Artificial intelligence with criterion.
How EstratégicaMente uses AI, under what conditions and what guarantees we offer. Compliance with Regulation (EU) 2024/1689 (EU AI Act).
This page last updated: 24 September 2026.
Legal content closed as at 8 September 2026, checked against the enacting terms and against our internal register of citations. Next revalidation: 8 November 2026.
Labels: anything stated as an obligation comes from the enacting terms and says who it binds; anything marked as our own criterion is EstratégicaMente methodology and is not required by the Regulation.
In plain language
- We use AI in our work, and we say so: our chat identifies itself as AI from the first message.
- No decision affecting you is taken alone: there is always a person who reviews it and answers for it.
- Your data is not used to train models. Neither ours nor our providers'.
- Before sending anything to a model, we minimise and pseudonymise personal data.
- We apply at home what we sell: a system inventory, recorded training and declared oversight.
- If you believe one of our AI systems has failed in something that affects you, write to us and we will review it with a person.
This summary helps you understand it quickly. It does not replace the full text below, which is the legally binding one.
Four non-negotiable principles.
Transparency first
We tell you which system we use, what for, and which provider is behind it. No small print.
Human oversight, always — beyond what the law requires
No report, analysis or document is delivered without prior professional review. AI proposes, EstratégicaMente decides.
Your data trains nothing
The data you provide is not used to train our own AI models or third-party ones. Full stop.
Honesty about the limits
AI can get things wrong. We accept that. That is why we review. And why we accept complaints within 5 working days.
EstratégicaMente · verified and documented compliance
Aligned with the timetable of Regulation (EU) 2024/1689. The Digital Omnibus, adopted on 8 July 2026 and published in the OJEU on 24 July 2026 (in force on 27 July), does not postpone Chapter IV: Art. 50 transparency has applied since 2 August 2026, and the Art. 4 AI literacy duty since 2 February 2025 — what changed on 27 July 2026 was its wording. It postpones Annex III high-risk systems to 2 December 2027 and Annex I to 2 August 2028. It also adds Art. 111.4, which allows until 2 December 2026 to comply with the Art. 50.2 marking, but only for providers of systems that generate synthetic content and were placed on the market before 2 August 2026. Anything made available after that date has no additional deadline: that is why our own systems that entered service from then on carry the marking from day one, as stated in the catalogue. The Art. 5 prohibitions already in force have applied since February 2025; the two new ones introduced by the Omnibus apply from 2 December 2026.
1 · Which AI systems do we use?
EstratégicaMente uses large language models (LLMs) as a professional assistance tool in the following services:
- Simplicity Advisory© — executive strategic advice with assisted analysis.
- Simplicity for Grants© — analysis of calls, eligibility assessment and document generation.
- Simplicity for Processes© — process diagnosis and AI adoption recommendations for SMEs.
- Simplicity Consilium© — support for strategic analysis and reporting for governing bodies.
- Simplicity LAB© — teaching support for training in AI management.
EstratégicaMente operates a vendor-agnostic artificial intelligence architecture: it does not depend on a single language model provider, but selects the most appropriate one for each task, always under the same guarantees of oversight and data protection.
2 · Risk classification
In line with the EU AI Act, EstratégicaMente has internally assessed its systems. None of those currently deployed falls within the high-risk categories of Annex III. Those that interact with people or generate content are classified as limited-risk systems, with the transparency obligations of Article 50. The CriterIA© Radar, which only reads public pages and neither interacts with people nor generates content, is classified as minimal-risk.
3 · Human oversight
AI at EstratégicaMente never acts autonomously nor takes final decisions. Our workflow:
The AI system produces an analysis or initial draft.
An EstratégicaMente professional reviews, cross-checks and validates it.
After human validation, the result is delivered to the client.
4 · Data protection
The data clients provide is processed solely to deliver the service. Relationships with technology providers are governed by data processing agreements and, where applicable, standard contractual clauses.
5 · Where processing takes place
EstratégicaMente is committed to full transparency about where data is processed. All storage, hosting and email infrastructure operates in the European Union.
- Database and authentication — data is stored in the EU region (Ireland) on infrastructure with SOC 2 Type II certification. The database behind the website, the CRM and the tools never leaves the European Union.
- Hosting and serverless functions — site content and server functions run in the EU region, with SOC 2 Type II certification.
- Transactional email delivery — email delivery operates from the EU region (Ireland). Account data, delivery metadata and activity logs may nonetheless be processed by the provider under standard contractual clauses.
- Language models · vendor-agnostic AI — EstratégicaMente operates a vendor-agnostic AI architecture: it does not depend on a single language model provider, but selects the most appropriate one for each task under the same guarantees.
- International transfers — the inference provider used as at this version is Anthropic, PBC, with processing in the United States. That transfer relies on standard contractual clauses and the supplementary measures described below.
Storage and the main processing of personal data take place entirely on European Union servers. As a general rule, no personal data is transferred outside the EU except for the inference calls described above.
Supplementary measures for international transfers · Schrems II and EDPB Recommendations 01/2020
Beyond contractual mechanisms, EstratégicaMente applies the supplementary measures recommended by the European Data Protection Board: data minimisation before any query, pseudonymisation of identifiers, and a ban on sending special categories of data to any model.
Express declaration on special categories of data · Art. 9 GDPR
No AI system deployed by EstratégicaMente processes special categories of personal data under Art. 9 GDPR: data on health, ethnic origin, political opinions, religious beliefs, trade union membership, biometrics or sexual orientation.
6 · Public register of AI agents
Article 50(1) of Regulation (EU) 2024/1689 requires the provider of an AI system intended to interact with natural persons to design it so that those persons are informed they are dealing with an AI system, unless this is obvious. We meet that duty where the interaction begins: in the first message of every assistant.
Publishing this catalogue is not required by the Regulation. No provision obliges anyone to make a register of their own systems public: it is EstratégicaMente's own decision, and the same criterion we ask of our clients under CriterIA©.
Naming of the agents. None of them takes on the identity of a person in the organisation. Those with a proper name —Sókrates, Solón, Breogán— refer to historical or mythical figures, never to an employee. The Regulation does not require this either: it is our own criterion, and it keeps the artificial nature of the interlocutor from depending on how the reader interprets a name.
| Agent | Purpose | Where it operates | Role and oversight |
|---|---|---|---|
| EstratégicaMente assistant | Guides on the six pillars and puts you in touch. It neither advises nor decides. | estrategicamente.info | Provider · identifies itself as AI from the first interaction · human escalation within 5 working days |
| Simplicity for Grants© (APEX · DRAFT · CRAFT) | Detection of calls, fit assessment and assisted drafting of technical proposals. | EstratégicaMente's own platform | Provider · mandatory professional review before every delivery |
| Digital Senior Advisors (NOVA · KAEL · VERA · LUMA · LEXIA) | Support for the team's strategic, financial, operational, brand and compliance analysis. | Internal use and Simplicity Consilium© | Provider · always identified as AI systems · human decision by the partners |
| Sókrates (Simplicity LAB©) | Maieutic agent for teaching use: it asks questions and refuses to give closed answers. | Campus and university classrooms | Provider · lecturer responsible in the classroom |
| CAUCE (EM-IA-021) | Turns a business idea into a first diagnosis of its fit with public funding. It gives no fit percentage and promises no award. | estrategicamente.info/cauce | Provider · human review before any commitment · in service since 15/08/2026, with the Art. 50.2 machine-readable marking in place from day one |
| Document reader of the management panel (EM-IA-022) | Finds data stated literally in the documentation provided and proposes it for human confirmation. It does not calculate, infer or write to the project record. | Internal use · management panel | Provider · no output enters the record without human confirmation logged with author and date · in service since 16/08/2026, with the Art. 50.2 marking in place from day one |
| Huella de IA© (EM-IA-024) | Free public checker: you type a domain and it returns an A to E score across fourteen controls. It measures what the organisation publishes, not what it does, and never says «non-compliant» but «not found». | huella.estrategicamente.info | Provider · in service since 09/09/2026, after 2 August and therefore with no transitional provision. Rule-based detection over the HTML served: it does not generate synthetic content, so Art. 50.2 does not reach it. Unlike the Radar, here there is interaction with a person, which is why it is presented as an automated checker from the home page and the report carries an express disclaimer. |
| Radar CriterIA© (EM-IA-023) | Reads public web pages of Spanish organisations —home page, up to two assistant routes, legal notice, privacy, cookies and one article— and records four objective signals about their use of AI. It is the system that produces the Índice de Criterio©. It does not access private areas and does not process personal data. | estrategicamente.info/indice-de-criterio | Provider · in service since 05/08/2026. Rule-based detection over the HTML served: it does not generate synthetic content, so Art. 50.2 does not reach it and the Art. 111.4 deadline is out of scope, which is not the same as being unprotected. Art. 50.1 does not reach it either: it does not interact with people, it reads served pages. The general obligation that does reach it is Art. 4. |
| CriterIA© 7 · CriterIA© 25 | Self-assessment of compliance with the Regulation and issuing of the report. | estrategicamente.info | These are not AI systems: the scoring is a deterministic calculation, with no language models involved. It is declared for transparency. |
This register is reviewed monthly from the internal AI control panel, where every agent is identified, classified and assigned to a responsible person.
7 · People, training and accountability
7.1 · AI governance owner
The AI governance owner at EstratégicaMente is General Management. This function maintains the register of systems, approves the introduction of each agent, reviews its risk classification monthly and is accountable for the human oversight decisions described in this policy. It is the point of contact for any matter concerning the firm's use of AI, at hola@estrategicamente.info.
Own criterion. Regulation (EU) 2024/1689 does not require a deployer of limited-risk or minimal-risk systems to appoint an AI governance owner. EstratégicaMente appoints one regardless: without an identified function there is no one to ask and no one to answer.
7.2 · AI literacy of the team
All EstratégicaMente staff who operate AI systems —management included— receive internal AI training before using them and maintain their competence through the firm's own learning pathway. Training is recorded, with its scope and date, in the same internal panel where the register of agents is documented.
EstratégicaMente is committed to maintaining the AI literacy of its team through continuous training, reviewed at least annually and whenever a new system is introduced or an existing one changes materially.
Applicable obligation. Article 4 of Regulation (EU) 2024/1689 requires providers and deployers to take measures to support the promotion of AI literacy among their staff. It has applied since 2 February 2025; its current wording since 27 July 2026. It is an obligation of means: the rule does not require guaranteeing a specific level, nor publishing the evidence. We publish this statement as our own criterion.
7.3 · Labelling of AI-generated content
Content published by EstratégicaMente that has been produced with AI assistance is expressly identified as such, stating that it has been reviewed and supervised by senior professionals who assume editorial responsibility. Images or illustrations generated or manipulated with AI are marked with a visible caption on the piece itself.
Own criterion. The marking obligation in Article 50(2) falls on providers of systems that generate synthetic content, not on those who use them. EstratégicaMente labels regardless: readers are entitled to know how what they are reading was made, irrespective of whom the rule binds.
Provider watermark. Anthropic, the provider of the models we use, signed the EU Code of Practice on Transparency of AI-Generated Content and marks what its models produce: text carries an imperceptible watermark in word choice, which neither changes meaning nor adds characters, and generated files carry C2PA Content Credentials. Coverage is being extended across its models in phases, with all covered by 2 December 2026. It is the provider's mark, not ours: it does not replace our own express labelling, which is why our AI-assisted texts always say so, even though the mark may be lost when they are edited or converted.
8 · User rights
You have the right to:
- Request the involvement of a human EstratégicaMente professional in any AI-assisted process.
- Obtain explanations about how the system works in general terms.
- Complain if you believe a result is incorrect or inappropriate.
9 · Contact and complaints
For any enquiry, exercise of rights or complaint regarding the use of AI at EstratégicaMente:
EstratégicaMente
Vía Norte, 12 Bajo · 36204 Vigo (Pontevedra)
N.I.F. B70399449
hola@estrategicamente.info
EstratégicaMente guarantees a reply and human review within a maximum of 5 working days.
10 · Updates
This policy is reviewed periodically as the applicable rules and the systems in use evolve. The date of the latest update appears above.