Twenty years running IT projects and programmes: ServiceNow and ITSM rollouts in international corporate environments, MES and ERP systems in industrial and regulated settings, and a PMO built from scratch. In recent years, specialising in what decides whether an AI project can be defended or not: governance and compliance — the EU AI Act, GDPR, evaluating and documenting before deploying.
Work done
Four projects from my professional career and three of my own — the latter described as what they are: AI projects managed end to end.
Led the ServiceNow rollout programme across three continents, with stakeholders and delivery teams in non-overlapping time zones. Migration from legacy tools and standardisation of the service management processes, with phased delivery, vendor and integration management, and training for end users and IT teams.
Established the reporting frameworks and the KPIs to measure adoption, service performance and business impact.
Since 2023, specialising in ITSM from inside the platform — and this is where the programme meets governance: applying Predictive Intelligence to incident categorisation, and evaluating AI assistants and agents in a controlled environment before taking them to production. It is the practice I later wrote up as an article: evaluate and document before deploying.
Platform tooling: Flow Designer, Business Rules, UI Policies and Service Catalog.
Rollout of an MES adapted to a process with high cultivar variety and small batches, which is exactly where a standard MES does not fit. Integration with the ERP and the IT infrastructure to get a consistent data flow to report and decide on.
Coordination of international teams and vendors, and process standardisation across all subsidiaries.
A programme for SEAT's employee coaches across the province of Barcelona: 61 routes and three daily shifts. Fleet geolocation with Frotcom giving both companies real-time visibility, on-board wifi and time tracking integrated into the vehicles.
Coordinated the development of a mobile app to check coach position and estimated arrival time.
Core healthcare IT projects: the clinical case management system and the digital nursing admission procedure. Strategic projects with IT leadership, such as the reengineering of accounting and billing and the Oncology Institute's workflows.
Every solution had to meet JCI/JCAHO accreditation standards, with patient safety as a precondition.
Reads invoices, delivery notes, contracts and payslips and extracts the fields into the system where people already work.
Deployment on the user's own server or cloud rather than on a platform of mine. That is a compliance decision before it is a technical one: once a company's document base passes through a third party, its data processor changes, and so does the documentation that has to be maintained.
Silent extraction errors: that is why the design keeps human review on any field involving amounts or deadlines. No document travels to an external service.
GDPR by design (minimisation, processor under control). Minimal-risk use under the EU AI Act: no specific obligations beyond transparency about the use of AI.
The question that determined the architecture was not technical — which model extracts best — but one of accountability: who answers if the data comes out wrong, and where that can be evidenced.
Gathers in one place the contacts arriving through different channels, with the status of each and a weekly summary of those waiting too long for a reply.
Three service tiers instead of one, because the jump from "notebook" to "CRM" is what makes most small implementations get abandoned. The tier with AI-generated summaries is optional and does not condition the rest.
A generated summary omitting or distorting a commitment made to a person: that is why the summary always links to the original message instead of replacing it. Data hosted in the EU.
Data minimisation and purpose limitation. Transparency about which parts are generated automatically.
The value was not the model but the information architecture: ordering what already exists solves more than adding intelligence on top of it.
AI literacy material built around Article 4 of the EU AI Act: training proportionate to actual use, a comprehension test, named certificates and a compliance dossier.
Sizing the training to each organisation's real use rather than producing a single course, because the rule asks for proportionality and because a generic course evidences nothing.
The sector's risk is fear marketing: selling training by waving million-euro fines that Article 4 does not provide for. The material was built explicitly not to do that.
When the Digital Omnibus (Reg. (EU) 2026/1744) softened Article 4 — from guaranteeing a level of literacy to adopting measures to foster its development — the material became outdated and, in places, alarmist. The correction was published with a dated note visible at the top of each article. That is the most valuable part of this project: a rule changing mid-delivery is a management scenario, not an accident.
I also built iaetica.cat end to end: a static site in four languages on Cloudflare Pages with automatic deployment, technical SEO, GDPR and a newsletter with double opt-in.
Profile
Programme management and AI governance, on a foundation of ITSM and process automation.
Career
Healthcare, manufacturing, transport and international corporate environments.
Since 2021 · international, remote
Global ServiceNow rollout (ITSM, ITOM and Security Incident Response) across Europe, North America and APAC. Since 2023, specialising in ITSM from inside the platform and evaluating AI applied to the service.
2019 – 2021 · Barcelona
Rollout of Siemens SIMATIC IT MES for production control and traceability, integrated with the ERP, in a process with high cultivar variety and small batches.
2016 – 2018 · Barcelona
Employee transport programme with on-board technology for SEAT: 61 routes and three daily shifts across the province of Barcelona.
2015 – 2016 · Barcelona
Built the IT department's PMO: project governance, a common methodology, and portfolio tracking and reporting tools.
2012 – 2015 · Parets del Vallès
Process automation on Aura Portal BPM and involvement in the analysis and rollout of Microsoft Dynamics ERP.
2006 – 2011 · Barcelona
Clinical and administrative projects in a Joint Commission accredited hospital: clinical case management, digital nursing admission, and accounting and billing reengineering.
Since 2026 · Catalonia
Own AI projects applied to administrative processes, training material on Article 4 of the EU AI Act, and a weekly column on the ethics and law of artificial intelligence.
The common thread
AI governance is not a recent turn: it is the same craft applied to a new rulebook.
2006 – 2011
At Teknon, no solution went into production without meeting JCI/JCAHO standards, with patient safety as a precondition. That is where I learned to work with a compliance framework overhead, and to document for somebody who would review it later.
Since 2023
On ServiceNow, Predictive Intelligence for incident categorisation and evaluation of AI assistants and agents in a controlled environment. Evaluating an AI feature before deploying it is, to a large extent, a documentation exercise.
2026
When the Digital Omnibus rewrote Article 4 of the EU AI Act, the material and articles I had published became outdated. I corrected them with a visible dated note instead of letting them run.
Samples of judgement
Four articles that show how I read the rules and where I draw the lines.
What it actually requires, of whom and from when, with the timetable updated after the Digital Omnibus.
Read →What the training rule really asks for and what is commercial noise. Including September's dated correction.
Read →What must always be verified and how to explain it to a team without banning anything.
Read →How to spot them — and why you should not buy them either. The advice in both directions.
Read →Approach
A decision that cannot be reconstructed six months later — who, when, on what information — is not documented. That holds for a project steering committee and for the documentation the EU AI Act asks for.
A ten-person company does not need what a thousand-person company needs. Part of the job is working out what does not need doing.
When something published becomes outdated or was wrong, it gets corrected and the correction is recorded, with a date, on the same page.
Person
I started in IT when projects were still defended with a printed Gantt chart, and I have spent twenty years getting them delivered: international corporate environments, teams spread across several countries, and the paperwork that lets a decision be reconstructed six months later. Artificial intelligence entered my work the way it enters everywhere — through the back door, without anyone having decided anything — and I realised the question that interested me was not what a model can do, but who answers for it when it gets things wrong.
I am now finishing an MSc in artificial intelligence and the PMI-CPMAI certification in AI project management, and I write weekly about ethics and legislation: the EU AI Act, bias, traceability, and what all of it asks of an organisation with no legal department. What occupies me is not the tools but what surrounds them: making sure what has been done can be audited, repeated and defended. I am interested in AI governance — systems that do not carry bias forward, projects run on ethical criteria, and compliance designed in rather than bolted on once it is too late.
What I write each week is on the iaetica.cat blog.
Education
Languages: native Catalan and Spanish; full professional English and French.