Digital Africa11 min read

Internal AI Assistant on a Consultancy Knowledge Base in New York

Mohamed Bah·Fondateur, Kolonell
October 15, 2026
Share:
Internal AI Assistant on a Consultancy Knowledge Base in New York

Internal AI Assistant on a Consultancy Knowledge Base in New York

Digital Africa

The verdict in three sentences

A firm of 120 consultants that can no longer find its deliverables loses the equivalent of several full-time staff every year to searching and rework. An internal RAG-based AI assistant connected to SharePoint, Google Drive or the document management system costs USD 27,000 to 55,000 (EUR 25,000 to 50,000) and USD 330 to 1,300 a month in inference. The non-negotiable condition is respecting per-engagement access rights: a consultant must never see, even in summary, a deliverable from a client they are not cleared for.

The hidden cost of 15 years of lost deliverables

A strategy or operations consultancy produces 400 to 800 deliverables a year: proposals, diagnostics, benchmarks, financial models, steering committee notes. After 15 years that adds up to 200,000 to 500,000 files spread across servers, SharePoint and the mailboxes of departed partners. Juniors redo benchmarks that already exist, managers rebuild proposals from scratch.

Indicator (2026 order of magnitude)Without assistantWith RAG assistant
Weekly search time per consultant5 to 6 h1 to 2 h
Hours saved per consultant per week04 h
Hours saved for 120 consultants over 45 weeks021,600 h
Average loaded hourly cost of a consultantUSD 110 to 160 (EUR 75 to 110 in Paris)same
Time to produce a proposal3 to 5 days1.5 to 2.5 days
Share of proposals reusing existing content20%55 to 65%
Files indexed and queryablea few %, via filename search200,000 to 500,000

Even if only half of the saved hours are counted as billable or reinvested in business development, the stake is well above USD 1 million a year for a firm this size.

Budget for a custom RAG assistant

RAG (retrieval augmented generation) means finding the relevant passages in your documents, then having a language model write the answer with source citations. Most of the budget does not go to the model but to ingestion, permissions and evaluation.

Item2026 range (EUR, excl. tax)Watch point
Scoping, source and permissions audit3,000 to 6,000Map SharePoint, Drive, DMS, CRM
Connectors and ingestion (PDF, PPT, Excel)6,000 to 12,000PowerPoint decks need dedicated processing
Vector and hybrid search engine4,000 to 8,000Combine semantic and keyword search
Per-engagement access filtering4,000 to 9,000Synced with Microsoft Entra ID groups
Interface (Teams, web) and source citations3,000 to 7,000Every answer links to the original file
Evaluation set and leak tests3,000 to 5,000150 to 300 sample questions validated by partners
Training and change management2,000 to 3,000Champions per practice
Total25,000 to 50,000 (about USD 27,000 to 55,000)10 weeks
Inference and hostingEUR 300 to 1,200/monthDepends on query volume and model

A New York firm will usually host in a US region it controls (Azure East US, AWS us-east-1), while a Paris firm needs EU hosting for its corporate clients. Either way, check contractually that your data is not used to train the provider's model.

Copilot or custom RAG?

Microsoft 365 Copilot costs about USD 30 per user per month, or USD 43,200 a year for 120 seats. It respects Microsoft 365 permissions, but only reads what lives in the Microsoft ecosystem and stays generic. A custom RAG costs more in year one, then USD 4,500 to 16,000 a year, and can include the DMS, the CRM, engagement records and business rules (for example, excluding deliverables under enhanced confidentiality clauses). Many firms combine both: Copilot for office productivity, RAG for engagement memory.

Mini case study

Need a professional website?

Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.

Prefer a call back?

Leave your WhatsApp number and a Kolonell expert will get back to you within 1 business day. Free, no strings attached.

You are :

Helen, operations partner at a 120-consultant firm, launches a EUR 38,000 project (about USD 41,000) with EUR 800 a month of inference. On the prudent basis of valuing 2 of the 4 hours saved per consultant, at EUR 85 per loaded hour over 45 weeks, the gain reaches EUR 918,000 a year. Dividing that figure by four to account for gradual adoption still leaves EUR 229,500 in year one, for a total cost of EUR 47,600. Payback in under three months after go-live. At New York billing rates, the same project pays back even faster.

FAQ

Can a consultant see documents from an engagement they are not on?

No, if filtering happens before generation: forbidden passages are never sent to the model. Test this with 30 to 50 trap questions before go-live.

Does the assistant make up answers?

The risk drops sharply when every answer cites its sources and the assistant says "I don't know" below a relevance threshold. On a 200-question evaluation set, aim for over 85% accurate, sourced answers.

What about compliance?

An internal document search assistant is not a high-risk system under the EU AI Act, but it requires informing staff, a processing record and documented hosting. In the US, client confidentiality clauses and SOC 2 expectations drive the same controls.

How long to index 300,000 files?

Allow 3 to 7 days of initial processing, then hourly incremental sync. PowerPoint files often account for 60% of ingestion time.

Should we clean up the knowledge base first?

A light cleanup is enough: excluding duplicates and draft versions cuts volume by 30 to 40% and improves relevance.

Let's scope your project. Describe your document sources, headcount and confidentiality constraints: we price a RAG assistant at EUR 25,000 to 50,000, delivered in 10 weeks. Detailed quote within 48 h. WhatsApp +221 77 596 93 33.

Tags:#internal AI assistant#RAG#knowledge base#consultancy New York#generative AI#knowledge management
Share:

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.