We automate customer service using AI.

There are many general opinions about what chatbots can offer but it is very hard to stumble across any specific information. That is why we decided to describe in detail how we create chatbots in K2.

1. General concept of Bot

At the beginning we create the idea of the bot and its widely understood environment, i.e. the whole organizational, technological and legal surrounding of the bot. This includes, in particular, a very deep analysis of business needs, challenges and goals in organizations and of processes that can be automated by the bot in part or in whole.

That way we have a very clear list of scenarios how the chatbot can be used, which is strictly tailored for the needs of users and organizations. After setting priority of available scenarios we define the MPV and device short-term and long-term plan for the bot development.

Depending on selected scenarios, we recommend specific technological and infrastructural solutions to build the chatbot of desired functionality.

2. Proof of Concept

Preparation of proof-of-concept allows to verify assumptions defined during creation of the bot’s vision, before investing substantial outlays to make it life.

PoC covers usually 1-3 scenarios that enable testing the key functions of the chatbot or feasibility verification of the most demanding integrations.

PoC covers closed group of users and conclusions thereof are used to calibrate, correct and update assumptions before creating final solution.

3. Soft Launch

Chatbots with NLP need so called „test data”, i.e. contact with real-life users to achieve optimal and stable quality of conversational skills.

To do so, we extend the “beta” bot to small and controlled recipients form the target group and the Quality Assurance team monitors chatbot’s conversations, verifies the efficiency of created scenarios of conversation and continues works to optimize configuration.

4. Public Launch

The soft launch stage ends after achieving expected quality indicators, to include among other things, the percentage of questions and orders that the bot understood correctly. Only when we are satisfied by the result, the green light for public launch is given.

In the initial stage of the public launch (approximately 1-3 months) the team monitors and optimizes chatbot’s operation with the same intensity.

Be assured that the bot’s development process is not finished yet. The users report new needs and requirements, thus stimulating constant evolution of chatbot’s functionalities and development of its knowledge database.

Read also:

If you want to discuss how to automate customer service with help of bots — feel free to contact us.