Key takeaways
- Scope, integrations and data preparation usually drive cost more than the choice of language model.
- Running costs (model usage, hosting, monitoring and maintenance) continue after launch and belong in every budget.
- Off-the-shelf tools are cheaper to start; custom builds pay off when workflows, data or compliance needs are specific.
- Published price ranges are only a guide; a short discovery phase gives a far more reliable estimate.
- Estimate ROI from measurable outcomes such as deflected tickets, faster responses and recovered after-hours leads.
Why AI chatbot cost varies so widely
Ask ten suppliers how much an AI chatbot costs and you may receive ten very different answers. That is rarely because anyone is being evasive. The word "chatbot" covers everything from a website widget that answers a few common questions to an assistant that reads your order system, books appointments, escalates complaints and logs every conversation in your CRM. These are different products with different engineering effort behind them.
The useful question is therefore not "what does a chatbot cost?" but "what does the chatbot I need cost?". Before you compare quotes, write down what the assistant must do, which systems it must reach, which channels it must run on and who will look after it once it is live. With that scope in hand, prices become comparable and the reasons for differences become visible.
The main cost drivers
Most of the budget for an AI chatbot is spent on the work around the model rather than the model itself. The factors below account for the bulk of the difference between a modest project and a large one. When reviewing a proposal, check that each one has been considered, even if the answer is "not needed for the first release".
- Scope: the number of tasks, conversation flows and languages the assistant must handle.
- Integrations: connections to your CRM, helpdesk, booking, e-commerce or internal databases, each needing design, testing and error handling.
- Data preparation: cleaning, structuring and approving the documents and FAQs the assistant will answer from.
- LLM usage costs: per-token charges from model providers, which scale with conversation volume and answer length.
- Voice: speech recognition, speech synthesis and telephony add components, latency work and per-minute costs.
- Compliance: data residency, retention rules, audit logging, access control and reviews for regulated sectors.
- Maintenance: monitoring, content updates, model upgrades and ongoing quality reviews after launch.
Typical market price ranges
The ranges below are broad figures commonly cited across the industry for custom projects. They are typical market ranges for orientation only, not Gen7Ven quotes, and real projects can fall outside them in either direction depending on scope, region and supplier. Treat them as a way to sense-check a proposal rather than as a price list.
Ongoing costs sit on top of the build. Model usage is often modest for a small business and grows with traffic, while monitoring, hosting and content updates are usually charged as a monthly retainer. Ask every supplier to separate one-off build costs from recurring costs so that you can compare the total cost of ownership over two or three years.
| Tier | What it typically includes | Typical market range (USD, one-off build) |
|---|---|---|
| Basic FAQ assistant | Answers from a small, approved knowledge base on one channel, with simple handoff to a person | Roughly $5,000 to $20,000 |
| Integrated business assistant | Retrieval over company documents, two or three system integrations, analytics and escalation rules | Roughly $20,000 to $80,000 |
| Advanced or multi-channel agent | Takes actions in several systems, web plus messaging or voice, custom workflows and reporting | Roughly $80,000 to $200,000 |
| Enterprise programme | Multiple agents, strict compliance, security reviews, high volumes and dedicated support | Often $200,000 and above |
Build vs off-the-shelf: which is better value?
Off-the-shelf chatbot platforms charge a monthly subscription and can be running within days. They suit businesses whose needs match the product: a standard helpdesk, common integrations and modest customisation. The trade-off is that you work within the vendor's features, pricing tiers and data policies, and costs can rise as conversation volumes grow.
A custom build costs more up front but fits your processes, data and compliance requirements exactly, and you control the roadmap. Many businesses take a middle path: they start with a platform to prove demand, then commission a custom assistant once the value is clear and the gaps are known. The table summarises the main differences.
| Factor | Off-the-shelf platform | Custom build |
|---|---|---|
| Upfront cost | Low; mostly setup and configuration | Higher; design, development and testing |
| Running cost | Subscription that often scales with usage or seats | Hosting, model usage and maintenance |
| Time to launch | Days to a few weeks | Several weeks to a few months |
| Integrations | Limited to the vendor's connectors | Any system with an accessible API |
| Data control | Governed by the vendor's policies | Defined by you, including hosting region |
| Differentiation | Similar to competitors on the same tool | Tailored to your workflows and brand |
How to estimate the ROI of an AI chatbot
Start with the work the assistant will take on. Count the enquiries your team handles each month, the share that are repetitive and the average time each one takes. Multiply the conversations the assistant can realistically resolve by the cost of handling them manually. That gives a conservative saving before any revenue effects are included.
Then add the upside: leads captured outside office hours, faster response times that reduce abandoned enquiries, and staff time redirected to higher-value work. Subtract the build cost and the annual running cost, and calculate the payback period. A credible business case uses your own figures, starts with cautious assumptions and is checked against real data during the first months after launch.
How to get an accurate quote
The quickest route to a reliable price is a short, well-prepared brief. Include the top tasks the assistant must handle, sample questions from real customers, the systems it must connect to, expected monthly volumes and any compliance constraints. Suppliers can then price the work instead of guessing, and you can compare proposals on equal terms.
For larger projects, a paid discovery phase of one to three weeks is worth considering. It produces a scoped plan, a technical approach and a firmer estimate, and it reduces the risk of costly changes later. If you would like a second opinion on scope or budget, Gen7Ven is happy to review your brief and suggest a sensible first release.
Ready to talk about your project? Explore AI Agents & Chatbots or send us a brief.
AI Agents & Chatbots →