The verdict in three sentences
A custom AI support chatbot connected to your document base (a RAG approach) costs between EUR 12,000 and EUR 45,000 in 2026, for a 4 to 9 week timeline and a 30 to 55 % deflection of recurring tickets. The real running cost sits in two lines: the LLM API (EUR 0.3 to 3 per 1,000 requests) and maintenance (EUR 300 to 900/month). Below ~8,000 tickets/month a packaged SaaS is enough; above that, bespoke becomes profitable and strategic.
Packaged SaaS or custom chatbot: the real trade-off
SaaS solutions (Intercom Fin, Zendesk AI) charge per resolution or per seat: easy to launch, but the cost climbs fast with volume and you stay tied to their engine. Bespoke requires an upfront investment, but you own the knowledge base, the integrations and the unit cost.
| Criterion | Packaged SaaS | Custom chatbot (RAG) |
|---|---|---|
| Upfront cost | EUR 0 - 3,000 | EUR 12,000 - 45,000 |
| Monthly cost | EUR 0.7 - 1.5 / resolution | API EUR 0.3 - 3 / 1,000 req. + EUR 300 - 900 maintenance |
| Time to launch | 1 - 3 weeks | 4 - 9 weeks |
| Typical deflection | 20 - 40 % | 30 - 55 % |
| Internal system integration | Limited (connectors) | Bespoke (API, CRM, ERP) |
| Data ownership | Hosted by the vendor | EU hosting of your choice |
| Tipping point | < 8,000 tickets/month | > 8,000 tickets/month |
What makes up a bespoke project budget
The price mostly depends on functional scope and level of integration. A chatbot that answers from a FAQ is simple; an agent that queries the CRM, checks an order status and opens a ticket is a project in its own right.
| Line item | 2026 range (EUR) | Indicative timeline |
|---|---|---|
| Scoping + conversation design | 2,000 - 5,000 | 1 week |
| RAG pipeline (ingestion, vectorisation) | 4,000 - 12,000 | 1 - 3 weeks |
| Integrations (CRM, ticketing, auth) | 3,000 - 15,000 | 1 - 3 weeks |
| Web interface + widget | 2,000 - 6,000 | 1 week |
| Guardrails, testing, go-live | 1,000 - 7,000 | 1 - 2 weeks |
| Total | 12,000 - 45,000 | 4 - 9 weeks |
The API cost varies by model: an economy model handles 1,000 requests for ~EUR 0.3, a premium model for ~EUR 3. Prompt caching and difficulty-based routing cut the bill by 30 to 60 %.
Mini case study
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Nadia, head of customer relations at an e-commerce retailer in London, handles 14,000 tickets/month, 60 % of them recurring (order tracking, returns, availability). She invests EUR 28,000 in a custom chatbot, with a running cost of EUR 550/month in API (optimised routing) and EUR 600/month in maintenance.
The bot deflects 45 % of tickets, i.e. 6,300 tickets absorbed. At a human handling cost of EUR 3.20/ticket, the saving is EUR 20,160/month. Even after the EUR 1,150/month of costs, the return on investment is reached in under 2 months, and the team refocuses on complex cases.
FAQ
How much does the LLM API really cost per month? For 14,000 requests/month with an optimised model and caching, expect EUR 150 to 550/month. Without optimisation or caching, the bill can triple.
What deflection rate is realistic to target? Between 30 and 55 % on a well-scoped perimeter. Beyond 55 %, quality drops because the bot starts answering cases it should escalate.
How long to reach production? Expect 4 to 9 weeks depending on the number of integrations. A pure FAQ bot ships in 4 weeks; an agent connected to the CRM and ticketing takes 8 to 9 weeks.
What maintenance budget should I plan? Between EUR 300 and 900/month for updating the document base, quality monitoring and adjusting guardrails. Without maintenance, relevance degrades within 3 to 6 months.
Is bespoke GDPR-compliant? Yes, with EU hosting and a model that does not reuse your data. That is precisely one of the key arguments against SaaS whose servers are often outside the EU.
Let's scope your project. Give us your ticket volume, your target integrations (CRM, ticketing) and your indicative budget, and we will define the profitable perimeter. 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.
