The verdict in three sentences
For a mid-size company in Amsterdam with 900 employees and 4,000 procedures and technical notes, an AI assistant built on RAG (retrieval-augmented generation) costs EUR 25,000 to 60,000 and goes live in 2 to 4 months. The non-negotiable condition is respecting access rights: the assistant must never quote a document a user could not open themselves. With 30 minutes saved per day per active user, 200 users are enough to pay the project back within months.
Why a RAG assistant rather than a generic chatbot
A consumer chatbot does not know your work instructions, ISO 9001 quality sheets or engineering notes. RAG indexes your documents, retrieves relevant passages and drafts an answer citing each source with a link. Employees verify in one click. The risk difference is major.
| Criterion | Generic AI chatbot | Internal RAG assistant |
|---|---|---|
| Knowledge of internal procedures | None | 4,000 documents indexed |
| Source citations | No | Link to document and page |
| Access rights enforced | Not applicable | Filtering by SharePoint or Entra ID group |
| Hallucination risk | High | Reduced, answers limited to sources |
| Data hosting | Often outside the EU | EU only |
| Content updates | Impossible | Daily re-indexing |
2026 budget and what it covers
| Component | 2026 range | Detail |
|---|---|---|
| Source connectors (SharePoint, DMS, wiki, PDF) | EUR 5,000 to 12,000 | Extraction, chunking, metadata |
| Access rights management | EUR 4,000 to 10,000 | Entra ID or LDAP sync |
| Vector and hybrid search engine | EUR 5,000 to 12,000 | Hosted in the EU |
| Interface (Teams, intranet, web) | EUR 4,000 to 10,000 | History, user feedback |
| Quality evaluation and guardrails | EUR 4,000 to 10,000 | Set of 200 reference questions |
| Security, logging, GDPR compliance | EUR 3,000 to 6,000 | DPIA, query traceability |
| Total project | EUR 25,000 to 60,000 | Usage EUR 500 to 2,000/month |
Monthly usage cost depends on question volume and model choice: about EUR 0.01 to 0.04 per question in 2026, i.e. EUR 500 to 2,000 a month for 300 to 600 active users.
Security: what your CISO will check
Three requirements come up every time. Rights filtering must apply before generation, not after, so no confidential excerpt ever reaches the model. Data and queries must stay in the EU, with no reuse for provider training. Every question and answer is logged for audit with a defined retention period. Documents marked restricted (HR, legal, sensitive R&D) can be excluded from the index by default.
Mini case study
Sanne, CIO of an industrial mid-size company in Amsterdam, launches a RAG assistant for EUR 48,000 with EUR 1,200 a month in usage. 250 employees use it daily and save 30 minutes on average. Over 210 working days, that is 250 × 0.5 h × 210 = 26,250 hours. At a loaded cost of EUR 55 an hour, the value of recovered time exceeds EUR 1.4 million a year. Even counting only 10% of it, the project pays back in under 5 months.
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
FAQ
How much does an internal RAG AI assistant cost in 2026?
Between EUR 25,000 and 60,000 for 2,000 to 10,000 documents, plus EUR 500 to 2,000 a month in usage depending on users.
How do we stop employees reaching confidential documents through the AI?
Entra ID or SharePoint permissions are synced and applied before retrieval. A test of 50 denied-access cases validates this before go-live.
Does data leave the European Union?
No, hosting, vector index and model are selected in EU regions, under a contract excluding training on your data.
What accuracy can we target?
As a 2026 order of magnitude, 85 to 92% of answers rated correct on a 200-question reference set, with systematic citations.
How long does a pilot take?
A pilot on 1,000 documents and 50 users launches in 5 to 6 weeks, before rollout across the company.
Let's scope your project. We map your sources, access rights and CISO requirements for an indicative budget delivered in 2 to 4 months. Detailed quote within 48 h. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
