Chatbits

AI chatbot platform as SaaS

Industry: Software & IT

Chatbits is our own product and a brand of the Convecto Group. The European platform turns company knowledge into AI chatbots. It reads a company’s websites, documents and texts, and the chatbot answers customers’ questions from them. For certain questions it brings up “Actions” in the chat: a form that collects leads, products from the company’s own online shop, or a button that leads to a set destination. We built Chatbits in Laravel, Python and PostgreSQL and offer it as a SaaS product of our own. The name, brand identity, corporate design with styleguide and the chatbits.ai marketing website are ours as well.

Services

Brief

Mid-sized companies in Germany, Austria and Switzerland had no European platform for AI chatbots with data protection that went far enough. The aim was a chatbot that takes in a company’s knowledge from websites, documents and texts and answers customer questions around the clock. The data had to stay in Germany, and setting it up should need no programming skills. Conversations were also meant to turn into enquiries: anyone asking in the chat for a call back or a product should be able to get there in the chat.

How it works

From company knowledge to an answer in the chat

Questions the chatbot cannot answer end up in the back end under “Unanswered questions”.

Result

Chatbits learns from whole websites or single pages, from Word, PowerPoint and PDF files and from pasted text. If you want, the platform retrains at fixed intervals, so the chatbot also knows about pages that have changed. Each chatbot goes onto the website with a JavaScript snippet or the WordPress plugin and answers in 120 languages. Questions it cannot answer are stored so the team can train it on them. Companies set up the “Actions” themselves in the back end. The data sits in an ISO/IEC 27001 certified data centre in Germany and is not used to train general language models.

Back end

“Collect leads” form in the Chatbits back end with the recipient email and a field for when to use it, a chat preview beside it

File list in the Chatbits back end with two uploaded PDFs

In the back end, the company decides when the chatbot collects leads or shows a button of its own.

In figures

120
Languages the chatbot answers in: 3
Steps: train, embed, answer: 4
Plans from Free to Pro: 44
Pages of styleguide

Project data

A product of our own, from the code to the brand

Product: Platform for AI chatbots built from company knowledge, as SaaS
Platform: Laravel, Python and PostgreSQL
Knowledge sources: Websites and single pages, Word, PowerPoint and PDF files, pasted text
“Actions”: Collecting leads, products from the online shop, links and buttons to set destinations
Integration: A JavaScript snippet per chatbot or the WordPress plugin
Operation: ISO/IEC 27001 data centre in Germany, customer data never trains general language models
Billing: Four plans from Free to Pro

Laravel
Python
PostgreSQL
SaaS
GDPR

Marketing website

www.chatbits.ai

The chatbits.ai home page on a smartphone with the same headline and its benefits as a list

We also designed and built the chatbits.ai marketing website. It is public and separate from the back end of the app.

The AI chatbot for awesome customer experiences.

Name, logo and corporate design

The name, brand identity and corporate design are ours. The primary logo puts the name in capitals inside a speech bubble, with “.ai” small and set low. Its square dot is the “bit”, which comes back across the design as a pattern and as a typing indicator of three bits in the chat. The colours are Chatbits blue, white and anthracite; on white, the blue reaches a contrast of 8.5:1. The brand addresses its customers informally. The styleguide has 44 pages, from tone of voice and the grid to the chat widget and the email signature.