Multilingual AI-Powered Vacation Chatbot for a SharePoint Intranet

Client

The Client is an international professional services company headquartered in Texas, US. The company provides their services to enterprise and middle-market customers across multiple business domains and industries, including healthcare, banking, retail, manufacturing, telecom and more.

Challenge

The Client runs an on-premises SharePoint intranet that hosts several departmental sites (HR, Marketing, Sales, ICT, etc.), business solutions (quality management, document management, etc.) and ensures enterprise-wide collaboration. Employees use the intranet actively to do daily tasks, collaborate with colleagues, manage documents and their personal information, as well as track their working time and vacations.

However, vacation scheduling was organized in a very inconvenient way. To check up available vacation time or book it, employees had to:

• Perform multiple actions on the intranet.
• Consult with an HR manager.
• Communicate with a line or a project manager to get their approval.

The company wanted to facilitate the process and minimize employees’ effort. To find a suitable solution, the Client turned to ScienceSoft.

Solution

After having analyzed the existing process of vacation booking on the Client’s intranet, ScienceSoft offered to develop a dedicated intelligent vacation chatbot. The chatbot was to assist employees in checking and booking vacations, thus replacing multistage human interaction completely.

Bot training

To ensure human-like conversation, ScienceSoft’s developers tried out several bot training services and chose Dialogflow (former API.ai) as the most suitable one. To teach the bot to support natural communication, the development team used a vast set of conversational phrases, including:

• Intent phrases: welcome, capability, bookVacation, showVacation, deleteVacation, cancel, balance, exit.
• Default response phrases available in Dialogflow.
• Custom response phrases to ensure adequate replies in case of complex or unexpected intent phrases.

Bot capabilities

The bot uses AI to understand users and respond like a living person. SharePoint developers integrated it with the SharePoint intranet to allow the bot to take data from the calendar and display relevant vacation periods. The bot has the following capabilities:

• Displaying available vacation days
• Showing a booked vacation
• Booking new vacations
• Deleting vacations

If an employee decides to book a vacation, he or she chooses the relevant command in the chatbot window. After that, the vacation request goes directly to a line or a project manager who can approve or decline it. When any changes are required, an employee can restart the communication with the bot to choose another vacation period.

In the framework of the chatbot project, ScienceSoft enabled the bot integration with the most used messengers in the company that are Skype and Telegram, thus making the bot even more helpful for employees. This way, employees can consult with the bot even outside the intranet.

Pic 1. Conversation with the Vacation Bot in the SharePoint intranet.

Pic 2. Conversation with the Vacation Bot in Skype.

Pic 3. Conversation with the Vacation Bot in Telegram. Desktop and mobile views.

The bot supports communication in several languages to cover the needs of the company’s local offices.

Results

ScienceSoft has successfully developed, trained, implemented and tested the vacation chatbot. Employees can access the bot directly from their intranet homepage, as well as from desktop and mobile messengers (Skype, Telegram). The bot facilitated the vacation process substantially. Now employees can check and book their vacation via a simple online conversation with the bot.

ScienceSoft’s developers keep collaborating with the Client and collecting end user feedback to detect the bot’s unnatural behavior and retrain it to ensure an excellent user experience.

Technologies and Tools

Microsoft SharePoint 2013, .NET, C#, Microsoft Bot Framework, ASP.NET Web API, Google Dialogflow, Visual Studio 2017.