# DataDream Full Reference

> AI agency based in Middelburg, Zeeland, Netherlands. We build working AI solutions that save businesses time and money.

Last updated: 2026-04-28

## Company Overview

DataDream is an AI agency founded by Laurens van Dijk, based in Middelburg, Zeeland, Netherlands. We help businesses implement AI solutions that deliver measurable results: time saved, costs reduced, revenue increased. We serve clients across the Netherlands and internationally.

Unlike traditional consultancies that deliver reports and leave, DataDream builds and deploys production AI systems. Our workflow is Claude/Anthropic-native, meaning we use the same advanced AI tools we recommend to our clients.

**KvK (Chamber of Commerce):** 76220540
**VAT number:** Available on request
**Founded:** 2024
**Founder:** Laurens van Dijk

## Why Businesses Use AI Now

Industry research consistently shows:
- 88% of organisations already use AI in their processes ([McKinsey, 2025](https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai))
- Average cost reduction of 35% through AI-driven automation (Deloitte)
- Frequent AI users save an average of 9 hours per week ([Federal Reserve, 2025](https://www.stlouisfed.org/on-the-economy/2025/feb/measuring-time-savings-generative-ai))
- 40% of applications will contain AI agents by 2026 (Gartner)

These numbers explain why Dutch SMEs (MKB) are actively looking for practical AI partners, not researchers or software vendors, but specialists who build and implement.

## Services in Detail

### AI Marketing & Content
Generate on-brand content at scale. Blog posts, social media, visuals, and video that maintain your brand voice while dramatically reducing production time. AI handles the heavy lifting, you approve the output.
- Automated content pipelines
- Brand-consistent AI copywriting
- AI-generated visuals and video
- Social media content automation

### AI Consulting & Strategy
Strategic AI advisory for businesses ready to implement. We start with an AI readiness assessment, identify the highest-impact opportunities, and create a practical implementation roadmap.
- AI readiness assessments
- Implementation roadmaps
- Change management and adoption support
- ROI-focused AI strategy

### AI Training & Workshops
Hands-on training that gets your team productive with AI tools. From prompt engineering fundamentals to advanced workflow automation. We train people to work with AI, not be replaced by it.
- Prompt engineering workshops
- AI tool training (ChatGPT, Claude, Copilot)
- Custom workshops for specific roles and industries
- Ongoing coaching and support

### AI Agents & Automation
Custom AI agents that handle real business tasks autonomously. From customer service calls to data processing, we build agents that work 24/7 with minimal oversight.
- Custom AI agent development
- Workflow automation (n8n, Make, custom)
- System integrations and API connections
- Voice AI agents for phone support

### AI Data & Insights
Turn your business data into actionable insights. Automated dashboards, analytics pipelines, and reporting that update themselves.
- Automated dashboards and reporting
- Data pipeline automation
- Business intelligence with AI
- Predictive analytics

### AI Customer Service
AI-powered customer support that handles inquiries 24/7. Chatbots, WhatsApp automation, and voice agents that speak your brand.
- AI chatbots (website, WhatsApp)
- Voice AI agents for phone support
- Automated ticket routing and resolution
- Multilingual support capabilities

## Free AI Readiness Scan

We offer a free AI Readiness Scan at https://datadream.nl/ai-scan. Answer a few questions about your business and receive a personalised AI readiness report with specific recommendations. No email required to start. The scan covers: current AI maturity, top automation opportunities, estimated time savings, and a prioritised action plan.

## Engagement model

We work in phases: first a short analysis to find the impact points, then a focused pilot to test assumptions, only then do we scale what works. Every engagement starts with a cost-benefit analysis upfront. No surprises, no hidden costs. Quotes are tailored to scope and discussed in the first meeting, not published as fixed tiers.

## Client Testimonials

**Jordi Dooge, Business Scout at Dockwize (maritime technology):**
"Laurens heeft ons geholpen in het tot leven brengen van een programma-aanbod dat eigenlijk nog gelanceerd moest worden. We hebben voor dit project erg fijn samengewerkt. Op naar meer mooie cases waar we AI in kunnen zetten om onze dienstverlening te verbeteren!"

**Theo Egginton, General Manager at Chillhop Music (international music brand):**
"De samenwerking met Laurens heeft een essentiële rol gespeeld bij de digitale transformatie van Chillhop Music, gevormd door zijn grondige kennis van de nieuwste technologie en AI integraties."

**Seth Colchester, CEO & Founder at Mycogenius (biotech):**
"Wat opvalt is dat er echt de tijd genomen wordt om elk probleem grondig te begrijpen voordat er met oplossingen gekomen wordt. Ik ben niet alleen tevreden, maar oprecht blij met de bijdrage aan onze projecten."

**Sam de Koning & Boaz Pleijte, Founders at Ortho Formulas (e-commerce):**
"We hebben waardevolle inzichten gekregen op het gebied van ecommerce, SEO, e-mail en andere digitale marketing. Wij hopen deze samenwerking nog lang voort te kunnen zetten."

**David Chapman, CTO at CBD Oil Europe:**
"De combinatie van diepgaande technische expertise en een sterke marketingmentaliteit zorgt ervoor dat complexe problemen niet alleen opgelost maar ook voorkomen worden."

## Method: Three Steps

**Step 1: Analysis**
We start by listening. What are your challenges? Where does your time go? We map your processes and find where AI has the most impact. Free and without obligation. Duration: 1-2 weeks.

**Step 2: Strategy**
You receive a concrete plan. Not 80 pages, but a clear document: what we will do, what it costs, and what it delivers. No surprises. Includes ROI projection and implementation timeline.

**Step 3: Implementation**
We build it, test it, and launch it. With training for your team so they can use it independently. We stay involved until it genuinely works. Most clients see first results within 2-4 weeks of go-live.

## EU AI Act (Q&A)

The EU AI Act is the world's first comprehensive regulation of artificial intelligence. It entered into force on 1 August 2024 and applies in stages. DataDream's implementations are designed to be AI Act-compliant by default, no separate compliance circus required.

**Q: What is the EU AI Act and when did it take effect?**
A: The AI Act is Regulation (EU) 2024/1689, published in the Official Journal on 12 July 2024 and in force since 1 August 2024. It is directly applicable across all EU member states, including the Netherlands. The Act takes a risk-based approach: the higher the risk an AI system poses to fundamental rights, safety, or health, the stricter the obligations. Application is staggered through 2027. Source: [artificialintelligenceact.eu](https://artificialintelligenceact.eu/) and [Commission Q&A](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai).

**Q: Who must comply and from when?**
A: The key dates: 2 February 2025: prohibitions on unacceptable-risk AI plus the AI literacy obligation under article 4 became enforceable. 2 August 2025: rules for general-purpose AI (GPAI) models, governance bodies, and penalties applied. 2 August 2026: the bulk of obligations for high-risk AI systems and limited-risk transparency rules. 2 August 2027: full application including high-risk systems embedded in regulated products. Providers, deployers (organisations using AI in their work), importers, and distributors all carry obligations, including SMBs. Source: [official timeline](https://artificialintelligenceact.eu/implementation-timeline/).

**Q: What are the risk categories?**
A: Four tiers. Unacceptable risk (banned): social scoring by governments, manipulative AI exploiting vulnerabilities, untargeted scraping of facial images, real-time remote biometric identification in public spaces (with narrow exceptions), emotion recognition at work or school. High risk: AI in critical infrastructure, education and vocational training, employment and HR, essential services (credit scoring), law enforcement, border control, justice administration, biometrics. Limited risk: chatbots, deepfakes, AI-generated content, emotion recognition systems (transparency obligation only). Minimal risk: spam filters, AI in video games, most everyday business uses (no specific obligations beyond AI literacy).

**Q: What is required for SMBs and MKB organisations?**
A: For most SMB applications (content generation, customer service bots, internal automation, marketing) you fall under limited or minimal risk. The concrete obligations: ensure AI literacy among staff who work with AI (article 4, in force since 2 February 2025), label AI-generated images, video, audio, and deepfakes, disclose to users when they interact with a chatbot, document basic information about which AI you use and how. High-risk obligations only kick in for specific use cases like CV screening, credit scoring, or biometrics. The Act explicitly avoids loading SMBs with disproportionate compliance burden, but it does not exempt them.

**Q: How does the AI Act interact with NIS2 and GDPR?**
A: They are layered. GDPR governs personal data: lawful basis, data subject rights, data minimisation, retention. The AI Act governs the AI system itself: risk classification, transparency, human oversight, documentation. NIS2 governs cybersecurity: incident reporting, network and system security for essential and important entities. An HR recruitment AI typically triggers all three: GDPR (it processes candidate personal data), AI Act (HR is high-risk), NIS2 (if your organisation is in scope as essential or important entity). We map per use case which framework applies and design controls that satisfy all of them at once, not three separate paper trails. Sister project [nis2-compliant.com](https://nis2-compliant.com) handles the NIS2 side.

**Q: What is the AI literacy obligation under article 4 in practice?**
A: Article 4 has been enforceable since 2 February 2025. It says organisations using AI must ensure "a sufficient level of AI literacy" for staff who operate AI systems and for those affected by them. The Act does not specify hours or curriculum. The Dutch supervisor (Autoriteit Persoonsgegevens, AP) and the European AI Office look for: do your people understand what the tool does, what its risks are, how to verify output, and where the limits are. In practice this means targeted training plus internal guidelines on paper. For most SMB organisations a half-day workshop plus a one-page policy is enough to demonstrate compliance. Source: [AP guidance on AI literacy](https://autoriteitpersoonsgegevens.nl/themas/algoritmes-ai/ai-literacy).

**Q: What are the penalties for non-compliance?**
A: Significant. Up to EUR 35 million or 7% of worldwide annual turnover for prohibited AI practices, EUR 15 million or 3% for high-risk and other obligations, EUR 7.5 million or 1% for supplying incorrect information to authorities. SMB caps are lower (the lower of the two amounts applies, not the higher). Penalties apply from 2 August 2025 for GPAI rules and from 2 August 2026 for the bulk of obligations.

**Q: How does DataDream help with AI Act compliance?**
A: Three concrete things. One: per use case we run a short risk classification so you know which tier applies and which obligations follow. Two: we build technical compliance into the implementation by default (audit logging, model version tracking, retrieval source citations, escalation paths, transparency disclosures), not as a separate add-on. Three: we provide AI literacy training that meets the article 4 standard plus the documentation that proves it, see https://datadream.nl/ai-training. No standalone compliance project, no certification theatre. Compliance is part of building it right.

## Sector expertise (Q&A)

### AI for legal (lawyers and notaries)

**Q: Is AI reliable enough for legal work?**
A: AI assists, it does not replace legal judgment. The lawyer always retains final responsibility. We build in safeguards: AI output goes through human-in-the-loop, sources are linked so a lawyer can verify the original case law or contract clause. We explicitly advise against AI use for cases where error tolerance is low without heavy review.

**Q: What about confidentiality?**
A: Confidentiality in legal work is a legal obligation, not a nice-to-have. We use GDPR-compliant tools where documents are not stored or used for training. For maximum confidentiality we build on-premise solutions where data does not leave the firm network. This meets bar association (NOvA, KNB) requirements and the most stringent DPA clauses from international clients.

**Q: How do you handle hallucinations and factual errors in AI output?**
A: Three-pronged. One: retrieval systems where the AI only cites from validated source documents (case law, own contract library, legislation), not from generic "training data". Two: every AI claim is linked to the source so the lawyer can verify. Three: we advise on use cases where AI is suitable. For pleadings or client-facing legal advice: not without heavy review. For contract analysis where sources are firm-controlled: fine.

**Q: Does this work for small firms too?**
A: Small firms benefit the most. Where a large firm has juniors and interns for the groundwork, AI gives a solo practice that same capacity. A solo lawyer can independently handle due diligences that previously needed a team. Start small, scale only what works.

**Q: Is this AI Act-compliant?**
A: Yes, if properly set up. Most legal AI applications are not high-risk under the AI Act. Transparency requirements apply (clients and counterparties knowing about AI use in certain contexts) and documentation is required. We document per use case which AI is deployed, how it was trained, what data it processes, and how human supervision is arranged.

Sector page: https://datadream.nl/ai-juridisch

### AI for HR

**Q: Doesn't AI screening of CVs lead to discrimination?**
A: That risk exists with poorly designed unchecked systems. We build differently. Transparent criteria set with the recruiter, bias checks (dropout reports per group, blind screening on name and address), AI assists and suggests but does not replace human judgment. The recruiter remains responsible for every rejection. The AI Act explicitly classifies HR recruitment as high-risk, so this oversight is mandatory.

**Q: Is HR recruitment really high-risk under the AI Act?**
A: Yes. Annex III of the AI Act lists employment, workers management, and access to self-employment as high-risk areas. From 2 August 2026, AI systems used to recruit, screen CVs, evaluate candidates, or take decisions affecting promotion or termination must meet high-risk obligations: risk management system, data governance, technical documentation, logging, transparency, human oversight, accuracy and cybersecurity. We make sure implementation meets these from day one.

**Q: What about employee data privacy?**
A: GDPR-compliant, EU-only data processing. Employee data is never used for AI training. Standard data processing agreement with the department. For sensitive data (absences, reviews, salary) we build on-premise solutions where data stays inside the company network. This also meets requirements from works councils (ondernemingsraden) and privacy officers.

**Q: Is this allowed under works council rules?**
A: Yes, when properly arranged. We help with the works council process: information sessions, demo, FAQ, drafting the consent request (instemmingsverzoek). We document what the AI does, which data it uses, and how to object. The works council has a right of consent for AI in personnel decisions under article 27 WOR.

**Q: Does this work for small HR teams?**
A: Especially then. A one-person HR function drowning in 80 vacancies and 200 leave requests per quarter benefits the most. Solutions scale from a single HR professional to departments of 50+. Many SMB clients start with one HR colleague and automate precisely because of that.

Sector page: https://datadream.nl/ai-hr

### AI for accountancy firms

**Q: Does AI work with our accounting software?**
A: Yes. We work with Twinfield, Exact Online, AFAS, Visma, and Yuki. Most packages have an open API or supported connection, so invoice recognition, automatic booking suggestions, and reporting export plug in without staff leaving their environment. For packages without open API we use scan-and-recognise or email-to-booking as an intermediate step. We only build once the connection is stable and bookings are correct, otherwise it produces net extra work.

**Q: Who is responsible if AI makes a mistake towards the tax authority?**
A: You remain professionally and legally responsible, that does not change. AI is a tool, not an accountant. We always build in a review step: AI drafts the suggestion, a staff member checks and authorises. For VAT returns, annual accounts, and tax positions, a human is in the loop before anything goes out. AI does the first 80%, the senior reviews the last 20%.

**Q: Can AI help with Wwft and KYC checks?**
A: Yes, this is one of the immediate-impact areas. AI can search UBO checks, sanctions lists, and adverse media, draft a first risk assessment, and record outcomes in a file. The compliance officer reviews the assessment, but the gathering and summarising is automated. The final risk assessment stays human work, especially in elevated risk cases. We tighten the audit trail so you can demonstrate to BFT or the tax authority during supervision which steps were taken.

**Q: What about confidentiality of client data?**
A: We work with EU-only AI providers where data is not used for training and not retained after processing. For firms with strict DPA requirements we build setups where data does not leave the firm network. We document per use case which data is processed, by which model, and for how long, useful for GDPR accountability and for client questions.

**Q: How does AI keep up with current legislation?**
A: A general language model does not know what changed yesterday in tax law or in an NBA guideline. We solve this by connecting AI to validated sources: an internal knowledge base with current NBA publications, tax handbooks, case law via [rechtspraak.nl](https://uitspraken.rechtspraak.nl/), and internal firm positions. The AI cites only from those sources with direct references. For fast-moving topics (box 3, business succession, CSRD) we always recommend a human final check.

Sector page: https://datadream.nl/ai-accountants

### AI for real estate agents

**Q: Can I use AI-generated text on Funda?**
A: Yes, provided you are transparent and the text is accurate. NVM and VBO do not ban AI-generated text, but the agent remains liable for the content. An AI description that mentions non-existent features or makes the property look better than it is, falls under misleading advertising. Our approach: AI drafts from your CRM object data (size, year, features), the agent reviews and adjusts. For AI-edited photos (virtual staging, tidied gardens) it gets stricter: NVM and consumer organisations push for clear labelling, and the AI Act in some cases requires disclosure that content is manipulated.

**Q: Does AI work with Realworks, Skarabee or MUS?**
A: Yes. Most Dutch real estate systems offer integration via NVM exchange, MAPI or their own API. We build AI tools that pull object data from your CRM, generate a Funda description in your office style, and write the draft back to the CRM for final check. Same for lead data: a chat intake on your site asks qualifying questions, scores them, and routes to the right agent.

**Q: What about GDPR and buyer lead data?**
A: Buyer leads contain name, phone, search criteria, sometimes income or financing indication. Privacy-sensitive data falling under GDPR. We build flows where data stays within the EU, is not used for training general AI models, and is only accessible to your office. For Wwft checks we work with validated Dutch providers that perform source checks within the correct legal framework. No US data parties for identity data.

**Q: Will AI replace my role at viewings and bidding?**
A: No. The value of an agent is personal contact, reading a buyer, negotiating, knowing the neighbourhood. AI takes over surrounding work: Funda text, confirmation emails, viewing reports, follow-up after open houses, bidding status updates to sellers. That gives you more time for what clients actually come to you for.

**Q: What if the AI writes something wrong about a property?**
A: We design around it. AI only writes from CRM object data, no invented features. Every draft goes through human review before Funda. For figures with legal impact (NEN 2580 area, year built, energy label) we use measurement-report values, not model estimates. We log which model version produced which text, in case questions arise later. Liability stays with the office, so the control loop must hold.

Sector page: https://datadream.nl/ai-makelaars

### AI for e-commerce

**Q: Does this work with my current e-commerce platform?**
A: Yes. We integrate with Shopify (Plus and standard), WooCommerce, Magento (Adobe Commerce), Lightspeed, BigCommerce, Centra, and custom platforms via API. Most integrations operational within a week. For unknown platforms we build API connections or work via Make.com / n8n as a bridge. No platform lock-in.

**Q: How many products do I need at minimum?**
A: AI recommendations based on purchase behaviour work best with 50+ products and 100+ orders per month. Smaller catalogs see immediate benefits from automated product descriptions, AI chatbot for standard questions, email campaign automation, and SEO content. A webshop with 20 products can already see ROI from content automation alone.

**Q: Does the AI chatbot work in foreign languages?**
A: Yes. Native NL, EN, DE, FR, ES, IT and extendable. Detection happens automatically via browser language and first client message. For sector-specific terminology (fashion, jewellery, technical products) we train the bot on your product catalog and prior customer conversations.

**Q: What about returns and fraud detection?**
A: Two AI applications with strong impact. Return prediction: AI identifies orders with high return risk based on product, customer, behaviour, and prior data, so you can intervene proactively. Fraud detection: AI flags suspicious patterns (unusual order size, atypical payment combinations, geographical anomalies) before processing. Clients typically see 30 to 50% lower fraud damage.

**Q: Does this work for B2B webshops or only B2C?**
A: Both. B2B brings complex catalogs, discount tiers, account-specific pricing, long sales cycles. The AI approach works particularly well because there is lots of data on recurring customers. AI recommendations based on purchasing patterns, automated quotes, smart sales follow-up, chatbot for account managers with full context.

Sector page: https://datadream.nl/ai-e-commerce

### AI for marketing agencies

**Q: Will AI not make our work generic and recognisable?**
A: Only if you use AI the way everyone else does. Generic prompt in ChatGPT, generic output. What makes the difference: your tone of voice, your brand guidelines, approved examples, and specific client context as input. We build on-brand AI environments with your own brand books, sample texts, and style guides baked in. The output carries your signature, not the language model's.

**Q: How do client briefings and strategy stay confidential?**
A: EU-only AI providers where input is not stored or used for model training. For agencies with clients in sensitive sectors (financial, pharma, defence) we build environments where data stays within the agency's own infrastructure. We document per workflow who can do what with which data, so you can demonstrate to clients and under GDPR that it is handled properly. No shadow IT.

**Q: Which work can AI actually take off our hands?**
A: First drafts of blogs, whitepapers, and SEO content. Social variations per platform and audience. Mailings and newsletter concepts. Monthly reports where campaign data and commentary come together. Image and video variations for a/b tests. Translations for international clients. Pitch decks where you do not start from scratch. Junior research for strategy teams. AI does not do the strategic thinking, the critical eye on concepts, or the client relationship.

**Q: How do you train an AI on our tone of voice?**
A: We work with existing material, not a questionnaire. We gather approved texts representative of your brand or specific client: campaigns, blogs, voice guidelines, mailings that worked. From that we distil style patterns, sentence rhythm, vocabulary, and explicit off-limits. Captured in prompt templates and a retrieval system that consults your own examples before the AI writes anything.

**Q: What do we have to communicate to clients about AI use?**
A: The AI Act sets transparency requirements. For most marketing output this is limited risk: end users must be able to know AI was involved. For deepfakes, generated images of people, and AI-generated audio, more specific rules apply. We help record per workflow where AI was used, in which stage, and how human review was arranged. Many agencies include AI use in client contracts as standard.

Sector page: https://datadream.nl/ai-marketingbureaus

### AI for education

**Q: Is AI safe to use in education?**
A: Yes, if set up properly. We exclusively use GDPR-compliant tools and ensure student data is never processed outside the EU. Standard data processing agreements. For schools under the Education Authority Act, we build on-premise solutions where data stays local.

**Q: Does AI replace the teacher?**
A: Absolutely not. AI takes over repetitive tasks (grading, reports, standard communication) so teachers have more time for guidance, creativity, and personal connection with students. The work that makes a teacher a teacher is exactly what we do not automate. AI does the boring part, you do the human work.

**Q: What about AI detection and student plagiarism?**
A: We advise schools on AI policy: when AI is allowed, when not, how to assess it. We also build tools for teachers that lighten the work: feedback on draft versions, alternative assignments where AI helps less, class discussion material on AI ethics. We help with curriculum innovation around AI literacy.

**Q: Which education sectors do you work with?**
A: Primary education, secondary education, vocational, higher professional education, and university. Different needs per sector: in primary more parent communication and differentiation, in secondary and vocational more test analysis and study guidance, in higher education more research assistance and essay feedback.

**Q: Does this work for smaller schools or only large institutions?**
A: For smaller schools AI is often a game changer. Every hour of the teacher counts double. Solutions scale from a one-person school automating parent communication to an education foundation with 30 locations rolling out central test analysis. Starting is possible with just one process.

Sector page: https://datadream.nl/ai-onderwijs

### AI for tourism and hospitality

**Q: Does AI work for a small B&B or holiday rental?**
A: That is exactly where it works well. A large hotel has front desk, revenue manager, and marketing team. As a vacation rental owner you do all of that yourself, often alongside another job. AI takes away multilingual listing texts, answers to standard booking questions, review responses during peak season, and social posts. Start small, with the part that takes the most time today. No big upfront investment, no annual PMS contracts that do not fit your scale.

**Q: May I use AI for guest contact under the AI Act?**
A: Yes, with a transparency obligation. Article 50 of the AI Act states that if guests chat with an AI system, they must know it. A short line at the bottom of the chat ("you are chatting with our AI assistant, for personal contact please email...") is enough. For review responses and email marketing this obligation does not apply the same way, since that is outbound communication you publish as the operator. AI-generated photos and videos have separate labelling requirements.

**Q: What about multilingual support for German, Belgian and British guests?**
A: This is the first win for most Zeeland and coastal businesses. AI does not just translate word for word, it adapts tone and context per language. A German guest expects different information and different phrasing than a Flemish or British one. Listing texts on Booking and Airbnb, automatic confirmation emails, house rules, welcome pages: all consistent across four languages. Update once, AI translates along.

**Q: Can AI make my nightly or weekly pricing dynamic?**
A: Yes. Yield management used to be reserved for large chains with expensive software. We build lighter solutions that look at your occupancy, season, local events (Liberation Day weekend, Concert at Sea, Veerse Marathon, German school holidays in North Rhine-Westphalia) and competitor pricing via public data. The system suggests prices, you approve or let it publish automatically. We integrate with Mews, Cloudbeds, Lodgify, Smoobu where possible. No black box.

**Q: Does AI replace my front desk or personal contact?**
A: No, and it should not. Tourism is hospitality, that cannot be automated. What AI does: catch the stack of standard questions that come in every day. Check-in time, dogs allowed, power outlet by the static caravan, early check-in. By removing those, your team keeps time for real contact: the guest just arriving, the couple wanting a restaurant tip, the problem that needs human attention.

Sector page: https://datadream.nl/ai-toerisme

## Service detail (Q&A)

### AI Marketing & Content

**Q: How do you keep content on-brand?**
A: Tone-of-voice is a trainable component, not a PDF in a drawer. We train models on your existing brand corpus: blogs, brochures, emails, customer communication. From that we distil style rules (sentence length, avoid-words, perspective, formality level) and feed those into reusable prompts. For imagery we do the same through brand-style training in Recraft, so AI illustrations feel consistent. Final quality check on every delivery.

**Q: Who actually writes: AI or human?**
A: Both, in a fixed division of roles. AI handles first drafts, outlines, variants, translations, repetitive volume work. Humans handle strategy, final editing, fact-checking, and pieces where context or nuance is decisive. For customer cases, opinion articles, sensitive topics: less AI, more handwork. For product descriptions, FAQ pages, social variants: AI carries the heavy lifting. We mark per content type whether it is "AI-first with review" or "human-first with AI assistance".

**Q: Does AI-generated content need to be labelled under the AI Act?**
A: Yes, for certain content types a transparency obligation applies from August 2026. The AI Act requires AI-generated or substantially modified imagery, video, audio, and deepfakes to be recognisably marked, both technically and visibly. For AI text published as journalism or informative material the same applies, with exceptions for human-edited text. We help decide per content type: metadata tags, a short disclosure line, or a separate mention in colophon or footer.

**Q: Who owns content created with AI?**
A: You do. We work with tools whose licence terms allow commercial use and where output rights sit with the client. For imagery we pay extra attention: Midjourney, Recraft, and Nano Banana have different licence structures. We document per project which tool was used and which licence applies. For text, AI-generated material does not get standalone copyright protection in most jurisdictions, so we advise on what that means for republishing.

Service page: https://datadream.nl/ai-content

### AI Strategy & Consulting

**Q: How do I start with AI when I have nothing in place?**
A: You start with a conversation, not a tool. What does your business do, where is the bottleneck, where does time get lost on repetitive work, where is the pain biggest. That is the AI Quickscan. From it come three to five concrete use cases. Pick one to start with, small and scoped, to see if it fits your way of working. Only when that pilot delivers value do you look at the next one. We do not talk about "AI transformation", we look at one work process at a time.

**Q: We have no data team. Can you compensate for that?**
A: Yes, that is exactly the SMB question. Large companies have a data department. At an SMB the inventory is often missing. We do the data readiness audit: what data do you have, where is it, what quality, and may you legally use it for AI. Often you have more than you thought, just spread across systems. Sometimes you need no own data at all for the first use case, off-the-shelf AI with good prompts gets you far.

**Q: How do I know if an AI use case really delivers value?**
A: Three questions. One: does this process today demonstrably cost time or money, and is that quantifiable in hours-per-week or errors-per-month. Two: is the AI solution proven elsewhere, or are we the first to try it (being first is expensive). Three: are the people who have to work with it willing to adapt. If any answer is no, we do not do it, or not yet. Better to honestly say a use case saves 4 hours a week than to invent a marketing-pretty number.

**Q: Do you build it yourselves or coach our team?**
A: Both, depending on what fits. Sometimes we build because the use case calls for it and it is faster. Sometimes we coach your own team or an external builder you already have, because that is cheaper and more sustainable. Sometimes the advice is: just buy this off-the-shelf tool. We have no interest in always building ourselves. The choice between build, buy, and coach we make per use case based on what delivers most.

Service page: https://datadream.nl/ai-strategie

### AI Training & Workshops

**Q: Is AI training really mandatory for our employees?**
A: Since 2 February 2025, article 4 of the AI Act requires organisations using AI systems to ensure "sufficient AI literacy" among employees who work with them. The law does not specify exactly what that means, but it is not a tickbox. The regulator looks at whether your people understand what the tool does, what the risks are, how they verify output, and where they draw the line. For most SMB organisations: targeted training plus internal guidelines on paper.

**Q: How much time will this cost my team?**
A: Depends on starting point. AI basics workshop for a team that has barely used AI: one half-day (3 to 4 hours) including coffee break. Prompting masterclass for people with experience: 4 hours. Team-specific training: typically 1 day or 2 half-days, with a week in between to practice. Ongoing coaching: monthly 1-hour check-in per team. We deliberately keep it compact, an 8-hour classroom day does not work. People learn by doing.

**Q: Does this work for seniors without a tech background?**
A: Especially for that group. Often the most valuable participants are people with years of experience in their field who simply have not touched AI tools yet. Their domain knowledge makes them spot what AI can and cannot do faster than beginners. We start with a live demo on a recognisable case from their work, not theory or jargon. After that they work themselves, with guidance.

**Q: How do we measure whether it works?**
A: Self-assessment before and after on concrete points: can I write a good prompt, do I know which tools we may use for which work, do I recognise when AI output is wrong. Feedback right after training and again after 4 weeks (where you see what stuck). The real signal is in behaviour: do people actually use AI in their work. For organisations that want to report formally on AI literacy (article 4) we deliver documentation of training content and participant list.

Service page: https://datadream.nl/ai-training

### AI Agents & Automation

**Q: How do I know if an AI agent is reliable enough for production?**
A: Reliability comes from design, not model name. Agents have clear boundaries: only the tools and data they need, defined task, explicit instructions on what to do when uncertain. We test on realistic scenarios including edge cases and adversarial use. We measure success rate, hallucination rate, escalation rate. Weekly dashboard showing what the agent did, what went well, what did not. Only when numbers are stable do we scale volume.

**Q: How does escalation to a human actually work?**
A: Human-in-the-loop is the rule, not the exception. For every agent we define when a human must step in: low confidence, sensitive decisions (finance, legal, complaints), unknown input patterns, or the customer asking. Escalation goes through your existing channel: Slack, Teams, ticketing, email. The team member receives full context, the agent's proposal, and can approve, adjust, or take over with one click. Stricter thresholds for new agents, looser as they prove themselves.

**Q: What happens with our sensitive data?**
A: For truly sensitive data we build on-premise or in an EU-only cloud you control. Nothing leaves your infrastructure. For less sensitive use cases we work with providers (Anthropic, OpenAI, Google) that contractually guarantee inputs are not stored or used for training. Vector databases like PGVector or Weaviate can run locally, as can open models like Llama or Mistral.

**Q: How does AI Act compliance work for agents?**
A: The AI Act imposes requirements on logging, transparency, and human supervision, especially for agents that affect people. We build audit trails by default: every decision recorded with input, output, model version, timestamp, any human approval. For agents in high-risk categories (recruitment, credit scoring, medical advice) logging is stricter: prompt version and retrieval sources retained too. Transparency to end users: they know they are talking to an agent and how to escalate. Compliance file per agent.

Service page: https://datadream.nl/ai-agents

### AI Data & Insights

**Q: Our data is a mess. Can you clean it up first?**
A: Yes, and that is often where we start. A dashboard built on dirty data produces nice charts that mean nothing, or numbers that contradict each other. We start with a data-quality audit: which sources, which fields are reliable, where are duplicates, empty records, broken formats, outdated definitions. Then a lightweight modelling layer (usually with [dbt](https://www.getdbt.com/)) where the logic lives once. Only on top of that clean layer do we build dashboards and AI analysis. Sometimes the groundwork is bigger than the dashboard, and we will say so.

**Q: Which tools do you use and is there vendor lock-in?**
A: BigQuery, Postgres, or DuckDB as warehouse, dbt for modelling, Fivetran or Airbyte for connectors (or custom scripts when cheaper), Looker Studio, Metabase, or Hex for dashboards. AI analysis via Claude or GPT API. Deliberately open stack: SQL models run on any warehouse, dbt is open source, dashboards can be moved. We deliver documentation and code in your repository, not in a platform that locks you in.

**Q: How does GDPR work for customer behaviour analysis?**
A: Customer segmentation falls under GDPR the moment you process personal data, even just an email address or customer ID. Legal basis must be clear: usually legitimate interest or contract performance for analysis on your own customer data. For heavier profiling with automated decision-making, stricter rules apply ([Article 22 GDPR](https://gdpr-info.eu/art-22-gdpr/)). Practical measures by default: pseudonymisation where possible, retention periods, no customer data sent to external AI vendors without DPA, processing register update.

**Q: How do we know the insights are correct?**
A: Three checks. One: every number has a definition and a traceable query. You can drill down to the underlying rows. Two: dbt has tests (uniqueness, not-null, referential integrity, business rules) running on every refresh. Alerts before you see the deviation in the dashboard. Three: for AI analysis (forecasts, segmentations) we run a back-test on historical data and show how output matches what actually happened. No blind trust in a model. If the back-test is poor, we say so.

Service page: https://datadream.nl/ai-data

### AI Customer Service

**Q: Customers hate chatbots. How do you do it differently?**
A: Customers hate bad chatbots: bots that loop, do not understand, refuse to hand off. We build differently. The bot lives off a real knowledge base (your FAQ, product docs, policy) via RAG, so it does not invent answers. As soon as the bot is unsure, it escalates with full conversation context. We test on hundreds of real questions before go-live. We keep training on actual conversations, not just on the FAQ. A good AI customer service is invisible when it works, honest about its limits when it does not know.

**Q: When does it escalate to a human?**
A: Not every conversation belongs with AI. Standard escalation rules: low confidence, customer angry or disappointed (sentiment), complex or non-standard question, customer asks for a human, or compliance / financial / legal matters. The bot hands off with full context: what was asked, which articles offered, where the customer got stuck. Your colleague does not start over. Rule of thumb: AI handles 70-80% of routine, humans pick up what counts.

**Q: AI Act article 50: does the customer need to know they are talking to AI?**
A: Yes. Article 50 requires customers to be informed when interacting with an AI system instead of a human. Applies to all chatbots, voice bots, and automated email replies where the average user would not realise it is AI. Standard in every implementation: brief disclosure at conversation start (for example "Hi, I am the AI assistant for [company], happy to help and I will hand you over to a colleague if you prefer"), visible branding (chatbot icon, naming). For voice AI: spoken disclosure in the first seconds.

**Q: Does this work with our existing helpdesk (Zendesk, Freshdesk, Intercom)?**
A: Yes, that is usually the starting point. We integrate with Zendesk, Freshdesk, Intercom, HubSpot Service Hub, Salesforce Service Cloud and most other helpdesks. AI sits on top of your existing system, does not replace it. AI-handled tickets are logged in your helpdesk including full conversation. Escalated tickets land in the right queue with relevant tags and context. Reporting, SLA tracking, team work distribution keep working as before.

Service page: https://datadream.nl/ai-klantenservice

## Frequently Asked Questions

Q: What is DataDream exactly?
A: DataDream is an AI agency from Middelburg. We help businesses use AI as a concrete tool, not a buzzword. From smart agents that answer your phone to content that writes itself. We build working solutions, not thick reports.

Q: How does DataDream work on engagements?
A: We work in phases. First a short analysis to find the impact points. Then a focused pilot to test assumptions. Only after that do we scale what works and stop where the business case doesn't close. Every engagement starts with a cost-benefit analysis upfront so you know what to expect. Quotes are tailored to scope and discussed in the first meeting.

Q: How do you determine project timeline?
A: Timelines depend entirely on scope. We discuss realistic phasing during the first meeting after the analysis. We work in evaluable phases so you can decide at each step whether we continue.

Q: Do we need technical knowledge?
A: No. We handle all technical aspects. Our training is designed so everyone can use it, regardless of background. If you can send email, you can work with AI.

Q: How does DataDream handle privacy and GDPR?
A: Privacy is our priority. We work exclusively with GDPR-compliant tools and advise on data handling with every implementation. Your data stays yours. We work only with European data storage when required.

Q: What businesses is DataDream suitable for?
A: Anyone who spends time on predictable work. From a sole trader who wants to do more without extra staff, to an SME that wants to automate processes. We work a lot in Zeeland but also throughout the Netherlands and internationally, from Rotterdam to internationally.

Q: How do you choose an AI agency in the Netherlands?
A: Look for three things: do they build working solutions or only advice? Do they have demonstrable results with similar businesses? Are they transparent about prices and approach? A good AI agency delivers results within weeks, not months.

Q: What is an AI readiness scan?
A: An AI readiness scan maps how far your business is with AI and where the biggest opportunities lie. At DataDream you can do a free online scan on our website. You immediately receive a personal report with concrete recommendations.

Q: What is the difference between AI consulting and AI implementation?
A: AI consulting is advice: which AI solutions fit your business and how do you approach it? AI implementation is the actual build and delivery. At DataDream we do both, from strategy to working product. No report that disappears into a drawer.

Q: Does DataDream also work outside Zeeland?
A: Yes. Our office is in Middelburg, but we work with clients throughout the Netherlands and internationally. AI projects can be executed perfectly remotely. We are happy to come by for an introduction, but daily work is digital.

## Technology Stack

DataDream builds with:
- Claude API (Anthropic) for AI agents and analysis
- Gemini (Google) for web grounding and research
- n8n and Make for workflow automation
- OpenAI GPT for specialised tasks
- Custom Python and Node.js backends
- Vercel for deployment
- Next.js for web applications

## Contact & Location

- Website: https://datadream.nl
- Email: info@datadream.nl
- Phone: +31 85 124 95 22
- Address: Middelburg, Zeeland, Netherlands
- LinkedIn: https://linkedin.com/company/datadream-ai
- Chamber of Commerce (KVK): 76220540
- Service area: Netherlands and international

## Sister project / partners

- nis2-compliant.com: DataDream's sister project covering NIS2 directive compliance for Dutch SMBs, in partnership with Burak Yazici. AI + NIS2 readiness audits available as a combined offering.

## Service pages (deep-dive per service)

- AI Content & Marketing: https://datadream.nl/ai-content
- AI Strategy & Quickscan: https://datadream.nl/ai-strategie
- AI Training & Workshops: https://datadream.nl/ai-training
- AI Agents & Automation: https://datadream.nl/ai-agents
- AI Data & Insights: https://datadream.nl/ai-data
- AI Customer Service: https://datadream.nl/ai-klantenservice

## Sector pages (industry-specific)

- AI for Education: https://datadream.nl/ai-onderwijs
- AI for HR: https://datadream.nl/ai-hr
- AI for Legal (lawyers, notaries): https://datadream.nl/ai-juridisch
- AI for E-commerce: https://datadream.nl/ai-e-commerce
- AI for Marketing Agencies: https://datadream.nl/ai-marketingbureaus
- AI for Accountancy Firms: https://datadream.nl/ai-accountants
- AI for Real Estate Agents: https://datadream.nl/ai-makelaars
- AI for Tourism & Hospitality: https://datadream.nl/ai-toerisme

## Content & Resources

- Blog: https://datadream.nl/blog
- AI Readiness Scan: https://datadream.nl/ai-scan
- Sitemap: https://datadream.nl/sitemap.xml
- Summary (llms.txt): https://datadream.nl/llms.txt
- Dutch reference (llms-nl.txt): https://datadream.nl/llms-nl.txt