What is an AI advisor?

An AI advisor is an independent specialist who guides an organisation's decisions about artificial intelligence without owning the build: which processes are worth automating and which are not, whether to buy a tool or commission custom software, what the EU AI Act and GDPR require of a given use case, and whether a supplier's proposal, pilot or invoice holds up. The advisor sells judgement rather than development hours, is engaged part-time or on a retainer rather than per project, and reports to the owner or the board rather than to the IT department.

The term overlaps with four others that buyers use interchangeably: the AI consultant, the AI strategy consultant, the fractional Chief AI Officer and the AI development agency. The difference is not seniority but ownership: the advisor owns the decision, the consultant a scoped analysis and often the first build, the fractional CAIO the AI function itself for a few days a week, and the agency delivery.

What does an AI advisor do?

Five responsibilities recur in almost every advisory engagement.

• Framing. Turning "we should do something with AI" into two or three measurable problems, each tied to a process and a system you already run: invoice capture in Exact or AFAS, order intake in the ERP, dock planning in a WMS. The output is a short list with a baseline number per problem, not a vision deck.
• Build-or-buy and vendor judgement. Whether an off-the-shelf tool covers the case, whether custom development is justified, what a fair quote looks like, and which clauses on data, model ownership and exit matter.
• Regulation. Placing each use case in its EU AI Act risk tier, checking what GDPR requires of the data, and flagging what the works council must be told.
• Governance and literacy. An AI policy staff can follow, an inventory of the AI systems in use, the Article 4 literacy measures on record, and a named person who signs off new tools.
• A second opinion. Reading pilots and results from other suppliers and saying plainly what they mean, including when the answer is not AI: a setting in software you already own solves a surprising share of these requests.

Equally important is what an advisor does not do. The role writes no code, runs no project and carries no delivery risk. That is why it can be independent, and why someone else has to turn the advice into working software.

How does an AI advisor differ from a consultant, a fractional CAIO, a strategy consultant and an agency?

The five roles separate on what each one owns and how it bills.

• AI advisor. Owns the decision. Part-time, retainer or fixed-scope audit; bills for judgement by the day or the month. Fits when the question is still open.
• AI consultant. Owns a scoped analysis and, in most Dutch practices, the first build. Per project; hourly or fixed price. Fits when you can name the process and want analysis and build from one team.
• AI strategy consultant. Owns the roadmap across departments: portfolio, operating model, budget. A defined study, usually by a strategy firm.
• Fractional Chief AI Officer. Owns the AI function itself, including governance, vendor decisions and board reporting, one to three days a week on a monthly retainer. Fits several initiatives running at once with no internal owner.
• AI development agency. Owns delivery. Per specification; bills for capacity. Fits a validated use case that needs working software.

The CAIO title has spread fast at the top of the market. The IBM Institute for Business Value CEO study of May 2026 (2,000 CEOs in 33 geographies, surveyed with Oxford Economics) found that 76% of surveyed organisations have a Chief AI Officer in 2026, up from 26% in 2025; IBM publishes no separate European figure. A 2024 Gartner poll of more than 1,800 executives found that 54% had a head of AI, and 88% of those did not carry the CAIO title. Both samples are large enterprises. For a Dutch company with 20 to 250 staff, the realistic form of the same role is fractional or advisory: a few days a month, not a seat on the board.

What does an AI advisor cost, and how is one engaged?

Four engagement models are common: a retainer (fixed days per month), a day rate (single days around a tender or supplier review), a fractional arrangement (a retainer with executive authority) and a fixed-scope audit (a defined deliverable priced up front).

Published Dutch rates exist for the individual, not the arrangement. The Knab Zzp Uurtarievenboekje 2026, based on more than 20,000 freelancers, puts the average hourly rate of a freelance AI specialist at €120 excl. VAT, against €83 across all freelancers; Consultant.nl reports €80 to €200 per hour. For firms, Consultancy.nl uses annual revenue per consultant as its proxy: roughly €300,000 to €400,000 at the Big Four and above €400,000 at strategy firms. Those firms publish no day rates, so any "€2,500 a day" figure is a market estimate, not a price list. For the fractional CAIO we found no independent European survey; US and UK vendor pages quote about $5,000 to $30,000 a month, which is a supplier's own price sheet rather than a benchmark.

Crux Digits does not sell advisory hours. The advisory role is the AI Audit & Strategy at a fixed €2,500 (excl. VAT): in one to two weeks you receive a ranked list of use cases with cost, saving and payback per opportunity, an EU AI Act and GDPR check per idea, and a plan you can execute with or without us. If you continue to the €20,000 Production-ready MVP within 60 days, the audit fee is credited in full; production follows from €50,000, and you own the code at every step. See the pricing page and AI consultancy pricing in the Netherlands.

Which EU AI Act and GDPR duties does an AI advisor handle for an SME?

The EU AI Act, Regulation (EU) 2024/1689, applies in stages. Since 2 February 2025 the prohibited practices of Article 5 and the AI literacy duty of Article 4 bind every provider and deployer, an SME included. Since 2 August 2026 the bulk of the regulation applies, including the Article 50 transparency duties: a chatbot must disclose that it is a machine and AI-generated content must be marked, with a grace period to 2 December 2026 for machine-readable marking on systems already on the market.

The Digital Omnibus on AI, Regulation (EU) 2026/1744, was published on 24 July 2026 and entered into force on 27 July 2026. It moved the stand-alone high-risk obligations of Annex III (recruitment, credit scoring, biometrics, education) to 2 December 2027, and those for AI embedded in regulated products under Annex I to 2 August 2028. It also reworded Article 4: providers and deployers must "take measures to support the development of AI literacy" of their staff, taking account of their knowledge, experience and the context of use. No individual competence level is guaranteed and no course or certificate is prescribed. In practice a short, recorded training for the people who use the tools is the evidence an inspector will ask for.

The risk tiers did not change: unacceptable, high-risk, limited and minimal. Most SME uses, such as reading invoices, forecasting demand or planning docks, are minimal risk. The duty is to know which tier each system sits in and to write it down; the first advisory deliverable is usually that inventory. The EU AI Act risk checker gives a first indication; the AI Act SME checklist lists the documents to have ready.

GDPR runs alongside the Act. The Dutch Data Protection Authority has made AI and algorithms its supervisory priority for 2026 to 2028 and in July 2026 published GDPR guidance for generative AI. A data protection impact assessment is mandatory where processing is likely to carry a high risk for individuals, which covers most AI applied to customers or staff. The advisor settles the legal basis, purpose limitation and the rule for staff pasting company data into public chat tools before the pilot starts, and briefs the works council, which under the Works Councils Act has a right of consent on rules for processing personal data and on systems able to monitor staff.

Why do most AI pilots stall, and what does an advisor do about it?

MIT's NANDA initiative reported in The GenAI Divide: State of AI in Business 2025, drawing on structured interviews, a leadership survey and a review of about 300 public deployments, that about 5% of AI pilots achieve rapid revenue acceleration while the vast majority stall with little or no measurable P&L impact. The same report found that purchased or partnered solutions reached deployment roughly 67% of the time against about one third for internal builds, and that budgets skewed to sales and marketing while back-office automation showed the better return.

Gartner points the same way. In February 2025 it predicted that through 2026, 60% of AI projects will be abandoned because they are not supported by AI-ready data; in June 2025 that over 40% of agentic AI projects will be cancelled by the end of 2027 through escalating costs, unclear business value or inadequate risk controls. Deloitte's State of AI in the Enterprise 2026 (3,235 leaders in 24 countries) found that only 25% had moved more than 40% of their AI pilots into production.

The causes are consistent: pilots chosen for visibility rather than a measurable baseline, data that was clean in the demo and messy in production, no owner for the metric, and no agreed path from pilot to production. An advisor earns the fee by refusing each of these in advance: the use case gets a baseline, the data is checked before budget is committed, the metric gets an owner, and the pilot is designed as the first release of a production system rather than a demo. See how to run an AI pilot and the AI pilot entry.

How many Dutch companies use AI, and what does that mean for the advisor question?

Statistics Netherlands (CBS) reported in February 2025 that 22.7% of Dutch companies with 10 or more employees used AI in 2024, up from 14% in 2023: 17.8% of companies with 10 to 19 staff, 22.2% with 20 to 49, and 59.2% with 500 or more. In December 2025 CBS published the 2025 figures on the wider population of companies with 2 or more employees: 17% used AI in 2025, against 13% in 2024 and 8% in 2023, with 27% at 10 to 50 employees, 45% at 50 to 250 and 66% at 250 and above.

Dialogic's study for the Ministry of Economic Affairs, AI-gebruik in het mkb: ambitie of aarzeling? (September 2025), names the barriers SMEs report: a lack of knowledge and skills, limited digitisation, cost, unclear added value and limited data, and finds that what they ask for is practical knowledge, guidance and support with vision and strategy. The Exact MKB Barometer of August 2026 (1,700 SMEs, reported by AG Connect) found that 70% of SMEs now use AI in some form, up from 33% in 2025, but 56% use it to write texts and generate images and only 36% to gain insight into business performance. The gap between CBS's 17% and Exact's 70% is one of definition: CBS counts AI embedded in business processes, the barometer any use, a chat tool included.

That is the advisory market in one sentence: adoption is broad, structural use is thin, and the stated need is for decisions rather than tools. The breakdown by sector and province is in AI adoption in the Dutch mkb 2026.

When do you need an AI advisor, a consultant, an agency, or nothing at all?

Read the checklist from the top; the first block that describes you is usually the right answer.

• Nothing at all, yet. The task is done a few times a week by one person, so no saving can pay for a project. The process has no digital record yet. Or the request can be met by a setting in software you already pay for.
• An AI advisor. You cannot yet name the process to start with. You hold quotes from several suppliers and cannot compare them. A supplier is proposing something your own team cannot assess. The board or the works council wants an independent view. You need an AI policy, an inventory and the literacy measures on record, and nobody inside owns it.
• An AI consultant. You can name the process and the measurable problem, want the analysis and the first build from one team, and want a price before you commit.
• An AI development agency. The use case is validated, a specification exists, and what you lack is capacity to build and run it.
• A fractional Chief AI Officer. Several initiatives run across departments, the combined budget exceeds what any single project justifies, and no one inside carries the function. In practice this starts around a hundred staff.

An audit that ends in a ranked list is only useful if someone then builds the first item; a build is only safe if someone independent has checked the case, the data and the risk tier first. The question for an SME is whether those two people must work for different firms.

Where does Crux Digits fit?

Crux Digits runs advice and build as one team from Utrecht. The €2,500 AI Audit & Strategy is the advisory step: one named expert maps your processes and systems, ranks the use cases by cost, saving and payback, places each in its EU AI Act tier with the GDPR check attached, and hands you a plan you can execute with or without us. The same expert stays on the €20,000 Production-ready MVP if you continue within 60 days, with the audit fee credited in full; production follows from €50,000, all excl. VAT and with the code yours at every step.

That structure changes what the advice is worth. An advisor who will never build has no reason to keep the numbers modest; one who must deliver the first use case at a fixed price has every reason to rank only what will survive its own delivery. The evidence is public: 13 published case studies describe what was built, for which process, and what it measured. Start with a free 30-minute conversation to check whether an audit is the right first step, or whether the honest answer is "nothing at all, yet".