Every founder asks this question eventually. Your support inbox is full. Your team answers the same questions every week. Someone suggests "just get a chatbot."
Then you start researching and the quotes are all over the place. Some vendors sell a cheap monthly plan. Others ask for a large upfront payment plus a monthly fee. A few send you a contact form and nothing else.
The reason is simple. AI chatbot costs depend on what you are actually building. A rule-based bot that answers FAQs is a small job. An AI agent that books appointments, reads from your CRM and passes complex issues to a person is a much bigger one.
This article explains what drives the cost, so you can budget sensibly.
Rule-Based Bots vs LLM Agents: The Cost Gap
Rule-based chatbots follow decision trees. The user says X and the bot replies with Y. They are predictable, easy to debug and cheap to run. You can set one up on a hosted platform with a low monthly plan, or have a developer build a simple one for your site.
The problem is that they break when someone asks something you did not plan for. "What is your return policy?" works. "Can I return this if my dog chewed the box but the product is fine?" does not.
LLM-powered agents (built on models such as GPT-4 or Claude) understand context. They can handle open-ended questions, pull data from your systems and reply in natural language. They cost more to run because every conversation calls an API that charges by usage.
In rough order of cost, from lowest to highest:
- Rule-based bot: lowest setup cost and a small monthly hosting and maintenance bill.
- Basic LLM chatbot (no integrations): a moderate setup cost. Monthly spend depends mainly on traffic.
- Production LLM agent (CRM integration, knowledge base, escalation logic): the highest setup cost, with monthly spend covering API usage, hosting and upkeep.
We build AI agents for tasks such as lead qualification, customer support and appointment scheduling. You can see the range on our AI agents page. Most small businesses do not need the most advanced option on day one.
Setup Costs: What You're Actually Paying For
Setup is not just installing a script. These are the things that drive the cost:
Knowledge base preparation. Your bot needs good source material. If you have clean documentation, this is quick. If your knowledge lives in old chat threads and scattered documents, someone has to organise it first. That takes real time.
Integration work. Does the bot need to check stock, pull order status from your ERP or create tickets in your CRM? Each integration adds development and testing time. The quality of the other system's API also matters. A well-documented API is faster to connect to than a legacy system with none.
Prompt engineering and testing. LLMs are powerful but unpredictable. You need to test edge cases, refine instructions and add guardrails so the bot does not make things up or drift off topic. This is often one of the largest parts of the project.
Interface design. A chat widget that looks dated hurts trust. It should look right, work on mobile and match your brand.
Escalation logic. When should the bot hand over to a person? How is your team notified? What happens if a customer asks for a manager? These rules take careful thought.
A small business can think about scope in three tiers:
- Basic chatbot: FAQ handling, no integrations. The smallest build.
- Mid-tier agent: CRM integration, order lookup and lead capture. A noticeably larger build.
- Advanced agent: multi-step workflows, payments and an analytics dashboard. The largest build, and the one that needs the most testing.
Most businesses do well starting with a mid-tier agent and expanding once they see how customers use it.
Monthly Run Costs: API Bills and Hosting
This is where LLM agents differ from traditional software. Every conversation has a usage cost.
API costs depend on:
- Model choice. More capable models cost more per request. Smaller models are cheaper but handle complex questions less well. Check the current price list from your provider, as these change often.
- Conversation length. A quick FAQ answer uses very little. A long troubleshooting session uses many times more.
- Traffic volume. More conversations means a higher bill. Estimate your monthly volume before you choose a model.
A practical way to forecast this is to take your average monthly conversations, estimate the length of a typical one, and multiply by the provider's usage rates. Then add a buffer for busy months.
Hosting and infrastructure. If the bot runs on your own server, you pay for that server. This suits businesses that want control over their data and are building a custom solution. Hosted platforms usually include infrastructure in their monthly fee, but you give up some flexibility.
Maintenance. Bots need regular attention. Products change, policies change and prompts need refining. If you have a support agreement with your development team, this may already be covered. If not, set aside a monthly allowance for updates and fixes.
Where the ROI Actually Comes From
An AI chatbot is only worth it if it replaces something more expensive, or earns something you were missing.
Here is a simple way to check. Work out how many hours your team spends each month on repetitive questions. Multiply that by the real hourly cost of those staff, including overheads. Then estimate what share of those questions a bot could realistically handle. Compare the result with the setup and running costs from the sections above. If the savings do not clearly cover them within a reasonable period, start smaller.
Other places value can come from:
- Lead qualification. A bot on your homepage can capture contact details and book calls. It can reduce the load on a sales or business development person.
- 24/7 availability. Small teams cannot staff support around the clock. A bot can. If prospects write in the evening and wait a day or more for a reply, you may be losing some of them.
- Faster response times. Quick replies generally help conversion. A bot answers in seconds.
- Upsell opportunities. A bot can suggest add-ons or related products based on what the customer is asking about.
There is a trade-off. If your business relies on high-touch, consultative sales, a chatbot may do more harm than good. No LLM can replace a skilled salesperson who builds rapport and reads between the lines. Know when not to automate.
What TechNova Clients Actually Pay
We do not publish a price list for AI agents, because no two projects are the same. Cost depends on the factors above: the number of use cases, how many systems need connecting, the state of your existing content, and the expected volume of conversations.
In practice, we scope projects in stages:
- Starter: one use case, minimal integrations.
- Growth: a few use cases, with CRM and knowledge base integration.
- Custom build: multiple agents, advanced workflows and analytics.
Starting small keeps the first investment manageable. It also gives you real conversation data before you commit to a larger build. If cash flow is a concern, tell us early. We can discuss phasing the work or agreeing a payment structure that suits you.
We have built business software for clients across several industries, including Best Estate Marketing, Bismillah Traders and EstateFlow, and we bring the same practical approach to AI work. You can browse examples on our case studies page.
Summary
For a small business, the cost of an AI chatbot comes down to four things: how complex the bot is, how many systems it connects to, how much usage it gets, and how much upkeep it needs. A rule-based bot costs less. A bot with deep links into your CRM or ERP costs more.
A sensible budget covers three parts: the build, the monthly running costs, and a small allowance for improvement after launch. Start with one clear use case, measure the results, then expand.
The better question is not "how much does it cost?" but "how much time and revenue am I losing without one?" If your team answers the same questions all day, that is worth looking into.
To see what is possible, look at our AI agents or get in touch. We will talk through your use case and give you a clear written quote.
