The verdict in three sentences
A credible AI customer support chatbot for an SME in London costs, in 2026, 8,000 to 30,000 EUR to set up, plus a variable inference cost of 0.002 to 0.02 EUR per query. The real lever is not the model but the knowledge base (RAG), the guardrails and clean human escalation. Scoped properly it automates 40 to 60% of tickets and drops first-response time from hours to seconds.
What it really costs in 2026
The budget splits into a project cost (one-off) and an operating cost (monthly). Here are 2026 orders of magnitude for an SME handling 1,500 to 4,000 requests per month.
| Item | 2026 range (EUR) | Comment |
|---|---|---|
| Scoping + ticket audit | 1,500 - 4,000 | Analysis of the top 20 intents |
| Knowledge base (RAG) | 2,500 - 8,000 | Ingesting FAQ, docs, support history |
| Site + helpdesk integration | 3,000 - 9,000 | Widget, Zendesk/Freshdesk, CRM |
| Guardrails + human escalation | 1,500 - 5,000 | Anti-hallucination, agent handover |
| QA + go-live | 1,000 - 4,000 | Testing, GDPR, logging |
| Total setup | 8,000 - 30,000 | By scope and integrations |
On the operating side, monthly cost depends on volume and model.
| Monthly element | 2026 estimate | Basis |
|---|---|---|
| Inference (LLM API) | 30 - 300 EUR | 0.002-0.02 EUR/query x volume |
| Hosting + vectors | 20 - 120 EUR | Managed vector database |
| Knowledge base upkeep | 200 - 600 EUR | Monthly content refresh |
| Quality supervision | 150 - 500 EUR | Reviewing failed conversations |
| Monthly total | 400 - 1,500 EUR | Excluding seasonal peaks |
Build vs SaaS: which model to choose
SaaS solutions (Intercom Fin, Zendesk AI, etc.) often bill per resolution (0.80 to 1.80 EUR per resolved ticket in 2026), easy to launch but pricey at volume. A custom build costs more upfront but amortizes fast above 1,500 tickets/month.
| Criterion | Per-resolution SaaS | Custom build |
|---|---|---|
| Entry cost | Low (< 2,000 EUR) | 8,000 - 30,000 EUR |
| Cost at 2,500 tickets/month | 2,000 - 4,500 EUR/mo | 400 - 1,500 EUR/mo |
| Data control | Limited | Full (GDPR, EU hosting) |
| Customization | Medium | High (tone, workflows) |
| ROI tipping point | < 1,000 tickets/mo | > 1,500 tickets/mo |
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Mini case study
Sophie runs customer service at a London e-commerce SME: 3,000 tickets/month, 3 agents at 38,000 EUR loaded/year. She invests 18,000 EUR in setup and 900 EUR/month to operate. The chatbot automates 52% of requests, i.e. 1,560 tickets absorbed. That frees the equivalent of one agent (redeployed to retention), worth 38,000 EUR/year, against an annual cost of about 28,800 EUR in year one (setup + 12 months). Payback arrives in month 10, then a net gain above 26,000 EUR/year thereafter.
FAQ
How long to launch a reliable AI chatbot? Plan 4 to 8 weeks in 2026: 1-2 weeks of scoping and knowledge ingestion, 2-4 weeks of integration and guardrails, 1-2 weeks of QA. A pilot on 20 intents can ship in 3 weeks.
Will the chatbot hallucinate and give wrong answers? The risk exists but is controllable: with strict RAG (the model answers only from your documents), a confidence threshold and human escalation, you get below 2% off-topic answers in production after tuning.
Is it GDPR compliant? Yes if data hosting stays in the EU, conversations are logged with a clear purpose, and no sensitive data goes to a non-compliant third-party model. A DPA with the inference provider is essential.
What automation rate is realistic? Between 40 and 60% of level-1 tickets in 2026 for standard support. Beyond that you need better data and deep integrations (order tracking, automated refunds).
Can we start small? Yes: a 6,000-9,000 EUR pilot on the 15 most frequent intents lets you measure the real resolution rate before expanding scope.
Let's scope your project. Send us your monthly ticket volume, your tools (CRM, helpdesk) and target budget, and we'll scope a costed pilot. 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.
