The verdict in three sentences
An internal AI assistant — a secure question-and-answer engine wired to your procedures, contracts, wikis and documents — costs 15,000 to 40,000 EUR depending on document volume and security level. Monthly usage sits between 100 and 500 EUR and the lead time between 6 and 12 weeks. The main gain: about -30 % of information-search time, provided you get EU hosting and data governance (per-document access rights) right.
Budget breakdown
The price difference from a simple chatbot comes from security and governance: here each user must only see what they're entitled to, and indexing must cover varied formats (PDF, Office, emails, wiki).
| Line item | 2026 range (EUR) | Note |
|---|---|---|
| Scoping + document audit | 2,000 - 5,000 | Sources, formats, volume |
| Indexing + RAG pipeline | 4,000 - 12,000 | Chunking, embeddings, updates |
| Per-document access control | 3,000 - 9,000 | Rights aligned to internal directory |
| Interface + tool integration | 3,000 - 8,000 | Web, Teams/Slack, search |
| Security + EU hosting | 3,000 - 6,000 | Encryption, logging, isolation |
| LLM usage + hosting / month | 100 - 500 | Depends on active users |
Where the assistant saves time
The assistant pays off where information is scattered and frequently searched. Here are 2026 orders of magnitude by internal use.
| Internal use | Search time avoided | Population |
|---|---|---|
| Procedures & work instructions | 2-4 h/person/month | Operations, support |
| Contracts & clauses | 3-6 h/person/month | Legal, procurement |
| Technical documentation | 4-8 h/person/month | R&D, IT |
| HR & payroll policies | 1-3 h/person/month | All staff |
| Tender responses | 5-10 h/case | Sales, pre-sales |
Governance: the real topic
An assistant that answers everyone across all documents is a risk, not an asset. The rule: the assistant inherits existing access rights, never shows a forbidden document to a user, logs queries, and cites its sources so they can be checked. We host the index and logs in the EU and contractually frame non-reuse of data on the LLM provider side. This foundation is a significant share of the budget — it's what separates a serious project from a demo.
Mini case study
Mehdi is CIO of a 120-person consulting firm in Bordeaux. On average, each consultant loses 3.5 h/month searching for a procedure, a template or an answer already written. Average loaded cost: 45 EUR/h.
Monthly time lost: 120 × 3.5 h = 420 h, i.e. 18,900 EUR/month. The AI assistant (30,000 EUR project) cuts this by 30 %, i.e. 126 h recovered = 5,670 EUR/month, minus ~350 EUR of usage. Net gain: ~5,320 EUR/month. Payback in ~6 months, with more consistent client answers as a bonus.
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
FAQ
How is this different from a customer-service chatbot?
The scope is internal and sensitive: per-document access control, EU hosting, logging. Governance weighs more than the interface, hence a 15,000 to 40,000 EUR budget.
How many users to make it profitable?
From 40 to 60 active users who search often, ROI is usually reached in under 12 months. Below that, we first target the teams that search the most.
How are document updates handled?
We set up automatic re-indexing (daily or on each change) so the assistant stays current. It's included in the RAG pipeline.
Can the assistant cite its sources?
Yes, and it's essential: each answer links to the source document, which allows verification and reassures legal and compliance teams.
Can data be kept strictly in-house?
We host the index in the EU and can, depending on budget, use models with isolated processing. Full zero-external-data is possible but raises infrastructure cost.
Let's scope your project. Give us the document volume, your sources (SharePoint, wiki, emails) and your access constraints, and we'll scope security and budget. 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.
