Introduction
Whether a new person joins the organisation, a new user joins your platform or a new customer starts using your product – all these situations share some kind of onboarding step. This is a necessary, but sometimes painful step. Some companies go manually through this process, some use several different scripts and automation elements, while others use dedicated pieces of software. We are advocates of the 3rd approach in the Botprise, more specifically – usage of bots that helps with the onboarding process.
Problem description
In this post we describe current user onboarding on the Botprise, specifically from the click on “Register User” button to full access to the platform. This process looks pretty similar across many organisations.
Each user is tied to an organisation. This gives rise to two different scenarios of user verification – if a user represents a new organisation or the existing one. If organisation exists – organisation admin’s verification is required. Also, registration links should be deactivated on expiration. To avoid managing a lot of different automation units in different places, we created a bot on our platform and called it BotpriseOnboardingSupportBot.
Bot description
At the moment of writing this post, Bot looks as follows. As the user clicks on register button, ServiceNow ticket is raised. ServiceNow ticket triggers the workflow associated with the Bot. Workflows are built in standard Business Process Model and Notation (BPMN). Shortly steps are:
- Bot processes SNOW ticket
- Bot sends registration request to the admin of organisation or create the new one
- If final steps are not done by admin or user, notification email is sent
- Depending on the previous steps bot activate account or close registration process
Self-healing capability of Bot
While the user onboarding process is not fundamentally dependent on AI, with the Botprise Workflows it is possible to add machine learning pieces to the whole process. We optimized the onboarding workflow described above by an ml algorithm which learns when better to send emails. Gaussian Process Classifier was used to model the distribution of response probability of organization admin based on time elapsed from registration start. Output of the algorithm is the best time when to send email based on historical response data of this account. Upon onboarding completion whether it completed successfully or not, the algorithm is updated with new data point.
Take it as a blueprint
The strong side of the platform is the ability to take existing workflows as blueprints and based on them develop new. It is possible to add new AU and DU blocks to existing workflow or to change existing ones.
How to create workflow?
Workflows are created in drag-in-drop Workflow Studio (another piece of platform). Machine learning algorithms are packed as Decision Units. User could use pre trained Decision Units, train existing algorithms on user’s data or provide own Python class that implements fit(), predict() APIs (most ml libraries are supported on Botprise).
Chatbot integration
It is also possible to involve chat bot in the onboarding process. Botprise platform provides Support ChatBot which is able to handle several tasks, such as registration, password reset, support ticket submission and other. User could simply pin the ChatBot and bot would trigger BotpriseOnboardingSupportBot based on the user’s slack information. Our team plans to introduce Chatbot Design Studio in Q3 2021 with capabilities to build custom Chatbots with specific responses and actions for different users intents.
Final thoughts
This post is about automating the user onboarding process with bots. We’ve seen how to do it with Botprise platform and how to add AI elements to workflows. Stay tuned for new posts and if you have any questions please contact us at support@botprise.com !