Director of AI implementation in operations
Twelve years turning large operations into systems you can measure. I have run the operation of a multinational education group —five cities, 26 programmes, more than 52,000 students— and today I answer for an operation of 1,150 people and a budget of more than a million dollars a month. I have taken artificial intelligence from promise to process: 40% optimisation and 66% cost reduction, with AI governance, indicators and automation that keep working when I am not there.
I use language models where they earn their place — drafting, summarising, classifying text — and leave them out where the result has to be explainable. The tools below are the second kind: they compute with rules, in your browser, and every step can be audited.
Where all of the above comes from. The data is what Iron publishes on his own site.
CUN — Corporación Unificada Nacional de Educación Superior
Operating strategy for a multinational education group: five cities, 26 programmes, over 52,000 students. Digital transformation, applied AI and executive indicators for the president’s office.
Results: 40% process optimisation through generative AI and re-engineering. 66% cost reduction via Lean Six Sigma. Multi-site scaling without adding headcount.
National entrepreneurship and management-innovation initiatives, connecting academia, the business ecosystem and sustainable models.
Results: “Biomolécula Empresarial” model: a management-innovation framework delivering over 66% financial savings through investment optimisation.
Full innovation project lifecycle: AI applied to intelligent tutoring, automated assessment and learning personalisation.
Results: Adoption, quality and pedagogical-return indicators with dashboards for deans. The MELDS model as a leadership framework for innovation teams.
Iron Fuentes Consulting
Consulting for C-suites and boards: bringing AI into operations, data-driven optimisation, efficiency audits and executive dashboards.
Results: AI adoption roadmaps with measurable return, governance and ethical frameworks, and recommendations ranked by impact and effort across higher education, public services, logistics and technology.
UNIEMPRESARIAL · Fundación San José
Training leaders who decide with data, using the MELDS model. Applied research and innovation groups, plus one-to-one mentoring.
Results: Curriculum in generative AI, cloud and modern architectures, tied to real digital transformation cases.
Servicios Postales Nacionales 4-72
National operational rollout projects across a public logistics network.
Results: Coordinated territorial rollout with progress traceable site by site.
Each one answers a concrete problem. The left column is the problem; the right one is what the method contributes.
Lean Six Sigma / DMAIC
Rework, errors, delays and hidden costs.
Diagnose, measure, find root cause, improve and control.
Scrum
Slow projects with nothing visible shipped.
Sprints, clear roles and measurable incremental delivery.
Kanban
Invisible work, overload and bottlenecks.
Visible flow, work-in-progress limits and cycle-time metrics.
OKR
Teams with no focus or shared direction.
Ambitious objectives with key results measurable each cycle.
KPI / Balanced Scorecard
Decisions made blind, with no indicators.
Actionable indicators and executive dashboards.
PMBOK
Projects with scope, cost and schedule out of control.
Formal lifecycle, risk and stakeholder management.
Design Thinking
Solutions nobody uses because they ignore the user.
Empathy, ideation, prototyping and early validation.
Lean Startup
Big bets placed without testing the hypothesis.
Minimal experiments, learning metrics and timely pivots.
AI Governance
AI adopted with no control, risk view or policy.
Responsible-use frameworks, risk and human review.
Data Governance
Data scattered, poor quality and nobody owns it.
Quality, traceability and information governance.
MELDS
Teams without the mindset or skills to decide with data.
Mindset, Experimentation, Leadership, Data, Skills. His own model.
Critical operating processes, slow and manual, across a multi-site operation.
Re-engineering with generative AI, Scrum/Kanban and automation.
40% sustained optimisation
DMAIC · Kanban
A growing cost structure threatening sustainability.
An end-to-end Lean Six Sigma operational efficiency programme.
66% annual reduction without touching quality or coverage
Lean Six Sigma · DMAIC
Leadership deciding without consolidated information.
Executive dashboards, payroll analysis and organisational redesign.
Data-based decisions at board level
OKR · KPI · Balanced Scorecard
A need to adopt AI without exposing data or creating risk.
Use cases, governance, vendor selection and success indicators.
An institutional AI strategy with ethical frameworks
AI Governance · NIST
A teaching model disconnected from emerging technology.
Intelligent tutoring, automated assessment and AI personalisation.
Measurable pedagogical return and faculty adoption
MELDS · EdTech
An operation across five cities with no traceability or visible owners.
An operational management and traceability system.
Scaling without adding administrative headcount
Scrum · OKR · KPI
Doctor of Education (EdD)
UMECIT · 2026
Master’s in Teaching
Universidad de La Salle · 2022
Specialisation in Data Analysis
CUN · 2026
Specialisation in Digital Transformation
CUN · 2026
Specialisation in Educational Innovation
Fundación Universitaria Panamericana · 2019
Business Administration (in progress)
CUN · 2026
Bachelor’s degree in Computing and Technology
Universidad Minuto de Dios · 2017
Psychology (in progress)
—
Real processes, each with its diagram, its error handling and a demo you can use right here.
You are the person accountable for dropout rates at a university. Today you find out a student has left once they have already left. With this you open it on Monday, see nine names in red and call those nine.
See the case →
You are the person running a mid-sized company. Today you know something is wrong in the operation, but not where, nor what to tackle first. With this you answer a questionnaire and get back a plan with priorities and indicators.
See the case →
Tools that run on rules, formulas and catalogues. None of them calls an artificial intelligence model, so they cost nothing to run and every result can be audited step by step. Everything is computed in your browser: what you paste never leaves your computer.
17 tools designed: 14 working and 3 pending · 250 rules · 1,446 catalogue rows
Paste your spreadsheet and it returns duplicates, broken emails and impossible dates, with the file cleaned.
P01 · 22 rules · 17 package cases
Try it →
Paste a long, messy prompt and it reorganises it, flags what is vague and counts the tokens.
P02 · 15 rules · 15 package cases
Try it →
Set retries and outages and it tells you how many events you lose and how long recovery takes.
P03 · 16 rules · 15 package cases
Try it →
Describe a manual process and it works out the yearly cost and what automating it saves.
P04 · 15 rules · 14 package cases
Try it →
Paste a webhook payload and it shows the structure, the types and the fields that shift.
P05 · 15 rules · 15 package cases
Try it →
Paste your list of people with indicators and it tells you who to attend to first, and why.
P06 · 15 rules · 14 package cases
Try it →
Paste two versions of a sheet and it tells you which row and which cell changed.
P07 · 14 rules · 14 package cases
Try it →
Write a cron expression and see the next runs, daylight-saving jumps included.
P08 · 12 rules · 13 package cases
Try it →
Give matching and non-matching examples and it returns the regular expression, explained.
P09 · 14 rules · 13 package cases
Try it →
Compare two API versions and it separates what is compatible from what breaks consumers.
P10 · 15 rules · 14 package cases
Try it →
Paste your projects and it returns the order by impact, effort and confidence.
P11 · 14 rules · 13 package cases
Try it →
Draws your value curve against competitors and builds the matrix of what to drop and what to create.
P12 · 13 rules · 12 package cases
Try it →
Paste your colours and it reports contrast, colour blindness and which corrected shade to use.
P16 · 14 rules · 14 package cases
Try it →
Describe your case and build the flow: it tells you what to build it with, which connections do not hold and which controls are missing.
L01 · 8 rules · 21 package cases
Try it →
These already have rules, catalogues and tests written in the specification package; what they lack is the engine and the screen. They are listed because the scope is defined, not because you can use them.
Résumé analyser
Upload CVs, set your criteria and it shows where each one appears. It scores nobody.
P14 · 16 rules · queued
Email tone corrector
Fixes the tone of an email and strips out what undercuts your authority.
P15 · 12 rules · waiting on the tone tables
Organisational diagnosis
Its questionnaire belongs to Neurona, a separate product with its own scope. Bringing it here would mean duplicating the logic in two places and maintaining both.
P13 · out of scope
If something here is useful for your operation, get in touch and we will look at it.
See the profile and message me
Full track record, recommendations and direct contact.
Half an hour to talk about your operation.
Next.js and TypeScript, with no third-party charting library: the graphics are hand-written SVG. The computation lives apart from the interface and is tested without a browser. There are no accounts, no data server and nothing of yours is stored.
What is checked before publishing — the command, the commit and the result of all 368 tests.