Custom AI Chatbot Builder
A no-code platform based on artificial intelligence and NLP technology.
Client
In 2016, a business owner came up with an idea to create a chatbot builder: an easy-to-use business tool that didn’t even require an in-house programmer to set up. The goal was to release a free version of the product, capture a niche, and monetize the platform by introducing paid features (for example, templates with a premium subscription or integrations with third-party services).
The idea looked promising: at that time, similar platforms were just gaining momentum. We evaluated the concept and the scope, estimated potential difficulties, and took on the development.
Challenges
The first challenge was speed: it was important to release the MVP faster than competitors. In the subsequent stages, it was essential to promptly add new features to enhance the product.
We already had significant developments in the field of data science and machine learning, but this was the first time such a product had been implemented.
It was necessary to quickly create a powerful AI-based platform solution.
1. Realistic Dialogue Interaction
In 2016, interactive voice response (IVR) was used to automate communication. However, IVR, with its rigidly defined response options and hierarchical routing, was not an ideal tool.
Users came up with specific questions but ended up on IVR. In 90% of cases, they received a multi-level menu that they had to listen to. Potential buyers hate playing “waiting games”—they hang up and never contact your company again.
2. Omnichannel
The few competitors at the time had a clear inconvenience: a separate bot was required for each communication channel.
Another challenge we faced was achieving omnichannel functionality. We needed to create a single bot that would operate seamlessly across all platforms.
How we solved the project’s challenges:
We used NLP
We decided to choose a Natural Language Processing (NLP) system. It uses machine learning algorithms for text and speech. It is NLP that allows chat robots to go beyond predefined scripts and make communication realistic, similar to a live operator.
We added popular channels:
— a button widget on the website;
— social networks and messengers (Facebook, Skype, Slack, Line, Viber);
— email;
— SMS messages;
— Zendesk web service for customer support.
Users create a bot in one channel, and all changes are automatically pulled into other channels. Even when multiple channels are utilized, all conversations are consolidated in a single location.
Enriched functionality
We enriched the NLP engine with many functions aimed at solving a wide variety of communication tasks. Ideas for functionality were developed together with the business owner, as well as based on user feedback. It is a classic MVP approach, which in our case proved to be very useful.
What we added:
— speech recognition and synthesis;
— intent recognition;
— extraction and transformation of user data;
— advanced analytics;
— media processing.
The chatbot’s functionality is ultimately constrained by the capabilities of the platform it is integrated into, whether that be a website, a social network, or a messaging app.
Implemented several communication scenarios
Bots created on the platform should not be limited to just a text scenario or a call; it is necessary to expand the number of interactions.
Users enter a query in any convenient way: type a message, make a call, or send a screenshot or photo. The bot “interprets” the query and provides the most accurate answer in a second.
Built-in customization capability
It’s a known fact: a personalized solution can only be obtained through custom development. However, the user interfaces of chatbots created on the platform should have the option of budget-friendly customization.
Users create the scenario themselves: greeting, trigger for starting a chain, and responses. In addition to text, cards with images, selection buttons, quick replies, and media files can be added. Widgets are fully customizable down to corporate identity elements. A template store has been created for full individualization.
Technologies
Backend
Go
Backend
Python
Frontend
Angular 8
Infrastructure
AWS
Infrastructure
Kubernetes
Infrastructure
Docker
RDBMS
PostgreSQL
DBMS
ClickHouse
Library
NLTK
Library
TensorFlow
External APIs
Social networks, Messengers
Result
The idea turned out to be excellent: creating a bot takes minimal time and requires neither special skills nor investments. The capabilities of such a constructor, combined with its simplicity, are suitable for businesses of any scale.
The numbers speak for themselves. Today, the platform’s reach is five billion people, including installations on websites, email clients, and SMS mailings.
The product is still active in the support phase and brings stable profit to the owner. Small and medium-sized businesses use the free version; large firms that strive for uniqueness contact the platform owner for paid customization.
A little later, at the owner’s request, we developed a messenger, which we added to the main product.
Services and communication channels were integrated into the product.
Were needed for an accelerated MVP product launch.
Installations (including websites, email clients, and SMS distributions).
What happens next:
Having received and processed your request, we will reach you shortly to detail your project needs.
After examining requirements, our analysts and developers devise a project proposal with the scope of works, team size, time and cost estimates.
We arrange a meeting with you to discuss the offer and come to an agreement.
We sign a contract and start working on your project as quickly as possible.