Nearshore vs offshore AI development — a decision framework for Nordic companies

Key takeaways

• A nearshore partner inside the EU/EEA that processes personal data for you needs an Art. 28 processor agreement but no Chapter V transfer tool for that EU-based access.
• An offshore partner in Vietnam that can access EU personal data – even remotely, on EU-hosted systems – also needs a transfer tool such as the Standard Contractual Clauses and a transfer impact assessment (EDPB).
• Vietnam (UTC+7) is six hours ahead of Copenhagen in winter and five in summer; a 14:00–17:00 Vietnam core window overlaps 08:00–11:00 or 09:00–12:00 Copenhagen time.
• Offshore is usually the wrong choice when data cannot be accessed from outside the EU, when procurement requires an EU-established supplier, or when the work needs engineers on site every week.
• Compare bids on total cost – the same scope, named team and acceptance criteria, plus travel, legal work and your own review time – rather than on day rates.

The three options and how they compare

In-house means hiring your own ML engineers, data engineers and an AI product owner in Copenhagen, Stockholm, Oslo or Helsinki. You keep direct control and build domain knowledge over time. The costs are recruitment time for senior AI roles and a fixed payroll commitment before you know whether the use case pays off.

Nearshore usually means a partner in Central or Eastern Europe, for example Poland, Czechia, Romania or the Baltic states. The time difference is generally zero or one hour, they are a short flight away, and many are inside the EU.

Offshore means a partner further away, for example in Vietnam. The potential draws are price and engineering capacity; the trade-offs are a real time-zone gap, a possible third-country transfer if the team can access personal data, and less spontaneous contact. A hybrid – product ownership in-house, build capacity from a partner – is often realistic with either route.

Two nuances the table cannot show. Talent varies by vendor more than by country: ask who will actually write the code and whether they have shipped retrieval, evaluation and monitoring for LLM systems – large language model applications – rather than demos. And not every nearshore location is in the EU: if a partner sits outside the EU/EEA without an adequacy decision, the same Chapter V analysis applies as for offshore.

Nearshore vs offshore AI development: a side-by-side example

Take one hypothetical use case: an internal assistant that answers employees' questions from your policy and procedure documents, some of which name employees. You receive two bids for the same scope, one from an EU-based nearshore partner and one from a partner in Vietnam:

How to read it:

• Normalise the bids. Same scope, same named team, same acceptance criteria. Then add the costs outside each bid: travel, legal work on transfers, and your people's review hours. A fixed-scope contract with clear acceptance criteria protects a budget better than a low rate on open-ended work; we price fixed-scope and do not bill hourly (pricing).
• Check whether the overlap fits how you decide. If decisions happen in morning stand-ups and written reviews, three shared hours are workable. If stakeholders need long afternoon workshops every week, the nearshore bid has a real advantage.
• Let the data decide early. If the documents cannot be accessed from outside the EU, or cannot be redacted for development, the offshore bid drops out regardless of price. Redaction is more work than it looks for an assistant like this: names in headers, signatures and file metadata all need handling, and the evaluation set must be rebuilt on the redacted text.
• Plan the handover before you sign. Whoever builds it, you should end up with code, documentation and an evaluation set your own team can run.

Working across the gap: hours, language and data

Hours. Copenhagen runs on CET (UTC+1) in winter and CEST (UTC+2) in summer; Vietnam is UTC+7 all year. Our core overlap is 14:00–17:00 Vietnam time – 08:00–11:00 in Copenhagen in winter and 09:00–12:00 in summer. That suits a morning stand-up, a mid-morning review and decisions before lunch, with afternoon questions answered by your next morning. It suits less well a team that decides things in long afternoon workshops; a CEE partner in your time zone has a genuine advantage there.

Language. Both routes normally work in English, and with offshore teams written specifications, recorded demos and clear acceptance criteria matter more. If your AI must handle Danish, Swedish or Norwegian text with fine domain nuance, plan for native speakers on your side to review evaluation sets and outputs – whichever partner you choose.

Data. An EU-based partner processing personal data on your behalf needs an Art. 28 processor agreement. Access by a processor in Vietnam, which has no EU adequacy decision, also calls for a Chapter V transfer mechanism, commonly the EU Standard Contractual Clauses supported by a transfer impact assessment. EU hosting, EU-hosted model endpoints and anonymised or synthetic development data reduce exposure but do not replace those safeguards. Our GDPR-compliant AI outsourcing checklist walks through the contracts, and our security page sets out our approach.

The EU AI Act applies whichever route you choose. As of October 2026, following the AI Omnibus, obligations for Annex III high-risk systems, which include recruitment uses, apply from 2 December 2027, so classify your use case at scoping. This article is not legal advice.

A decision framework, including when offshore is the wrong choice

Work through these five questions in order. The first one that gives a hard answer usually decides the model.

1. Can the data be accessed from outside the EU at all? If customer contracts, public-sector rules or your own risk appetite say no, even with SCCs, offshore is out.
2. Is this a capability you must own for years? If the AI is your product's core differentiator, build an in-house team; a partner can still help you start faster, with a planned handover.
3. How much real-time collaboration does the work need? Daily co-design, frequent on-site workshops or exploratory research favour in-house or nearshore.
4. How well can the scope be defined? A clear use case with measurable acceptance criteria, such as a RAG system over internal documents or AI agents for a back-office workflow, suits fixed-scope offshore delivery.
5. What are the budget and timeline? If you need a working result in weeks on a limited budget, compare fixed-scope bids from both routes.

Offshore is usually not the right choice when the data cannot leave EU-controlled access, even remotely; when the work is open-ended research with no internal product owner; when you need engineers on site every week; or when procurement requires an EU-established supplier or EU governing law. On that last point, our contracting entity is NKKTech Global Pte. Ltd. in Singapore, under a Singapore-law MSA, so confirm early whether that fits your procurement.

When you do compare offshore vendors, ask for long engagements with clients in other countries, because they test the handover and documentation habits the knowledge-retention row depends on. The two published examples below are Japanese. For a Japanese e-commerce retailer, a 22-week, USD 140,000 legacy modernisation reported 90% fewer manual operations, three times faster page loads and zero downtime during migration (published case study). PrismLab in Japan has worked with us for more than three years on AI agent systems and AR; its AI agent systems reduced processing time by about 30% (published case study).

If none of the stop conditions apply, test the question with evidence rather than debate. Our AI Pilot runs one use case on your real data in 3–4 weeks for a fixed fee (from USD 5,000 as of October 2026), with any GDPR safeguards defined in the scope, so you can judge the offshore route on delivered work. Our page on offshore development in Vietnam covers the working model.

Related reading: AI development in Denmark – offshore options, costs and GDPR · GDPR-compliant AI outsourcing checklist

Frequently asked questions

Is nearshore or offshore better for AI development for a Danish company? Neither is better across the board. Nearshore in the EU gives you full time-zone overlap and no third-country transfer for EU-based access, while offshore can suit well-scoped builds where the bids compare well on total cost. If the data cannot be accessed from outside the EU, or the work needs constant real-time collaboration, choose nearshore or in-house.

How much cheaper is offshore AI development in Vietnam than hiring in Denmark? It depends on scope, seniority and the costs outside the bid, so treat any single percentage with caution and compare actual quotes. For a concrete reference, as of October 2026 our fixed prices start at USD 5,000–15,000 for an AI Pilot and USD 20,000 for a full build (indicative, at our reference rate, about EUR 18,000 or DKK 134,000).

Can an offshore AI team in Vietnam work with GDPR-protected data? Yes, with safeguards that match the data flow: an Art. 28 processor agreement (a databehandleraftale in Danish) and, if the team can access personal data, the EU Standard Contractual Clauses and a transfer impact assessment showing effective protection. Keeping production data in an EU cloud region with controlled, logged access is a useful extra safeguard, not a legal requirement in itself.