Skip to main content
The AI Guys

Large organisations

AI your board, procurement and data protection can sign off

We build private deployments and air-gapped systems for better quality of work at lower cost, involving each department that signs off from the start.

What we build for you

Four examples for a large organisation, each logging every action, with a point where a person approves.

  • Assistants that answer from your documents

    Answers from your own policies and files, with the source passage shown.

  • Document review with traceable answers

    Contracts, loan files or claim files checked, every finding pointing to its passage.

  • Decision support with a full record

    Credit, underwriting or claims decisions prepared with reasons, approved by a person, logged.

  • Agents that work under approval

    Work of several steps carried out, with a person approving each step that matters.

These are examples of what we can build. They do not describe client work.

For a large organisation the choice is usually a private deployment or an air-gapped system, made per type of work. The three kinds of deployment are explained on the solutions page.

How we work

The free briefing costs nothing. Each paid step has its own fixed price, in writing before it starts.

  1. 1. Free briefing
  2. 2. Discovery
  3. 3. Design
  4. 4. Build
  5. 5. Run and improve

Discovery serves the whole table, and it ends with a presentation to your board by a partner.

How each step works, on the solutions page

What each person at your table gets in writing

A solution goes live only when each of them agrees, so discovery ends with something on paper for each.

  • The board

    One short file to approve: the systems in use, where each may run, one owner.

  • Procurement and supplier management

    A supplier file per supplier, and our own answers to your supplier check, in writing.

  • The technology department

    The deployment decision per type of work, checked against the real configuration.

  • The data protection officer and compliance

    Where the data goes, where a person reviews, and one description of each system.

Read the article: Obstacles for large organisations, and why the work stalls between departments

Obstacles for large organisations, and why the work stalls between departments

A large organisation has the people and the budget, and no single list of what is running. The work moves again when each department gets its own page and one executive owns the whole.

A large organisation has the people and the budget for this work, and still it stalls. Each department holds one piece and no one holds the whole. Procurement has the contracts. The data office has a list of models. Security knows the settings, and the data protection officer knows which decisions touch people. Pedro Machado of the Supervisory Board of the European Central Bank described the ownership of artificial intelligence at banks in February 2026 as fragmented across information technology, data science, business lines and control functions. A law firm with many practice groups shows the same pattern in smaller rooms, each group with tools it chose and will defend. At a Dutch employer the works council must consent, under Article 27 of the Wet op de ondernemingsraden, the Dutch works councils act, to arrangements on staff appraisal and on facilities suited to observing behaviour or performance, and that consent can take longer than the purchase.

The usual finding is two registers that do not join. Since 17 January 2025 the Digital Operational Resilience Act has required every financial entity to keep a register of information on all its contracts for information and communication technology services. The data office keeps a second list, the inventory of artificial intelligence applications. Different departments built them for different rules. Laid side by side at a large bank, the register shows a cloud tenancy, a document processing vendor, a customer service platform and a coding assistant as four unrelated suppliers, and the inventory shows that all four reach the same small group of model makers. That is concentration risk, which Article 29 of the Act tells the bank to assess before it contracts for a critical or important function. De Nederlandsche Bank, the Dutch central bank and prudential supervisor, reports that many banks have not completed implementation in full and that the quality of the register has to rise. The same holds outside finance. The Algemene Rekenkamer, the Dutch court of audit, found in 2024 that only a small share of the artificial intelligence systems at central government stood in the public register. Expect the first joined list to be wrong.

Each party at the table needs something different on paper. The board needs a short file: the joined list, the decision per type of work on which route the data may take (an online interface under contract, a private deployment or the organisation's own premises), and the name of the one executive who owns both. Under Article 5 of the Digital Operational Resilience Act the management body bears the ultimate responsibility for technology risk, so directors have to be able to read what they approve. Procurement needs a supplier file for each supplier: where the data is processed, which subcontractors sit behind the product down to the model maker, what is retained, and which rights of audit the contract gives. Technology needs the route decision in writing and the task of checking it against the real configuration. Microsoft's documentation of 18 May 2026 states that a sample of prompts in a private cloud deployment may by default be stored for abuse monitoring and reviewed by the provider's staff unless an exemption was applied for, and that a deployment labelled Global may process in any geography. The data protection officer needs, for each system, the point at which a person reviews the outcome. Article 22 of the General Data Protection Regulation gives people the right not to be subject to a decision based solely on automated processing, and on 21 August 2026 the Autoriteit Persoonsgegevens, the Dutch data protection authority, fined Uber 824,990,000 euros because software deactivated drivers' accounts without human review. Uber has announced an objection, so that fine is not final.

The exposure of a large organisation is a percentage of worldwide turnover. Article 99 of the Artificial Intelligence Act sets a ceiling of 15,000,000 euros or 3 percent of total worldwide annual turnover, whichever is higher, for breaches of a list of duties, and of 35,000,000 euros or 7 percent for a prohibited practice. Those amounts are maximums. The obligations for high risk systems behind the first ceiling apply from 2 December 2027 for the uses in Annex III and from 2 August 2028 for regulated products. From those dates Article 25 also turns an organisation into the provider of a high risk system when it puts its own name on one, modifies it substantially or changes its intended purpose. A model on the organisation's own premises changes where the data goes. It does not make the organisation compliant, and the same governance sits over all three routes.

By next month the accountable board member should have four things in hand. One joined list in which every application is matched to its contract and to the model maker at the end of the chain. A route decision per type of work, with the reasons written out. A supplier file for the suppliers behind the most critical functions. And the name of the executive who owns the list. The AI Guys offers a free briefing as a first step: you answer five questions and receive one signed page within two working days that says what each of the board, procurement, technology and the data protection officer should receive.

The sources of this article

  1. Digital Operational Resilience Act, Official Journal text: Article 5 (the management body), Article 28 (register of information) and Article 30 (rights of access, inspection and audit)
  2. Digital Operational Resilience Act, Article 64: applies from 17 January 2025
  3. Digital Operational Resilience Act, Article 29: concentration risk assessed before contracting
  4. De Nederlandsche Bank, Toezicht in beeld, edition for 2025 and 2026: implementation at banks and the quality of the register
  5. Paperjam, on the speech of Pedro Machado of 24 February 2026 (a secondary source)
  6. Banken.nl, 22 January 2026, on the agenda of the Autoriteit Financiële Markten for 2026
  7. Algemene Rekenkamer, report of 2024 on artificial intelligence at central government
  8. Wet op de ondernemingsraden, Article 27: consent of the works council
  9. Artificial Intelligence Act, Article 99, paragraphs 3 and 4: fine ceilings (consolidated version of 27 July 2026)
  10. Regulation (EU) 2026/1744, Official Journal: obligations for high risk systems apply from 2 December 2027 (Annex III) and 2 August 2028 (regulated products); Article 111, paragraph 2, on systems already in service and the date of 2 August 2030
  11. Artificial Intelligence Act, Article 25: when a deployer becomes a provider
  12. Artificial Intelligence Act, Annex III, point 8: alternative dispute resolution
  13. Microsoft, documentation dated 18 May 2026 on data, privacy and abuse monitoring in private cloud deployments
  14. General Data Protection Regulation, Article 22: decisions based solely on automated processing
  15. Autoriteit Persoonsgegevens, 21 August 2026: fine of 824,990,000 euros on Uber, objection announced
  16. Autoriteit Financiële Markten, letter of 11 June 2026 on the implementation of the Artificial Intelligence Act
  17. Inspectie Gezondheidszorg en Jeugd, supervision of digital care
  18. Jubel, 3 March 2026: comparison of the Belgian and Dutch professional rules for lawyers
  19. European Banking Authority, factsheet of 21 November 2025 on the Artificial Intelligence Act and the banking sector
  20. TNO, report of 2024 on the dependence of care institutions on large foreign technology companies

Last checked 5 October 2026.

What you risk

For a large organisation the ceiling is the higher of the fixed amount and the percentage of worldwide annual turnover.

For breaking the obligations the Act lists, such as the transparency duties

€15 million or 3%

Source and caveat

Breaking the other duties the Act lists, such as the transparency duties: up to €15 million or 3% of worldwide annual turnover, whichever is higher.

European Union Artificial Intelligence Act (Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744), Article 99(4). Applies from 2 August 2026.

This is the most the law allows, not what is usually imposed. A fine can only follow a duty that applies: the transparency duties from 2 August 2026, the high-risk duties from 2 December 2027 and 2 August 2028.

Source: EU AI Act, Article 99(4), consolidated version of 27 July 2026, EUR-Lex, accessed 5 October 2026.

For breaking the basic principles of the GDPR or the rights of data subjects

€20 million or 4%

Source and caveat

Breaking the basic principles, the rights of data subjects or the rules on transfers outside the EU: up to €20 million or 4% of worldwide annual turnover, whichever is higher.

General Data Protection Regulation (Regulation (EU) 2016/679), Article 83(5). Applies from 25 May 2018.

This is the most the law allows, not what is usually imposed.

Source: Regulation (EU) 2016/679, Article 83(5), Official Journal text on EUR-Lex, accessed 5 October 2026.

The ceiling is the most the law allows, not what is usually imposed. All fines, with their sources.

Worked examples

These are worked examples. They do not describe real clients.

All worked examples

If your firm has no compliance department, read the page for small firms.

This page was last checked on .