Victor Arias
An AI chatbot makes sense when you get the same 5-10 questions every day, you have out-of-hours traffic and quality documentation; it hurts the experience if you can't guarantee answers or if it hides human support. Typical cost: $1,200-3,500 to implement plus $80-300/month for an AI chatbot with lead capture. Success is measured by qualified leads and hours returned to the team, not by messages handled.
A visitor lands on your website at 11:40 pm, has three questions about prices and timelines, and nobody is there to answer. Or it's a repeat customer asking the same status question every week, or your support team spends half its time answering the same thing. AI chatbots have been promising to solve this for years; in 2026 they're no longer a toy, but they're still a bad project when they're implemented "because everyone has one".
The difference between a chatbot that generates customers and one that drives visitors away isn't the AI model: it's the design, the data you give it and when it decides to stay quiet and hand the conversation to a person. In this article I explain what a chatbot can actually do on your website, when it makes sense (and when it doesn't), what it costs to implement and how to measure whether it's working.
What an AI chatbot is (and what it isn't)
Rule-based chatbot vs generative AI
There are two big families, and confusing them explains most of the poor results:
- Rule-based chatbot (decision tree): follows a script you define. Cheap, predictable and perfect for 5-6 fixed questions (opening hours, basic prices, "how do I return an item?"). If the visitor goes off-script, it gets stuck.
- Generative AI chatbot: understands natural language and answers using your documentation as the source. Costs more, requires curation and is far better at conversation: it resolves doubts you never anticipated.
The right option is almost always hybrid: rules for the predictable and AI for everything else, with a clear criterion for when to hand over to a human.
What it can really do
With a well-built implementation, a chatbot on your website can:
- Resolve purchase objections (prices, timelines, availability) exactly when the visitor is deciding.
- Qualify and capture leads: it collects name, email, need and budget before a human gets involved.
- Reduce support load by answering the usual recurring queries.
- Guide through the catalog: in stores, it helps find the right product in two questions.
- Book appointments or hand over to WhatsApp when the conversation gets serious.
The best use case isn't "replace the team", it's giving them back the hours spent answering the same thing.
When it makes sense (and when it doesn't)
Cases where it really works
- You get the same 5-10 questions every day by email or WhatsApp.
- You have traffic outside your working hours (another country, weekends, early mornings).
- Your sales process has clear steps the chatbot can walk through and book.
- You have quality documentation: FAQs, terms, product sheets, up-to-date prices.
Cases where it makes the experience worse
- You can't guarantee the answers. A chatbot that invents prices or deadlines destroys trust faster than having none.
- Your business is highly sensitive (health, legal, finance) and every answer has consequences.
- You use it to hide support. If the chatbot has no clear path to a person, it turns frustration into bounce.
- Nobody maintains it. An out-of-date chatbot is worse than none at all.
If the visitor repeats the question three times and never speaks to anyone, the chatbot is losing the sale.
How much a chatbot costs for your website
Realistic market ranges for 2026 (project plus monthly subscription):
| Type | Initial project | Monthly |
| Basic rule-based widget (5-10 flows) | $300 - $900 | $0 - $50 (tool) |
| AI with your documentation + lead capture | $1,200 - $3,500 | $80 - $300 (API + hosting) |
| Integrated with your CRM/store (orders, appointments, tickets) | $3,500 - $9,000 | $250 - $800 |
| Enterprise multichannel solutions | from $9,000 | volume-based |
The typical mistake is looking only at the monthly fee: a cheap, poorly trained chatbot costs a lot in lost sales. What really moves the needle is answer quality and integration with your sales flow.
How to implement it well: 4 steps
1. Knowledge sources and limits
Gather what you already know how to answer: FAQs, prices, policies, product sheets and real cases. Define what it must NOT say (timeline promises, discounts, legal topics) and what to do when it doesn't know: say so honestly and escalate.
2. Conversation, handoff and metrics
- Clear path to a human in one click (WhatsApp or form), always visible.
- Conversation logs to review failed answers weekly.
- Minimum metrics: % resolved without intervention, leads captured, transfers to human and closing satisfaction.
If the chatbot captures leads, those leads must land in your CRM and your follow-up, not in a forgotten spreadsheet. If your priority is the website behind it, start with a review of your WordPress or store to make sure speed and tracking are right before adding more layers.
3. Test with real questions
Ask it the questions your customers actually ask you (not the ones you think they ask). If it answers more than 1 in 10 wrong, go back to step 1 before publishing.
4. Gradual rollout
Start with a small scope (one area, one channel), measure for two weeks and expand. Correcting a narrow chatbot is far cheaper than fixing a full rollout.
How to know if it's working
Forget "messages handled" as your main metric. What matters:
- Qualified leads generated and their close rate versus other channels.
- Hours returned to the team through queries resolved without intervention.
- Transfer rate to a human: very high = poorly trained; very low with unhappy customers = it's hiding the problem.
- Mid-conversation abandonment: the best indicator that the answers aren't useful.
With those four numbers you'll know within two weeks whether the chatbot pays for itself or needs adjusting.
Conclusion: a chatbot is a channel, not a trick
AI doesn't save a website with bad information or a confusing sales process: it amplifies what you already are. Done well, it answers when you can't, captures when the visitor hesitates and gives hours back to your team. Done badly, it's a revolving door that frustrates and pushes people away.
If you want to know whether a chatbot fits your business and what it would cost to implement it properly, tell us about your case: we'll propose a concrete scope, with metrics and no promises of magic robots.
Frequently asked questions
What is an AI chatbot and how is it different from a classic one?
A classic one follows a decision tree; an AI one understands natural language and answers from your documentation. A hybrid setup is usual. We implement it on your website.
How much does it cost to add a chatbot to my website?
$1,200-3,500 for the project plus $80-300/month for an AI chatbot with lead capture. Get a quote with a concrete scope.
Can a chatbot actually bring in customers?
Yes, if it collects name, email and need before handing over to a person, and those leads land in your CRM. We set that up in a free review.
When should I NOT use a chatbot?
If you can't guarantee answers, your sector is highly sensitive or nobody will maintain it. Fix the process before automating it badly.
Does it work for online stores?
Yes: it guides through the catalog, resolves shipping and returns questions and helps close the sale. We integrate it into your store.
What happens when the chatbot doesn't know?
It should say so and escalate to a human in one click. If you hide support, the chatbot loses sales. That handover is also covered by our IT support.
