Subscribe to GEN
Login to GEN
A good chatbot answers the question quickly, from what your business has actually published, and knows when to hand over to a person. A bad one goes round in circles, invents answers and makes promises you would never make. Most customers have met far more of the second kind.
GEN-Chat is our own website chatbot, built to be the first kind. It answers from your content on local models, admits when it does not know, and brings a member of your team into the conversation when the customer needs one.
Most chatbot failures have the same cause: the model answers from whatever it learned in training, and fills the gaps with confident invention. GEN-Chat works the other way round. Your website, knowledge base, FAQs and documents are indexed, and for every question the most relevant passages are found first and given to the model with it. Anything not relevant enough is left out, so the answer comes from your material or not at all.
It can use live information too. Our own deployment reads the current service status, so a customer asking about a fault is told what is actually happening. For stock, orders or bookings, GEN-Chat connects to a backend of your own and answers from the system that holds the facts.
A chatbot that keeps a frustrated customer talking to it does more damage than having no chatbot at all. GEN-Chat measures two things as the conversation goes on: how certain the model is of its answers, and whether the customer is becoming unhappy. When either passes a threshold, it escalates.
A member of your team is notified and joins the same conversation through a secure link, with the whole history in front of them, so the customer does not have to start again. When the matter is dealt with, they resolve the conversation and it is kept for review.
Outside the hours your team is available, it does not leave the customer waiting for someone who is not there. It takes their details instead, so the enquiry is picked up first thing.
Added to any website with a single tag. The widget follows your site's light or dark theme, and its position, greeting and appearance can be set to suit.
The conversation carries on as the visitor moves from page to page, and can pop out into its own window if they want to keep it open while they browse.
It only runs on the domains you authorise, inside a sandboxed frame, and each site has its own configuration, content and backend.
Every session is visible: live and past conversations, the full transcript, and the pages the visitor viewed on the way, so your team can see what customers ask, where the bot helped and where it needed a person.
Our retail demonstration store has GEN-Chat answering questions about its stock. Ask it for a particular item in a particular size, and see what it does.
From your own content. Your website, knowledge base, FAQs and documents are indexed, and for each question the most relevant passages are found and given to the model with it. Passages that are not relevant enough are left out rather than stretched to fit, so the bot answers from what you have published, not from what a model picked up elsewhere.
It says so, and hands over to a person. GEN-Chat watches for uncertainty in the model's answers and for signs that the customer is frustrated, and either one can bring a member of your team into the same conversation through a secure link, with the whole history in front of them. Outside your team's hours it takes the customer's details instead, so nobody is left waiting.
Not with GEN's own models, which run locally on our hardware. GEN-Chat can also connect to a backend of your own, such as your stock or order system, so answers about live information come straight from the source.
With a single script tag. The widget matches your site's light or dark theme, keeps the conversation going as visitors move between pages, and can pop out into its own window. It only runs on the domains you authorise.
Yes. Every conversation is kept with its full transcript and the pages the visitor viewed, so your team can see what people asked, where the bot helped, and where it needed a person.
Yes. Our retail demonstration store has GEN-Chat answering questions about its stock, at chat.gensoftware.dev.
Send us the questions your team answers most often, and tell us where the answers live. We will tell you how many GEN-Chat could handle from your own content, which need live data from your systems, and which should always go to a person.
For answering questions from internal documents rather than on your website, see business AI.