we load every night» are descriptions, not conclusions. A bad connection turns an AI project into an unbudgeted integration project.
- 02
Data
For this use case, do we have a shared definition, a traceable source and enough quality to make the decision?
«Customer», «sale» or «available inventory» can mean different things depending on the system and the department. AI does not correct that ambiguity: it processes it and scales it.
- 03
Documentation
For this use case, can we understand, operate and validate the critical systems without depending on one person's memory?
You can hire knowledge of the language; you cannot immediately buy decades of your own context.
- 04
Security
For this use case, do we know what can be attacked or abused, who can act, what risk we accept and how we will recover?
An agent with write permission must be governed like an identity with operational capability: least privilege, action logging, limits and a way to stop it.
- 05
Adoption
Who will change the way they work, why does it help or hurt them, and what evidence will show that the new process is used and produces the expected result?
If you cannot point to which decision, task or control disappears or changes, there is no adoption design yet: there is a purchase.
Tool
The 15 minute test
Before approving the next AI investment, the board can run this exercise. It needs no technical preparation and usually gives its most valuable finding at minute four. The investment is identified beforehand and the question answered for that case: are we prepared to reduce its risk and accelerate its value?
0 to 4
Reveal
Everyone shows their answers at the same time. They are not averaged and not corrected.
4 to 11
Discuss
The board contrasts the differences against the evidence. Each statement is classified as verified fact, assumption or missing data.
11 to 15
Decide
The board picks one of two routes: take the case to a foundations diagnostic with evidence, or pause its approval until the missing evidence is gathered, with a date.
If everyone answers the same with evidence, there is a common basis to decide. If they differ, the session already produced value: it revealed a visibility gap before it became an investment gap.
Living annex
Updates and errata
Last reviewed: 21 August 2026
The book points here for what can change after printing: regulation, prices, vendors and volatile figures. Check them here and against the primary source.
Dated figures and references
Updates to the market figures, vendor prices and regulatory references cited in the book, each with the date it was verified.
No updates published yet. The book is in preparation.
Errata
Corrections to the printed text, with chapter, page and the corrected version.
No errata recorded.
I will let you know when a new report is out.
One email per report, no forced cadence and no sponsored content. If a month brings nothing worth sending, nothing arrives.
Only to announce our own publications. You can unsubscribe by replying to the email.