How much does building a chatbot cost?
From a simple FAQ bot to fully integrated conversational AI that lives within your business systems: costs range from a few thousand to tens of thousands of euros. This article explains where those differences come from, so you can draw up a realistic budget.
No standard price, but a spectrum
The question "what does a chatbot cost?" is like asking "what does a car cost?" It depends entirely on what you need. A chatbot that answers frequently asked questions based on a fixed decision tree is a fundamentally different product from an AI assistant that consults your order system, CRM and knowledge base in real time.
The biggest cost variable is not the technology itself, but the depth of integration with your existing systems. A standalone chat widget on your website can be realised within a limited budget. Once the chatbot needs to look up orders in your ERP, schedule appointments in your calendar or update customer records in your CRM, the complexity grows, and with it the investment.
The chosen AI approach also plays a role. Older chatbots work with predefined flows (intent-based). Modern conversational AI uses large language models (LLMs) that understand natural language and generate contextual responses. The second route requires more knowledge of prompt engineering, RAG architecture (Retrieval-Augmented Generation) and vector databases, but delivers a considerably better product.
Three typical budget levels for chatbot development
The breakdown below gives a realistic picture of what you can expect at each budget level. The amounts are indicative: every project is custom, and the final investment depends on your specific requirements and technical environment.
FAQ chatbot or decision tree
A chatbot that answers frequently asked questions based on a fixed set of answers. Well suited to relieving customer service of repetitive questions about opening hours, delivery terms or product information.
- Fixed question-and-answer flows
- Web chat widget on your site
- Simple fallback to email
- Basic analytics dashboard
- Training on your content
AI chatbot with system integration
A chatbot that, alongside free-text conversation, also consults your business systems. Think of checking order status, scheduling appointments, or searching a knowledge base using RAG technology.
- LLM-based (GPT-4, Claude, Gemini)
- RAG with your own documentation
- API integrations with CRM/ERP
- Multichannel (web, WhatsApp, email)
- Handover to a live agent
- Extensive analytics and logging
Fully autonomous AI assistant
A conversational AI that independently carries out tasks within your processes: processing orders, generating contracts, escalating tickets. Includes a compliance layer, audit trail and on-premises or private cloud hosting.
- Agentic AI with tool use
- Deep integration into business processes
- Custom fine-tuning or private LLM
- RBAC and audit logging
- On-premises or private cloud option
- Availability SLA
- Ongoing management and optimisation
Seven factors that determine the cost
Every chatbot project is unique. Nevertheless, there are seven factors that influence the budget in almost every project. Below we explain how each factor affects the outcome.
Complexity of the conversation
A linear decision tree with ten questions is fundamentally simpler than an open conversation in which the bot must retain context across multiple messages. The more open the conversation, the more NLU logic is required.
Number of channels
A web chat widget alone is simpler than a bot that must also be active on WhatsApp, Facebook Messenger, Teams and email. Each channel has its own API requirements and UX constraints.
System integrations
Each integration with an external system (CRM, ERP, ticketing system, knowledge base) requires API development, authentication handling and error handling. Two integrations do not double the cost, but they add considerably.
Compliance and security
Does the chatbot process personal data? Then GDPR requirements apply: data minimisation, retention policies, right to erasure. In sectors such as finance and healthcare, sector-specific requirements come on top (DNB, NEN 7510).
Choice of AI model
A chatbot based on a commercial LLM (OpenAI, Anthropic, Google) has lower development costs but ongoing API costs per conversation. A self-hosted open-source model (Llama, Mistral) requires more upfront work but can work out cheaper at high volumes.
Size of the knowledge base
Does the chatbot need to search ten FAQ pages or thousands of product sheets, manuals and contract documents? In RAG implementations, the volume of source data determines the complexity of indexing, chunking and retrieval tuning.
Three routes: SaaS platform, custom or hybrid
There are broadly three ways to realise a chatbot. Each has its own advantages and disadvantages in terms of cost, flexibility and management.
| Aspect | SaaS chatbot platform | Custom development | Hybrid approach |
|---|---|---|---|
| Initial budget | Low — subscription costs | High — one-off build | Medium: SaaS core + custom layers |
| Ongoing costs | Fixed monthly fee + per-conversation fees | Hosting + AI API costs + maintenance | Subscription + maintenance of custom layers |
| Flexibility | Limited to platform features | Fully tailored | Flexible where it matters |
| Time to market | Fast — weeks | Long — months | Average |
| Depth of integration | Standard integrations | Unlimited | Extensible |
| Suitable for | Simple customer service | Complex business processes | Growing organisations |
Appfront often takes a hybrid approach: we use proven AI platforms as a foundation (OpenAI, Google Gemini) and build custom integrations and business logic around them. That way you get the best of both worlds: a solid AI core with the flexibility to integrate deeply into your systems.
Ongoing costs: don't forget the running costs
The build costs are only the beginning. A chatbot running in production carries ongoing monthly operating costs. This is often underestimated in the initial budget.
Fixed costs
- Hosting and infrastructure (cloud compute, vector database)
- LLM API costs per conversation (tokens in/out)
- Monitoring and logging tooling
- SSL certificates and domain management
Variable costs
- Knowledge base updates when products or policies change
- Prompt optimisation following analysis of failed conversations
- New integrations as your systems landscape changes
- LLM version upgrades (GPT-4 → newer model)
As a rule of thumb, expect monthly operating costs equal to a fraction of the initial build costs. The exact ratio depends on conversation volume and the complexity of your integrations. Alongside the initial build, Appfront also offers ongoing management, so your chatbot keeps performing at its best.
How does a chatbot project work?
A typical chatbot project runs through five phases. Each phase has a clear go/no-go decision point, so you stay in control of the budget.
Discovery and scope definition
We map out your use case: which questions the chatbot answers, which systems it communicates with, and which channels are needed. The result is a functional design with a clear scope and budget range.
Proof of Concept
A working prototype that handles the core conversations. Often built with a limited knowledge base and one or two system integrations. It lets us test whether the chosen approach delivers the quality you want.
Full development
The chatbot is expanded with all integrations, channels and business logic. Including edge-case handling, fallback flows and the integration with your live systems in a staging environment.
Testing and optimisation
User testing with your team and a select group of end users. We measure success rate, average handling time and customer satisfaction. Prompt engineering is fine-tuned based on real conversations.
Go-live and continuous development
Rollout to production with monitoring. After go-live comes a period of intensive analysis in which we review conversation logs and continuously improve the chatbot. This is not a one-off task but an ongoing process.
Frequently asked questions about chatbot costs
After delivery, you pay for hosting, LLM API usage (per conversation/per token), and optional maintenance. At low volumes the API costs are negligible; at thousands of conversations per month they can add up. A management contract for knowledge base updates and prompt optimisation is recommended for the best results.
Certainly. Many of our clients start with an FAQ chatbot or a limited RAG solution and expand it step by step with system integrations and additional channels. A phased approach limits risk and gives you the chance to decide, based on real usage data, which extensions deliver the most value.
A traditional chatbot works with predefined flows: if the user says X, reply Y. Conversational AI uses a language model that understands natural language and responds in context. The difference lies in flexibility: conversational AI can handle unexpected questions, synonyms and complex sentence structures. The costs are higher, but the result is a significantly better user experience.
A simple FAQ chatbot can go live within a few weeks. An AI chatbot with multiple system integrations and channels typically takes several months. The timeline depends heavily on the availability of API documentation for the systems to be integrated and on how quickly your organisation gives feedback during the testing phase.
Yes, provided you apply privacy by design from the outset. That means clearly defining which data the chatbot processes, where it is stored, how long it is retained, and how users can view or have their data deleted. When using commercial LLM APIs, it is important to choose providers that offer European data processing, or to consider a self-hosted model.
Modern LLMs support dozens of languages out of the box. The challenge lies not in the language model itself but in the knowledge base: if your documentation is only available in Dutch, we will need to either add translations or implement cross-lingual retrieval. This adds cost, but is technically very feasible.
Cost estimate for your chatbot project?
We are happy to think along with you about the right approach and the corresponding budget. In a no-obligation conversation, we will map out your requirements and provide a realistic estimate, with no strings attached.
A chatbot is one of several channels. On applatenmaken.com you will find the broader overview of customer contact and service software.