The chatbot market is full of noise. SaaS vendors promising "deploy in minutes." Agencies quoting $50,000 for an "AI solution." And somewhere in the middle, confused small business owners not knowing what to believe.
This article cuts through it. Here's what an AI chatbot for a small business actually costs, what it actually does, and how to decide if it's worth it for you.
First, What Kind of Chatbot Are We Talking About?
Not all chatbots are equal. There are three main types:
- Rule-based chatbots — decision trees. "Press 1 for X, press 2 for Y." Fast to build, brittle to use. If a customer doesn't phrase their question exactly right, the bot fails.
- SaaS chatbot platforms — tools like Intercom, Drift, Tidio, or Freshchat. You configure the flows, the AI is limited and generic. Monthly subscription of $50–$500/month. You never own it.
- Custom LLM-powered chatbots — built specifically for your business, trained on your products, policies, and tone. Uses models like Claude or GPT-4 under the hood. Understands natural language. Handles complex queries. Escalates gracefully.
When we talk about AI chatbots that actually replace human work, we mean option 3.
What Can a Custom AI Chatbot Actually Handle?
For a small business, a well-built chatbot can handle:
- Answering product/service FAQs (hours, pricing, returns, availability)
- Qualifying leads (budget, timeline, requirements) before handing to sales
- Booking appointments directly into your calendar (Calendly, Google Calendar, practice management systems)
- Handling order status, tracking, and basic support queries
- Collecting customer information before a human needs to get involved
- Running 24/7 — answering enquiries at 2am that used to go unanswered until the next morning
Realistically, a well-built bot handles 60–80% of incoming enquiries without human involvement. The remaining 20–40% are escalated to your team — with full context already captured.
What Does It Replace?
Let's be concrete. Here's what a chatbot typically replaces or reduces:
- Reception / front desk time — businesses with physical or service-based operations often have staff spending 2–4 hours/day on routine phone and chat enquiries. A chatbot can absorb 60–70% of this.
- After-hours misses — leads that came in at 8pm and got a reply at 9am often already contacted a competitor. Instant response converts more.
- VA or customer support staff — a good chatbot handles the volume of 1–2 part-time support roles. At $2,000–$4,000/month for those roles, the math works quickly.
What Does a Custom AI Chatbot Cost?
Here's the honest breakdown:
SaaS Platform (ongoing subscription)
- Cost: $50–$500/month
- You own: nothing
- AI quality: limited, generic
- Customisation: low-medium
- Risk: price increases, platform shutdowns, lock-in
Custom-Built (one-time project)
- Build cost: $1,500–$4,500 depending on scope
- You own: the entire system — code, prompts, integrations
- AI quality: state-of-the-art (Claude, GPT-4)
- Ongoing API cost: $30–$150/month at SMB conversation volumes
- Risk: minimal — you control it
A custom chatbot at $2,800 with $80/month in API costs = $3,760 in year one.
A SaaS platform at $200/month = $2,400/year — every year, forever, with limited AI and no ownership.
By year 2, custom is cheaper. By year 3, it's dramatically cheaper. And custom is better on every quality metric.
How to Calculate Your ROI Before You Spend Anything
Use this quick formula:
- Count how many customer enquiries (chat, phone, email) you receive per month
- Estimate what percentage are routine (FAQs, bookings, status checks) — usually 60–75%
- Calculate how much staff time that represents (hours × hourly cost)
- Add the value of missed after-hours leads (leads × average deal value × close rate)
If that number is over $1,000/month, you have a clear ROI case for a chatbot. Most businesses we talk to find it's $2,000–$5,000/month once you account for everything.
What to Ask Before Hiring Anyone to Build One
- Do I own the code and system when it's done? (The answer must be yes.)
- What happens if I want to change something after launch?
- How does it handle questions it doesn't know the answer to? (Escalation to human is critical.)
- How is it trained on my business?
- What are the ongoing running costs?
The Bottom Line
An AI chatbot isn't a magic solution — it needs to be properly built, trained on your business, and integrated with your systems. Done right, it's one of the highest-ROI investments a small business can make.
Done wrong (generic SaaS bot, badly configured, no integration), it frustrates customers and gets turned off within 3 months.
The difference is in the build quality — and having clear ownership of the system.