The verdict in three sentences
A domain-specific chatbot connected to your knowledge base (RAG) costs in London between GBP 15,000 and GBP 45,000 in 2026, delivered in 6 to 12 weeks, with GBP 90-1,300/month in token consumption depending on volume. It does not replace your team: it absorbs 30-50 % of tier-1 tickets (passwords, order status, recurring questions) and drives +20 % satisfaction through instant 24/7 answers. The trap is not the AI model, it is the quality of your knowledge base: without clean documentation, no budget will save the project.
What a domain-specific chatbot actually covers
A serious AI assistant is not a widget wired to ChatGPT. It runs on a RAG (Retrieval-Augmented Generation) architecture: your documents (FAQ, terms, product sheets, procedures) are indexed in a vector store, and the model answers only from your verified sources, with citations. This cuts hallucinations and keeps editorial control.
| Line item | Scope | 2026 ballpark (GBP) |
|---|---|---|
| Scoping & use cases | Workshops, decision tree, brand tone | 2,200 - 5,500 |
| RAG ingestion | Doc indexing, chunking, embeddings | 3,500 - 11,000 |
| Widget integration | Website, WhatsApp, agent handoff | 2,800 - 8,000 |
| Business connectors | CRM, ticketing, order base | 3,500 - 13,000 |
| Guardrails & testing | Filters, human escalation, UAT | 1,400 - 7,000 |
| Project total | Complete domain chatbot | 15,000 - 45,000 |
The 6-to-12-week timeline mostly depends on content availability and API access to your internal tools.
Recurring costs: the real subject
Build cost is one-off; what weighs over 3 years are the recurring items. Token consumption varies by model and conversation volume.
| Monthly volume | Model type | Tokens (GBP/month) | Hosting & maintenance |
|---|---|---|---|
| 1,000 conversations | Budget model | 90 - 260 | 130 - 350 |
| 5,000 conversations | Mid-tier model | 350 - 800 | 260 - 520 |
| 15,000 conversations | Advanced + RAG | 800 - 1,300 | 430 - 870 |
| 30,000+ conversations | Optimised multi-model | 1,300 - 2,600 | 700 - 1,300 |
A 2026 best practice: route simple questions to a small model and reserve the large model for complex cases. This often halves the token bill.
Mini case study
Sophie, head of customer relations at a fashion e-commerce brand in London, handles 8,000 tickets/month with 6 agents. 45 % concern order status and returns. She deploys a domain chatbot at GBP 28,000, with GBP 620/month in tokens and maintenance.
The bot absorbs 40 % of tickets, i.e. 3,200/month. At a loaded cost of GBP 4 per human-handled ticket, gross savings are GBP 12,800/month. After recurring costs (GBP 620), net gain is roughly GBP 12,180/month. The project pays back in under 3 months, and Sophie redeploys 2 agents to high-value cases, lifting premium satisfaction.
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
FAQ
Can an AI chatbot hallucinate and give wrong info?
With a well-built RAG architecture, the model answers only from your verified sources and cites references. Risk drops below 2-3 %, and an escalation guardrail hands off to a human when in doubt.
How long before go-live?
Expect 6 to 12 weeks. The limiting factor is rarely the technology: it is the quality and availability of your documentation, plus internal API access.
What happens if the bot cannot answer?
It escalates to a human agent with full conversation context. Tuned well, this smart handoff spares the customer from repeating themselves and cuts handling time by 20-30 %.
Is GDPR an obstacle?
No, provided you host data in the UK/EU, anonymise logs and avoid sending sensitive data to the model. A DPA with the AI provider and a clear retention policy suffice in most cases.
Can we start small then expand?
Yes, it is recommended. Launch on 3-5 high-volume use cases (GBP 15,000-20,000), measure deflection, then add business connectors once ROI is proven.
Let's scope your project. Tell us your ticket volume, channels (web, WhatsApp) and current tools: we will price a RAG domain chatbot with an indicative budget and timeline. 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.
