The short version
A small business can run AI customer support for under 50 dollars per month and handle thousands of conversations. A medium business might spend 200 to 500 dollars per month. Enterprise plans run 1000 dollars and up.
That is the headline. Now let us look at what is actually in those numbers.
The three cost components
Every AI chatbot setup has three cost layers:
- The AI model fees (OpenAI, Anthropic, etc.)
- The platform fees (the software that runs your chatbot)
- Your time setting it up and maintaining it
People usually only think about the second one. That is a mistake.
AI model costs are surprisingly cheap
Modern AI models are dramatically cheaper than they were two years ago. Here are typical costs in 2026:
GPT-4o mini: Around 0.0015 dollars per conversation GPT-4o: Around 0.01 dollars per conversation Claude Haiku: Around 0.001 dollars per conversation Claude Sonnet: Around 0.008 dollars per conversation
For a business with 1000 conversations per month, the AI bill is 1 to 10 dollars total. That is less than a single hour of a support agent.
The cost is so low that worrying about it is a waste of time. Pick a model based on quality and speed, not price.
Platform fees are where the real money goes
This is where you actually pay. Platforms charge for:
- Hosting the chatbot
- Storing your training documents
- Providing the dashboard and analytics
- Handling integrations with Slack, WhatsApp, etc.
- Customer support for you
- Updates and security
Typical pricing in 2026:
Free plans: Most platforms offer 50 to 200 free messages per month. Good for testing.
Starter plans (10 to 30 dollars per month): 1000 to 5000 messages, 1 chatbot, basic features. Perfect for small businesses.
Growth plans (50 to 150 dollars per month): 10000 to 30000 messages, 3 to 5 chatbots, advanced features like API access.
Scale plans (200 to 500 dollars per month): 50000 plus messages, unlimited chatbots, white label options, priority support.
Enterprise (1000 dollars and up): Custom limits, SLA, dedicated support, security features like SSO and data residency.
The hidden cost: your time
This is what nobody tells you. Setting up an AI chatbot takes time even with the easiest platforms.
Initial setup: 2 to 4 hours
- Creating the account
- Uploading documents
- Configuring the system prompt
- Embedding the widget
- Testing common questions
Content prep: 4 to 16 hours (one time)
- Writing or organizing help documents
- Cleaning up FAQ pages
- Creating training material the bot can learn from
Ongoing maintenance: 1 to 2 hours per week
- Reviewing bot conversations
- Updating outdated content
- Improving answers that were wrong
- Monitoring analytics
If you value your time at 50 dollars per hour, the hidden cost in the first month is 300 to 1000 dollars. That is more than the platform fees.
Comparison with hiring a support agent
This is the comparison people care about most.
A part time support agent in the US costs around 2000 to 3000 dollars per month. Full time is 4000 to 6000 dollars. They handle maybe 30 to 50 conversations per day.
An AI chatbot handles unlimited conversations for 30 to 200 dollars per month. It works 24/7. It never calls in sick.
The math looks like this:
| Setup | Monthly Cost | Conversations Handled | |---|---|---| | 1 part time agent | 2500 dollars | ~600 | | AI chatbot only | 100 dollars | unlimited | | Hybrid (1 agent + AI) | 2600 dollars | unlimited |
The hybrid setup is what most businesses end up with. You save on agents because the AI handles the easy questions, freeing the human to focus on the hard ones.
DIY vs platform: which is cheaper
You might think building it yourself is cheaper than paying a platform. It usually is not.
Building it yourself:
- 4 to 8 weeks of developer time (10000 to 30000 dollars in salary)
- Ongoing maintenance (5 to 10 hours per week forever)
- Vector database costs (Pinecone, Weaviate, etc.)
- Hosting and infrastructure
- No analytics dashboard (unless you build one)
- No connectors to Slack, WhatsApp, etc.
Total first year cost: 30000 to 80000 dollars.
Using a platform:
- 100 to 500 dollars per month
- Setup in a few hours
- Everything included
Total first year cost: 1200 to 6000 dollars.
The math is obvious. Build it yourself only if you have a specific technical reason to do so.
Costs nobody warns you about
A few sneaky costs:
Per agent fees: Some platforms charge per team member who can access the dashboard. Watch for this if you have a big support team.
Conversation limits: Going over your plan can be expensive. Some platforms charge 0.10 dollars per extra message. At scale this adds up fast.
Premium model fees: Cheaper plans use cheaper AI models. If you want GPT-4 quality, you usually pay extra.
Storage fees: Uploading lots of documents can hit storage limits and trigger upgrades.
API access: If you want to integrate the chatbot into a custom app, this is often a paid feature.
White label: Removing the platform branding is usually a paid feature.
How to keep costs down
A few tips:
- Start on the free plan and only upgrade when you hit limits
- Use cheaper AI models like GPT-4o mini or Claude Haiku
- Write a focused system prompt so the bot does not waste tokens
- Limit the conversation history sent to the model
- Cache common answers if possible
- Review usage monthly and downgrade if you are over paying
When to spend more
Pay for the better tier if:
- You handle sensitive information and need security features
- Your team is large and needs collaboration tools
- Your traffic is high enough to justify the cost
- You need integrations the cheaper plan does not include
- The cheaper plan is too restrictive on AI model quality
The bottom line
For most small to medium businesses, the right setup costs between 30 and 200 dollars per month. That is a tiny fraction of what hiring even one part time agent would cost, and you handle far more conversations.
The technology has gotten cheap enough that the question is no longer "can we afford this?" but "why are we still answering the same five questions all day?"
If you have not added an AI chatbot to your business yet, you are spending money the slow way.