AI Development in Denmark – Offshore Options, Costs and GDPR

Key takeaways

• Danish companies can resource AI development in-house, through a Danish agency, with an EU nearshore partner or offshore; a hybrid that keeps product ownership in-house and adds outside build capacity is often workable.
• As of October 2026, our AI Pilot costs USD 5,000–15,000 fixed-fee – indicatively EUR 4,500–13,500 or DKK 34,000–101,000 at our reference rate – and full builds start from about DKK 134,000.
• Vietnam has no EU adequacy decision, so if staff there can access EU personal data, the transfer needs a safeguard such as the EU Standard Contractual Clauses plus a transfer impact assessment, even when the data is hosted in the EU (EDPB).
• After the AI Omnibus (Regulation (EU) 2026/1744), obligations for stand-alone Annex III high-risk AI systems apply from 2 December 2027, wherever the system was built (European Commission).
• Three costs sit outside any quote: model usage and hosting, a product owner's weekly time, and legal review of the data-protection paperwork.

Four ways to resource AI development in Denmark

If you began by looking for an AI development company in Copenhagen, the real choice is usually between four models. None is right for everyone:

For Nordic AI outsourcing, a workable split keeps product ownership and domain knowledge in-house while a partner adds build capacity. Decisions stay with you; the engineering goes to the partner.

A worked example. Take a hypothetical Danish logistics company that wants an assistant to answer staff questions from internal procedures written in Danish and English. Before any build starts:

• The head of operations signs off the success criteria – for example, the share of test questions answered correctly with a source cited, and the maximum acceptable rate of confident wrong answers.
• Two experienced Danish-speaking staff own the Danish test set. They write questions the way colleagues actually ask them, agree the correct answers and grade the pilot's output. The partner builds the evaluation tooling and reports the scores, but does not mark its own work.
• The DPO or legal adviser signs off the data flow – what data the pilot uses, where it is hosted and who can access it.

A partner who welcomes that structure is easier to work with than one who asks you to judge a polished demo. Modern language models handle Danish, but quality varies by task, so the Danish test set matters more than any English benchmark.

What AI development costs, from pilot to dedicated team

Partners usually price in one of three ways: hourly or time-and-materials, a dedicated team per month, or a fixed scope for a fixed fee. Hourly billing moves the risk of overruns to you. We do not bill hourly; work is priced by fixed scope, agreed before it starts.

Our prices as of October 2026, with EUR and DKK figures indicative, at our reference rate (EUR ≈ USD × 0.9; DKK ≈ EUR × 7.46):

An LLM (large language model) is the kind of model behind tools such as ChatGPT or Claude; a RAG system (retrieval-augmented generation) makes it answer from your own documents rather than from its training alone.

What the later stages look like in practice, from two published case studies – neither client is Danish:

• A full build. For a US fintech company, we built LLM document intelligence – software that reads, extracts and checks information in financial documents – in 14 weeks for USD 85,000 (about DKK 571,000, indicative). The published case study reports 80% less processing time, 95% accuracy and USD 200,000 in annual savings.
• A dedicated team. A Japanese software company has worked with a dedicated team of 6–10 of our engineers since January 2023; the published case study records a 5.0 on-time delivery rating over a partnership of more than three years.

Most Danish teams should not start at either size. Our AI Pilot for Danish companies is the smaller first step: one use case on your real data, a fixed fee, 3–4 weeks, code you own, and a full build only if the pilot earns it. The full breakdown is on our pricing page.

Three costs are easy to forget:

• Running costs – language-model usage and cloud hosting, which scale with volume. Estimate them during the pilot, not after.
• Your own time – a product owner who answers questions and reviews demos each week.
• Compliance work – processor agreements, transfer assessments and, for some systems, a data protection impact assessment (DPIA). Budget legal review for these.

GDPR and the EU AI Act, in brief

GDPR. Where a partner processes personal data on your behalf, you are normally the controller and the partner a processor, which calls for a data processing agreement under Article 28 of the GDPR. Vietnam does not have an EU adequacy decision, so if staff there can access EU personal data, treat it as a transfer that needs a safeguard – typically the EU Standard Contractual Clauses, backed by a transfer impact assessment. Hosting in an EU region, such as AWS Stockholm (eu-north-1), does not remove the transfer if a team outside the EU can reach the data, as the EDPB's guidance on transfers explains. Pseudonymised data is still personal data; only properly anonymised data falls outside the GDPR. Datatilsynet, the Danish Data Protection Agency, publishes guidance in Danish and English. Our GDPR-compliant AI outsourcing checklist covers the contracts in detail.

EU AI Act. The AI Act (Regulation (EU) 2024/1689) is risk-based. Many first projects – internal document search, summarisation, email triage – are unlikely to be high-risk, although chatbots carry transparency duties such as telling people they are talking to an AI. Annex III lists high-risk uses, including AI that filters job applications or evaluates candidates and AI that assesses individuals' creditworthiness. Two points matter for outsourcing:

• The Act applies where the system is placed on the market or used, not where it is built. An offshore developer does not take a system outside its scope.
• You may be the "provider". Putting a system into service under your own name can make you the provider, with the heavier duties. Get legal advice on your role.

As of October 2026, after the AI Omnibus entered into force on 27 July 2026, stand-alone Annex III high-risk obligations apply from 2 December 2027, and those for high-risk AI embedded in Annex I products from 2 August 2028. That is time, not an exemption: if your use case could be high-risk, build documentation, logging and human oversight in from the pilot.

This is general information, not legal advice.

How to structure the first project

The first project with any new partner should be small enough to stop and real enough to learn from:

1. One use case with a measurable outcome – search over internal documents in Danish and English using retrieval-augmented generation, customer email triage, document extraction, or an AI agent for a repetitive back-office task.
2. Success defined first – a baseline, a metric and a threshold, signed off by the business owner.
3. Data access agreed before data moves – hosting region, access model (anonymised data or controlled access) and the applicable GDPR safeguards, documented.
4. Fixed scope and fee – three to four weeks, one deliverable, no open-ended hours.
5. Your outputs – code, prompts, configuration and evaluation sets all belong to you.
6. A decision gate – continue to a full build, adjust, or stop.

On hours: Vietnam is UTC+7 all year, six hours ahead of Copenhagen in winter and five in summer. The shared window is your morning – 08:00–11:00 in winter, 09:00–12:00 in summer – which suits stand-ups, demos and decisions; questions raised in your afternoon are picked up at the start of the next Vietnamese working day. Written specifications, a shared decision log and one named product owner on your side do more for a distributed team than extra meetings.

For due diligence: contracts are with NKKTech Global Pte. Ltd. in Singapore under a Singapore-law MSA, so have your counsel review it alongside the DPA. We are ISO 9001 and ISO 22301 certified and rated 5.0/5.0 on Clutch from 9 verified reviews.

Related reading: GDPR-compliant AI outsourcing checklist · Nearshore vs offshore AI development for Nordic companies · EU AI Act for recruitment AI after the AI Omnibus

Frequently asked questions

How much does it cost to outsource AI development from Denmark to Vietnam? It depends on scope and pricing model. As a reference point, as of October 2026 our AI Pilot is USD 5,000–15,000 fixed – indicatively DKK 34,000–101,000 at our reference rate – and full builds start from USD 20,000, about DKK 134,000. Add running costs and your own team's time.

Is it GDPR compliant to use an AI development company outside the EU? It can be. Map which personal data the partner can access, agree a data processing agreement, and put a transfer safeguard such as Standard Contractual Clauses in place with a transfer impact assessment. EU hosting helps but does not remove the transfer if staff outside the EU can access the data, so take legal advice for your case.

Does the EU AI Act apply if my AI system is built by an offshore developer? Yes. The Act follows where the system is placed on the market or used, not where the code is written. As of October 2026, stand-alone Annex III high-risk obligations apply from 2 December 2027, and if you put the system into service under your own name, you may carry provider obligations yourself.