One AI Chatbot for the whole website can be a great starting point. It can answer common questions, guide visitors, and help them find what they need.
But as your site grows, visitors may need help with very different things — pricing, product choice, support, and more. At this point, one all-in-one assistant may become less effective.
One possible solution is to create separate assistants for specific pages/tasks. Here’s how we use this approach at Elfsight:
This gives each assistant a clearer purpose and makes the conversation more focused.
At the same time, there should be a balance: too many assistants can make the website experience feel fragmented.
[!question] Have you tried using separate chatbots for different pages or tasks?
How did you decide where one assistant was enough and where a focused one made more sense?
I think this is a really useful way to look at it, and one small rule of thumb I’d add is: split the assistant when the visitor’s intent changes, not just because the page changes.
Keep one assistant if:
visitors ask similar questions across pages
the same tone and instructions work well everywhere
the chatbot doesn’t need much page-specific context
Create a focused assistant if:
the page has a very specific goal
visitors need different next steps
the assistant needs a different welcome message, form, or set of instructions
Would love to hear how you approach this on your own sites
Good point and very important. To me, Elfsight’s use of three different chatbots and UIs (two provided above and the third one below) makes your website(s) feel somewhat fragmented or confusing.
For our team, keeping things simple at the frontend is our way of doing business, so we only use one chatbot and UI.
As Richard Branson once said: “Keeping things simple is difficult; making things complicated is easy.”
Great point, Petar, totally agree that simplicity matters here
Quick clarification though: the tool in your screenshot isn’t actually a third AI Chatbot - it’s our support widget, combining instant answers with a way to reach our Support Team if needed.
And your site is a good example of when one chatbot makes perfect sense. Visitors move through a pretty connected journey: they learn about you, check the services, then get in touch or book a call. No sudden context switch, so one well-trained assistant can guide the whole thing.
So yes, I’d say your approach fits the “keep it simple” philosophy perfectly. One chatbot works because it keeps the experience clear.
Have you noticed the chatbot helping visitors make a decision, or is it more of a first-touch tool before they reach out directly?
Thanks for the clarification. In my opinion – from a user’s perspective – all three can be interpreted or viewed as AI Chatbots.
If you ask me, Elfsight should be using only one chatbot and train it using a more advanced LLM so it can capture everything that all three, current chatbots (i.e., widgets) are trying to accomplish. That’s where “keeping things simple is difficult” comes into play.
Our AI chatbot has been trained to handle just about anything related to our business and the industries we serve (aviation, aerospace, and defense). If you have a consulting, business, or technical question, please ask away. You’re welcome to challenge it as many times as you wish – as long as message credits are not deducted from our account. If you don’t like the answers provided by the chatbot, please give them a “thumbs down” to alert us and we’ll fix them.
Fair point, I can see how all these tools may feel like chatbots from a user’s perspective.
And I really like your “one well trained chatbot” approach and how it works in your case! A huge thank you for sharing your thoughts and examples, we really appreciate it
